CN104965980A - Outlier eliminating method and apparatus - Google Patents

Abstract

The invention discloses a kind of method of abnormal value removing and correction and devices, belong to intelligent transportation field. The described method includes: obtaining measuring value Xi (k) of i-th of sensor at current time to obstructing objects in n sensor being arranged on vehicle; The current time of i-th of sensor is filtered the measuring value Xi (k) of obstructing objects to obtain estimated value of the current time to obstructing objects of i-th of sensor The predicted value at current time is obtained to the estimated value of target obstacle according to the last moment of each sensor in the n sensor Based on the measuring value Xi (k), estimated value And predicted value It determines test value λ (k), judges whether the test value λ (k) is less than preset value TD; If the test value λ (k) is not less than the preset value TD, it is determined that the current time of corresponding i-th of the sensor of the test value λ (k) is outlier to the measuring value Xi (k) of obstructing objects, and by the unruly-value rejecting. A kind of method of abnormal value removing and correction calculation amount disclosed by the invention is small, and computation complexity is low.

Description

Method of abnormal value removing and correction and device

Technical field

The present invention relates to intelligent transportation field, particularly a kind of method of abnormal value removing and correction and device.

Background technology

In order to the road environment in monitor vehicle front, multiple sensor such as vision sensor and radar sensor is generally provided with in vehicle, wherein vision sensor can identify the target object that vehicle front occurs, and obtains the measurement information such as orientation and translational speed of this target object; Radar sensor can obtain the translational speed of target object that vehicle front occurs, apart from the measurement information such as distance of this car.In order to better assess the road environment that vehicle travels, need to merge the measurement information of multiple sensor such as this vision sensor and radar sensor etc., the measurement information namely obtained the plurality of sensor carries out complementation and optimal combination to generate more reliable more accurate measurement information.

In correlation technique, when the measurement information obtained multiple sensor merges, get each sensor to the measurement information of target object after, due to exist in this measurement information some depart from most of data present the fraction information of trend, these information are called as outlier, therefore need detect these outliers and reject, improve the accuracy rate of the information that sensor obtains with this.The detection method of current outlier mainly comprises fitting of a polynomial method of prediction, Differential Detection method and likelihood ratio test method etc.

Realizing in process of the present invention, inventor finds that prior art at least exists following problem:

Containing much information of the measurement information that vision sensor and radar sensor obtain, the complexity of information is higher, when adopting the outlier detection algorithm in correlation technique detect outlier and reject outlier, the calculated amount of algorithm is comparatively large, and computation complexity is higher.

Summary of the invention

In order to solve the problem of prior art, embodiments provide a kind of method of abnormal value removing and correction and device.Described technical scheme is as follows:

On the one hand, provide a kind of method of abnormal value removing and correction, described method comprises:

N the sensor that acquisition vehicle is arranged is at measuring value X (k) of current time to obstructing objects, and a described n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k), described i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n;

To the current time of described i-th sensor to the measuring value X of obstructing objects _ik () is carried out filtering and is obtained the current time of described i-th sensor to the estimated value of obstructing objects

The predicted value of current time is obtained according to the upper estimated value of a moment to target obstacle of each sensor in a described n sensor a described upper moment differs t with described current time, and described t is greater than 0, and described target obstacle is the target object that described each sensor was determined in a upper moment;

Determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), described filtering error e ₁k () is for the current time of described i-th sensor is to the estimated value of obstructing objects with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), described predicated error e ₂k () is the predicted value of described current time with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k);

According to described filtering error e ₁(k) and described predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k), described test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)；

Wherein, β ^tk () represents the transposition of β (k);

Judge whether described test value λ (k) is less than preset value T _d;

If described test value λ (k) is not less than described preset value T _d, then determine that the current time of i-th sensor that described test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier;

From a described n sensor measuring value X (k) of current time to obstructing objects by described unruly-value rejecting.

Optionally, described from a described n sensor measuring value X (k) of current time to obstructing objects by described unruly-value rejecting after, described method also comprises:

Described n sensor after rejecting outlier is carried out time-space relation at current time to measuring value X (k) of obstructing objects;

To measuring value X (k) of obstructing objects, association process is carried out at current time to described n sensor after time-space relation, obtains the measuring value belonging to target object;

The described measuring value belonging to target object is carried out merging to export.

Optionally, described at measuring value X (k) of current time, association process is carried out to described n sensor after time-space relation, comprising:

Obtain described n sensor after described time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

According to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object;

If there is obstructing objects is described target object, from described n sensor after described time-space relation current time to measuring value X (k) of obstructing objects obtain and belong to the measuring value of described target object.

Optionally, described according to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object, comprising:

Obtain each sensor in a described n sensor in the measuring value to obstructing objects of current time and estimated value;

Determine the estimated value probability of described each sensor in the estimated value to obstructing objects of current time, described estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, for described i-th sensor is in the estimated value to obstructing objects of current time, A _ifor the estimated value probability of described i-th sensor;

Determine the measuring value probability of described each sensor at the measuring value to obstructing objects of current time, described measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _i(k) for described i-th sensor is at the measuring value to obstructing objects of current time, B _ifor the measuring value probability of described i-th sensor;

Estimated value probability based on described each sensor determines that each obstructing objects of a described n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\∩...\∩u_k^i...\∩u_k^n=u_j}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\∩...\∩u_k^i...u_k^n=\mathrm{\φ}}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that described i-th sensor exists the measuring value to described each obstructing objects at current time, represent that described i-th sensor does not exist the measuring value to described each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to described each obstructing objects, then if there is not the measuring value to described each obstructing objects at current time in i-th sensor, then

Be the probability m (u of target object by described each obstructing objects ₁) and described each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m;

Judge whether described interactive calculation value m is greater than the second preset value ε;

If described interactive calculation value m is greater than described second preset value ε, judge in a described n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ;

If the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ, then there is obstructing objects is described target object, and described target object is the barrier that the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ.

Optionally, the described upper estimated value of a moment to target obstacle according to each sensor in a described n sensor obtains the predicted value of current time comprise:

Obtained a upper moment of described each sensor to the estimated value of target obstacle;

Based on upper moment of described each sensor to the described predicted value of the estimated value determination current time of target obstacle described predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\μ}}_i{\hat{X}}_i(k-1);$

Wherein, for a upper moment of described i-th sensor is to the estimated value of target obstacle, μ _ifor the predicted value coefficient of described i-th sensor, and

Optionally, described the described measuring value belonging to target object is carried out merging export, comprising:

Obtain the mean value belonging to the measuring value of described target object;

Described mean value is exported.

On the other hand, provide a kind of unruly-value rejecting device, described device comprises:

First acquiring unit, for obtaining n sensor that vehicle is arranged at measuring value X (k) of current time to obstructing objects, a described n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k), described i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n;

Filter unit, for the current time of described i-th sensor to the measuring value X of obstructing objects _ik () is carried out filtering and is obtained the current time of described i-th sensor to the estimated value of obstructing objects

Second acquisition unit, for obtaining the predicted value of current time according to the upper estimated value of a moment to target obstacle of each sensor in a described n sensor a described upper moment differs t with described current time, and described t is greater than 0, and described target obstacle is the target object that described each sensor was determined in a upper moment;

First determining unit, for determining filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), described filtering error e ₁k () is for the current time of described i-th sensor is to the estimated value of obstructing objects with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), described predicated error e ₂k () is the predicted value of described current time with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k);

Second determining unit, for according to described filtering error e ₁(k) and described predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k), described test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)；

Wherein, β ^tk () represents the transposition of β (k);

Judging unit, for judging whether described test value λ (k) is less than preset value T _d;

3rd determining unit, for being not less than described preset value T described test value λ (k) _dtime, determine that the current time of i-th sensor that described test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier;

Culling unit, for from a described n sensor in measuring value X (k) of current time to obstructing objects by described unruly-value rejecting.

Optionally, described device also comprises:

Registration unit, for carrying out time-space relation at current time to measuring value X (k) of obstructing objects by described n sensor after rejecting outlier;

Association process unit, for carrying out association process at current time to measuring value X (k) of obstructing objects to described n sensor after time-space relation, obtains the measuring value belonging to target object;

Output unit, exports for the described measuring value belonging to target object being carried out merging.

Optionally, described association process unit comprises:

First acquisition module, for obtaining described n sensor after described time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

Judge module, for according to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object;

Second acquisition module, for when there is obstructing objects and being described target object, from described n sensor after described time-space relation current time to measuring value X (k) of obstructing objects obtain and belong to the measuring value of described target object.

Optionally, described judge module also for:

Obtain each sensor in a described n sensor in the measuring value to obstructing objects of current time and estimated value;

Determine the estimated value probability of described each sensor in the estimated value to obstructing objects of current time, described estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, for described i-th sensor is in the estimated value to obstructing objects of current time, A _ifor the estimated value probability of described i-th sensor;

Determine the measuring value probability of described each sensor at the measuring value to obstructing objects of current time, described measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _i(k) for described i-th sensor is at the measuring value to obstructing objects of current time, B _ifor the measuring value probability of described i-th sensor;

Estimated value probability based on described each sensor determines that each obstructing objects of a described n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\∩...\∩u_k^i...\∩u_k^n=u_j}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\∩...\∩u_k^i...u_k^n=\mathrm{\φ}}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that described i-th sensor exists the measuring value to described each obstructing objects at current time, represent that described i-th sensor does not exist the measuring value to described each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to described each obstructing objects, then if there is not the measuring value to described each obstructing objects at current time in i-th sensor, then

Be the probability m (u of target object by described each obstructing objects ₁) and described each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m;

Judge whether described interactive calculation value m is greater than the second preset value ε;

If described interactive calculation value m is greater than described second preset value ε, judge in a described n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ;

If the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ, then there is obstructing objects is described target object, and described target object is the barrier that the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ.

Optionally, described second acquisition unit comprises:

3rd acquisition module, for obtaining a upper moment of described each sensor to the estimated value of target obstacle;

Determination module, for the upper moment based on described each sensor to the described predicted value of the estimated value determination current time of target obstacle described predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\μ}}_i{\hat{X}}_i(k-1);$

Wherein, for a upper moment of described i-th sensor is to the estimated value of target obstacle, μ _ifor the predicted value coefficient of described i-th sensor, and

Optionally, described output unit comprises:

4th acquisition module, for obtaining the mean value of the measuring value belonging to described target object;

Output module, for exporting described mean value.

The beneficial effect that the technical scheme that the embodiment of the present invention provides is brought is:

A kind of method of abnormal value removing and correction that the embodiment of the present invention provides and device, by obtain n sensor that vehicle is arranged at current time to measuring value X (k) of obstructing objects and i-th sensor at the measuring value X of current time to obstructing objects _i(k), estimated value and predicted value determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), and determine test value λ (k) according to this residual error β (k) and covariance W (k), finally will be less than preset value T _dthe current time of i-th sensor corresponding to test value λ (k) to the measuring value X of obstructing objects _ik () is defined as outlier, and from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.The calculated amount when method of abnormal value removing and correction that the embodiment of the present invention provides and device rejecting outlier is little, and computation complexity is low.

Accompanying drawing explanation

In order to be illustrated more clearly in the technical scheme in the embodiment of the present invention, below the accompanying drawing used required in describing embodiment is briefly described, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.

Fig. 1 is the process flow diagram of a kind of method of abnormal value removing and correction that the embodiment of the present invention provides;

Fig. 2 is the process flow diagram of the another kind of method of abnormal value removing and correction that the embodiment of the present invention provides;

Fig. 3 is the process flow diagram of another method of abnormal value removing and correction that the embodiment of the present invention provides;

Fig. 4 is the process flow diagram of another method of abnormal value removing and correction that the embodiment of the present invention provides;

Fig. 5 is the structural representation of a kind of unruly-value rejecting device that the embodiment of the present invention provides;

Fig. 6 is the structural representation of the another kind of unruly-value rejecting device that the embodiment of the present invention provides;

Fig. 7 is the structural representation of the association process unit that the embodiment of the present invention provides;

Fig. 8 is the structural representation of the second acquisition unit that the embodiment of the present invention provides;

Fig. 9 is the structural representation of the output unit that the embodiment of the present invention provides.

Embodiment

For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, embodiment of the present invention is described further in detail.

The embodiment of the present invention provides a kind of method of abnormal value removing and correction, and as shown in Figure 1, the method comprises:

N the sensor that step 101, acquisition vehicle are arranged is at measuring value X (k) of current time to obstructing objects, and this n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k).Perform step 102.

This i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n.

Step 102, to the current time of i-th sensor to the measuring value X of obstructing objects _ik () carries out the estimated value of current time to obstructing objects that filtering obtains i-th sensor perform step 103.

Step 103, obtain the predicted value of current time according to the upper estimated value of a moment to target obstacle of each sensor in this n sensor on this, a moment differs t with current time, and this t is greater than 0, and target obstacle is the target object that each sensor was determined in a upper moment.Perform step 104.

Step 104, determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k).

This filtering error e ₁k () is the estimated value of current time to obstructing objects of i-th sensor with the current time of i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), predicated error e ₂k predicted value that () is current time with the current time of i-th sensor to the measuring value X of obstructing objects _ithe difference of (k).Perform step 105.

Step 105, according to filtering error e ₁(k) and predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k).Perform step 106.

This test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)。

Step 106, judge whether test value λ (k) is less than preset value T _dif test value λ (k) is not less than preset value T _d, perform step 107; If test value λ (k) is less than preset value T _d, perform step 108.

Step 107, determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier.Perform step 109.

Step 108, determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is not outlier.

Step 109, from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.

In sum, a kind of method of abnormal value removing and correction that the embodiment of the present invention provides, the method by obtain n sensor that vehicle is arranged at current time to measuring value X (k) of obstructing objects and i-th sensor at the measuring value X of current time to obstructing objects _i(k), estimated value and predicted value determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), and determine test value λ (k) according to this residual error β (k) and covariance W (k), finally will be less than preset value T _dthe current time of i-th sensor corresponding to test value λ (k) to the measuring value X of obstructing objects _ik () is defined as outlier, and from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.The method of abnormal value removing and correction calculated amount that the embodiment of the present invention provides is little, and computation complexity is low.

Optionally, after step 109, the method also comprises:

N sensor after rejecting outlier is carried out time-space relation at current time to measuring value X (k) of obstructing objects.

To measuring value X (k) of obstructing objects, association process is carried out at current time to the sensor of the n after time-space relation, obtains belonging to the measuring value of target object to obstructing objects.

By belonging to target object, merging output is carried out to the measuring value of obstructing objects.

Optionally, this carries out association process to the sensor of the n after time-space relation at measuring value X (k) of current time, comprising:

Obtain n sensor after time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

According to n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is target object.

If there is obstructing objects is target object, from the sensor of the n after time-space relation current time to measuring value X (k) of obstructing objects obtain and belong to the measuring value of target object.

Optionally, according to this n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is target object, comprising:

Obtain each sensor in this n sensor in the measuring value to obstructing objects of current time and estimated value;

Determine the estimated value probability of each sensor in the estimated value to obstructing objects of current time, this estimated value probability A _ifor:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, be the estimated value to obstructing objects of i-th sensor at current time, A _iit is the estimated value probability of i-th sensor;

Determine the measuring value probability of each sensor at the measuring value to obstructing objects of current time, this measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\Σ}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _ik () is the measuring value to obstructing objects of i-th sensor at current time, B _iit is the measuring value probability of i-th sensor.

Estimated value probability based on this each sensor determines that each obstructing objects of n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\∩...\∩u_k^i...\∩u_k^n=u_j}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\∩...\∩u_k^i...u_k^n=\mathrm{\φ}}{\mathrm{\Σ}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that i-th sensor exists the measuring value to each obstructing objects at current time, represent that i-th sensor does not exist the measuring value to each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to each obstructing objects, then if there is not the measuring value to each obstructing objects at current time in i-th sensor, then $m_i\left(u_2^i\right)=1-A_i.$

Be the probability m (u of target object by each obstructing objects ₁) and each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m.

Judge whether interactive calculation value m is greater than the second preset value ε.

If interactive calculation value m is greater than the second preset value ε, judge in this n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ.

If the difference of the estimated value probability of this each sensor and measuring value probability is all less than the 3rd preset value τ, then there is obstructing objects is target object, and target object is the barrier that the estimated value probability of each sensor and the difference of measuring value probability are all less than the 3rd preset value τ.

Optionally, the predicted value of current time is obtained according to the upper estimated value of a moment to target obstacle of each sensor in n sensor comprise:

Obtained a upper moment of each sensor to the estimated value of target obstacle.

Based on upper moment of each sensor to the predicted value of the estimated value determination current time of target obstacle predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\μ}}_i{\hat{X}}_i(k-1);$

Wherein, be a upper moment of i-th sensor to the estimated value of target obstacle, μ _ibe the predicted value coefficient of i-th sensor, and

Optionally, the measuring value belonging to target object is carried out merging and exports, comprising:

Obtain the mean value belonging to the measuring value of target object;

Mean value is exported.

In sum, a kind of method of abnormal value removing and correction that the embodiment of the present invention provides, the method by obtain n sensor that vehicle is arranged at current time to measuring value X (k) of obstructing objects and i-th sensor at the measuring value X of current time to obstructing objects _i(k), estimated value and predicted value determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), and determine test value λ (k) according to this residual error β (k) and covariance W (k), finally will be less than preset value T _dthe current time of i-th sensor corresponding to test value λ (k) to the measuring value X of obstructing objects _ik () is defined as outlier, and from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.The method of abnormal value removing and correction calculated amount that the embodiment of the present invention provides is little, and computation complexity is low.

The embodiment of the present invention provides another kind of method of abnormal value removing and correction, and as shown in Figure 2, the method comprises:

N the sensor that step 201, acquisition vehicle are arranged is at measuring value X (k) of current time to obstructing objects, and this n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k).Perform step 202.

This i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n, i.e. 0 < i≤n.

In embodiments of the present invention, n sensor vehicle arranged can comprise radar sensor, laser sensor and vision sensor etc.This n sensor can continue the barrier detecting vehicle front, and obtains measuring value X (k) to this barrier, and this measuring value X (k) comprises the speed of the orientation of this barrier, the Distance geometry movement of distance vehicle.

Step 202, to the current time of i-th sensor to the measuring value X of obstructing objects _ik () carries out the estimated value of current time to obstructing objects that filtering obtains i-th sensor perform step 203.

Example, to measuring value X _ik operation that () carries out filtering can be Kalman filtering, and the detailed process of Kalman filtering can with reference to correlation technique, and the embodiment of the present invention does not repeat.

Step 203, obtain the predicted value of current time according to the upper estimated value of a moment to target obstacle of each sensor in this n sensor on this, a moment differs t with current time, and this t is greater than 0, and target obstacle is the target object that each sensor was determined in a upper moment.Perform step 204.

Concrete, as shown in Figure 3, step 203 comprises:

Step 2031, obtained a upper moment of each sensor to the estimated value of target obstacle.

Step 2032, based on upper moment of each sensor to the predicted value of the estimated value determination current time of target obstacle

This predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&mu;}}_i{\hat{X}}_i(k-1);$

Wherein, be a upper moment of i-th sensor to the estimated value of target obstacle, μ _ibe the predicted value coefficient of i-th sensor, and

Example, the predicted value coefficient μ of this i-th sensor _imeet: 0≤μ _i≤ 1, in actual applications, suppose that number of probes n equals 2, then the predicted value coefficient μ of first sensor ₁with the predicted value coefficient μ of second sensor ₂can be: μ ₁=μ ₂=0.5.

Step 204, determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k).Perform step 205.

This filtering error e ₁k () is the estimated value of current time to obstructing objects of i-th sensor with the current time of i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), predicated error e ₂k predicted value that () is current time with the current time of i-th sensor to the measuring value X of obstructing objects _ithe difference of (k).

Example, filtering error e ₁(k) be:

$e_1\left(k\right)={\hat{X}}_i\left(k\right)-X_i\left(k\right);$

Predicated error e ₂(k) be:

$e_2\left(k\right)=\tilde{X}\left(k\right)-X_i\left(k\right);$

This filtering error e ₁(k) and predicated error e ₂k the residual error β (k) of () is:

$\mathrm{\&beta;}\left(k\right)=e_1\left(k\right)-e_2\left(k\right)={\hat{X}}_i\left(k\right)-{\tilde{X}}_i\left(k\right);$

Filtering error e ₁(k) and predicated error e ₂k the covariance W (k) of () is:

W(k)＝E{β(k)β ^T(k)}；

Wherein, β ^tk () represents the transposition of β (k).

Step 205, according to filtering error e ₁(k) and predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k).Perform step 206.

Test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)。

Step 206, judge whether test value λ (k) is less than preset value T _d.If test value λ (k) is not less than preset value T _d, perform step 207; If test value λ (k) is less than preset value T _d, perform step 208.

Step 207, determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier.Perform step 209.

Step 208, determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is not outlier.

In embodiments of the present invention, detected value λ (k) can reflect that the current time of i-th sensor that this detected value λ (k) is corresponding is to the measuring value X of obstructing objects _ik the error in measurement of (), therefore when this detected value λ (k) is greater than preset value T _dtime, the measuring value X of the current time of i-th sensor to obstructing objects can be confirmed _ik the error in measurement of () is comparatively large, therefore can determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier; Same, when test value λ (k) is less than preset value T _dtime, can determine that the current time of i-th sensor that test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is not outlier.

Step 209, from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.Perform step 210.

Example, due to the outlier in measuring value for depart from substantial amount measured value present the fraction measuring value of trend, therefore from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting after, this n sensor can be improved in the degree of accuracy of current time to measuring value X (k) of obstructing objects.

Step 210, at current time, time-space relation is carried out to measuring value X (k) of obstructing objects by rejecting n sensor after outlier.Perform step 211.

Example, the time interval of measuring obstructing objects due to this n sensor is different, therefore n sensor is needed to be synchronized to the same moment at measuring value X (k) of current time to obstructing objects, namely carry out temporal registration at current time to measuring value X (k) of obstructing objects to n sensor, the method for temporal registration mainly comprises timeslice technology and synchronized sampling plays point methods etc.; Again due to coordinate system that the measuring value of each sensor to obstructing objects is based on this sensor self, therefore to need n sensor at current time measuring value X (k) of obstructing objects by the method migration such as rotation of coordinate and translation in same coordinate system, namely carry out spatial registration at current time to measuring value X (k) of obstructing objects to n sensor, this same coordinate system can be bodywork reference frame.After time-space relation is carried out to measuring value X (k) of this n sensor, just can better carry out association process to measuring value X (k) of this n sensor.The detailed process of above-mentioned temporal registration and spatial registration can with reference to correlation technique, and the embodiment of the present invention does not repeat.

Step 211, to measuring value X (k) of obstructing objects, association process is carried out at current time to the sensor of the n after time-space relation, obtain belonging to the measuring value of target object to obstructing objects.Perform step 212.

In embodiments of the present invention, in order to better assess the road environment that vehicle travels, need to carry out association process at current time to the measuring value of obstructing objects to the sensor of the n after this time-space relation, and then improve accuracy and the reliability of the information that this n sensor obtains.

Concrete, as shown in Figure 4, step 211 can comprise:

Step 2111, obtain n sensor after time-space relation in measuring value X (k) to obstructing objects of current time and estimated value and in this n sensor each sensor in the measuring value to obstructing objects of current time and estimated value.Perform step 2112.

Step 2112, determine the estimated value probability of each sensor in the estimated value to obstructing objects of current time, and each sensor is at the measuring value probability of the measuring value to obstructing objects of current time.Perform step 2113.

Example, this estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, be the estimated value to obstructing objects of i-th sensor at current time, A _iit is the estimated value probability of i-th sensor;

This measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _ik () is the estimated value to obstructing objects of i-th sensor at current time, B _iit is the estimated value probability of i-th sensor.

Step 2113, determine that each obstructing objects of n sensor measurement is the probability m (u of target object based on the estimated value probability of each sensor ₁) and each obstructing objects be not the probability m (u of target object ₂).Perform step 2114.

Example, j=1 or 2, m (u _j) can represent with following formula (1):

$m\left(u_j\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_j}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}---\left(1\right)$

Concrete, during j=1, m (u ₁) can be expressed as:

$m\left(u_1\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_1}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

During j=2, m (u ₂) can be expressed as:

$m\left(u_2\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_2}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, example, and A ∩ B represents the common factor getting A and B, and φ represents empty set, $u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_j$ Represent common factor be u _j, $u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=\mathrm{\&phi;}$ Represent common factor be empty set. represent that i-th sensor exists the measuring value to each obstructing objects at current time, represent that i-th sensor does not exist the measuring value to each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to each obstructing objects, then if there is not the measuring value to each obstructing objects at current time in i-th sensor, then

Example, suppose the number n=2 of the sensor arranged in vehicle, wherein the estimated value probability A of first sensor ₁the estimated value probability A of=0.9, second sensor ₂=0.8, if first sensor is at the measuring value of current time existence to each obstructing objects, then if there is not the measuring value to each obstructing objects at current time in first sensor, then it is in like manner known, $m_2\left(u_1^2\right)=A_2=0.8,m_2\left(u_2^2\right)=1-A_2=1-0.8=0.2.$ Each obstructing objects that can obtain these two sensor measurements according to above-mentioned formula (1) is the probability m (u of target object ₁) be:

$m\left(u_1\right)=\frac{\underset{u_k^1\&cap;u_k^2=u_1}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)}{1-\underset{u_k^1\&cap;u_k^2=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)};$

Wherein, represent that first sensor and second sensor all exist the measuring value to each obstructing objects at current time, meet this condition with be combined as to represent in first sensor and second sensor have at least a sensor to there is not the measuring value to each obstructing objects at current time, meets this condition with be combined as then now each obstructing objects of these two sensor measurements is the probability m (u of target object ₁) can be expressed as:

$m\left(u_1\right)=\frac{m_1\left(u_1^1\right)m_2\left(u_1^2\right)}{1-[m_1\left(u_1^1\right)m_2\left(u_2^2\right)+m_1\left(u_2^1\right)m_2\left(u_1^2\right)+m_1\left(u_2^1\right)m_2\left(u_2^2\right)]};$

Will $m_1\left(u_1^1\right)=0.9,m_1\left(u_2^1\right)=0.1,m_2\left(u_1^2\right)=0.8$ With $m_2\left(u_2^2\right)=0.2$ Bring above-mentioned formula into, can obtain:

$m\left(u_1\right)=\frac{0.9\&times;0.8}{1-[0.9\&times;0.2+0.1\&times;0.8+0.1\&times;0.2]}=1;$

Same, each obstructing objects of these two sensor measurements is not the probability m (u of target object ₂) can be expressed as:

$m\left(u_2\right)=\frac{m_1\left(u_1^1\right)m_2\left(u_2^2\right)+m_1\left(u_2^1\right)m_2\left(u_1^2\right)+m_1\left(u_2^1\right)m_2\left(u_2^2\right)}{1-[m_1\left(u_1^1\right)m_2\left(u_2^2\right)+m_1\left(u_2^1\right)m_2\left(u_1^2\right)+m_1\left(u_2^1\right)m_2\left(u_2^2\right)]};$

Will $m_1\left(u_1^1\right)=0.9,m_1\left(u_2^1\right)=0.1,m_2\left(u_1^2\right)=0.8$ With $m_2\left(u_2^2\right)=0.2$ Bring above-mentioned formula into, can obtain:

$m\left(u_2\right)=\frac{0.9\&times;0.2+0.1\&times;0.8+0.1\&times;0.2}{1-[0.9\&times;0.2+0.1\&times;0.8+0.1\&times;0.2]}=0.39.$

Step 2114, be the probability m (u of target object by each obstructing objects ₁) and each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m.Perform step 2115.

Example, interactive calculation value m is: m=m (u ₁)-m (u ₂).As m (u ₁)=1, m (u ₂during)=0.39, interactive calculation value m=1-0.39=0.61.

Step 2115, judge whether interactive calculation value m is greater than the second preset value ε.If interactive calculation value m is greater than the second preset value ε, perform step 2116; If interactive calculation value m is not more than the second preset value ε, perform step 2118.

In embodiments of the present invention, the span of this second preset value ε can be 0.05 to 1, supposes this second preset value ε=0.05, therefore when interactive calculation value m is 0.61, is greater than this second preset value ε, now performs step 2116.

Step 2116, judge in n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ.If the difference of the estimated value probability of each sensor and measuring value probability is all less than the 3rd preset value τ, perform step 2117; If the estimated value probability of arbitrary sensor and the difference of measuring value probability are not less than the 3rd preset value τ in n sensor, perform step 2118.

Example, because each sensor exists error between the measuring value of obstructing objects and estimated value, therefore when judging whether this obstructing objects is target object, also need to consider the error between the measuring value of each sensor and estimated value, when this error is less than the 3rd preset value τ, just can determine that this obstructing objects is target object.The span of the 3rd preset value τ can be 0.05 to 0.1.In n sensor, the estimated value probability of each sensor and the difference of measuring value probability can be expressed as: Δ _i=A _i-B _i, wherein A _ibe the estimated value probability of i-th sensor, B _ibe the measuring value probability of i-th sensor, Δ _ibe the estimated value probability A of i-th sensor _iwith measuring value probability B _idifference.Suppose that the 3rd preset value τ is 0.05, number of probes n=2, wherein the estimated value probability of first sensor and measuring value probability are respectively: A ₁=0.9, B ₁=0.88; Estimated value probability and the measuring value probability of second sensor are respectively: A ₂=0.8, B ₂=0.77, then the difference of the estimated value probability and measuring value probability that can obtain this first sensor is Δ ₁=A ₁-B ₁=0.02; The estimated value probability of this second sensor and the difference of measuring value probability are Δ ₂=A ₂-B ₂=0.03.Due to Δ ₁, Δ ₂all be less than the 3rd preset value 0.05, now perform step 2117.

Step 2117, determine that there is obstructing objects is target object, and this target object is the barrier that the estimated value probability of each sensor and the difference of measuring value probability are all less than the 3rd preset value τ.

Example, if interactive calculation value m is greater than the second preset value ε, and in n sensor, the estimated value probability of each sensor and the difference of measuring value probability are all less than the 3rd preset value τ, then determine that there is obstructing objects is target object.

Step 2118, determine that there is not obstructing objects is target object.

Example, if interactive calculation value m is not more than the second preset value ε, or the estimated value probability of arbitrary sensor and the difference of measuring value probability are not less than the 3rd preset value τ in n sensor, then determine that there is not obstructing objects is target object, does not now export the measuring value of this obstructing objects.

Step 212, the measuring value belonging to target object carried out merging and export.

Concrete, the measuring value belonging to target object is carried out merging output and comprises:

Obtain the mean value belonging to the measuring value of target object, mean value is exported.

Example, the measuring value belonging to target object in two sensors is respectively X ₁(k) and X ₂(k), then the mean value exported is:

$\frac{X_1\left(k\right)+X_2\left(k\right)}{2}.$

In embodiments of the present invention, owing to determining to exist after obstructing objects is target object, each sensor is all less than the 3rd preset value τ to the estimated value probability of this target object and the difference of measuring value probability, therefore the estimated value of the estimated value of target object can also be carried out merging in step 212 and export.

In sum, a kind of method of abnormal value removing and correction that the embodiment of the present invention provides, the method by obtain n sensor that vehicle is arranged at current time to measuring value X (k) of obstructing objects and i-th sensor at the measuring value X of current time to obstructing objects _i(k), estimated value and predicted value determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), and determine test value λ (k) according to this residual error β (k) and covariance W (k), finally will be less than preset value T _dthe current time of i-th sensor corresponding to test value λ (k) to the measuring value X of obstructing objects _ik () is defined as outlier, and from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.The method of abnormal value removing and correction calculated amount that the embodiment of the present invention provides is little, and computation complexity is low.

Embodiments provide a kind of unruly-value rejecting device 300, see Fig. 5, this device 300 comprises:

First acquiring unit 301, for obtaining n sensor that vehicle is arranged at measuring value X (k) of current time to obstructing objects, this n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _ik (), this i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n.

Filter unit 302, for the current time of this i-th sensor to the measuring value X of obstructing objects _ik () carries out the estimated value of current time to obstructing objects that filtering obtains this i-th sensor

Second acquisition unit 303, for obtaining the predicted value of current time according to the upper estimated value of a moment to target obstacle of each sensor in this n sensor on this, a moment differs t with this current time, and this t is greater than 0, and this target obstacle is the target object that this each sensor was determined in a upper moment.

First determining unit 304, for determining filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), this filtering error e ₁k current time that () is this i-th sensor is to the estimated value of obstructing objects with the current time of this i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), this predicated error e ₂k predicted value that () is this current time with the current time of this i-th sensor to the measuring value X of obstructing objects _ithe difference of (k).

Second determining unit 305, for according to this filtering error e ₁(k) and this predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k), this test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)；

Wherein, β ^tk () represents the transposition of β (k).

Judging unit 306, for judging whether this test value λ (k) is less than preset value T _d.

3rd determining unit 307, for being not less than this preset value T at this test value λ (k) _dtime, determine that the current time of i-th sensor that this test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier.

Culling unit 308, for from this n sensor in measuring value X (k) of current time to obstructing objects by this unruly-value rejecting.

Optionally, as shown in Figure 6, this device 300 also comprises:

Registration unit 309, for carrying out time-space relation at current time to measuring value X (k) of obstructing objects by this n sensor after rejecting outlier.

Association process unit 310, for carrying out association process at current time to measuring value X (k) of obstructing objects to this n sensor after time-space relation, obtains the measuring value belonging to target object.

Output unit 311, exports for the measuring value belonging to target object being carried out merging.

Optionally, as shown in Figure 7, this association process unit 310 comprises:

First acquisition module 3101, for obtaining this n sensor after this time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

Judge module 3102, for according to this n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is this target object.

Second acquisition module 3103, for when there is obstructing objects for this target object, from this n sensor after this time-space relation current time to measuring value X (k) of obstructing objects acquisition belong to the measuring value of this target object.

Optionally, this judge module 3102 also for:

Obtain each sensor in this n sensor in the measuring value to obstructing objects of current time and estimated value.

Determine the estimated value probability of each sensor in the estimated value to obstructing objects of current time, this estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, for this i-th sensor is in the estimated value to obstructing objects of current time, A _ifor the estimated value probability of this i-th sensor.

Determine the measuring value probability of each sensor at the measuring value to obstructing objects of current time, this measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _i(k) for this i-th sensor is at the measuring value to obstructing objects of current time, B _ifor the measuring value probability of this i-th sensor.

Estimated value probability based on each sensor determines that each obstructing objects of this n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_j}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that this i-th sensor exists the measuring value to each obstructing objects at current time, represent that this i-th sensor does not exist the measuring value to each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to each obstructing objects, then if there is not the measuring value to each obstructing objects at current time in i-th sensor, then $m_i\left(u_2^i\right)=1-A_i.$

Be the probability m (u of target object by each obstructing objects ₁) and each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m.

Judge whether this interactive calculation value m is greater than the second preset value ε.

If this interactive calculation value m is greater than the second preset value ε, judge in this n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ.

If the difference of the estimated value probability of each sensor and measuring value probability is all less than the 3rd preset value τ, then there is obstructing objects is this target object, and this target object is the barrier that the estimated value probability of each sensor and the difference of measuring value probability are all less than the 3rd preset value τ.

Optionally, as shown in Figure 8, this second acquisition unit 303 comprises:

3rd acquisition module 3031, for obtaining a upper moment of this each sensor to the estimated value of target obstacle.

Determination module 3032, for the upper moment based on this each sensor to this predicted value of the estimated value determination current time of target obstacle this predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&mu;}}_i{\hat{X}}_i(k-1);$

Wherein, for a upper moment of this i-th sensor is to the estimated value of target obstacle, μ _ifor the predicted value coefficient of this i-th sensor, and

Optionally, as shown in Figure 9, this output unit 311 comprises:

4th acquisition module 3111, for obtaining the mean value of the measuring value belonging to this target object.

Output module 3112, for exporting this mean value.

In sum, a kind of unruly-value rejecting device that the embodiment of the present invention provides, this device by obtain n sensor that vehicle is arranged at current time to measuring value X (k) of obstructing objects and i-th sensor at the measuring value X of current time to obstructing objects _i(k), estimated value and predicted value determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), and determine test value λ (k) according to this residual error β (k) and covariance W (k), finally will be less than preset value T _dthe current time of i-th sensor corresponding to test value λ (k) to the measuring value X of obstructing objects _ik () is defined as outlier, and from n sensor measuring value X (k) of current time to obstructing objects by unruly-value rejecting.Calculated amount when the unruly-value rejecting device that the embodiment of the present invention provides is rejected outlier is little, and computation complexity is low.

Those skilled in the art can be well understood to, and for convenience and simplicity of description, the unit of foregoing description, the specific works process of module, with reference to the corresponding process in preceding method embodiment, can not repeat them here.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (12)

1. a method of abnormal value removing and correction, is characterized in that, described method comprises:

N the sensor that acquisition vehicle is arranged is at measuring value X (k) of current time to obstructing objects, and a described n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k), described i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n;

To the current time of described i-th sensor to the measuring value X of obstructing objects _ik () is carried out filtering and is obtained the current time of described i-th sensor to the estimated value of obstructing objects

The predicted value of current time is obtained according to the upper estimated value of a moment to target obstacle of each sensor in a described n sensor a described upper moment differs t with described current time, and described t is greater than 0, and described target obstacle is the target object that described each sensor was determined in a upper moment;

Determine filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), described filtering error e ₁k () is for the current time of described i-th sensor is to the estimated value of obstructing objects with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), described predicated error e ₂k () is the predicted value of described current time with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k);

According to described filtering error e ₁(k) and described predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k), described test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)；

Wherein, β ^tk () represents the transposition of β (k);

Judge whether described test value λ (k) is less than preset value T _d;

If described test value λ (k) is not less than described preset value T _d, then determine that the current time of i-th sensor that described test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier;

From a described n sensor measuring value X (k) of current time to obstructing objects by described unruly-value rejecting.

2. method according to claim 1, is characterized in that, described from a described n sensor measuring value X (k) of current time to obstructing objects by described unruly-value rejecting after, described method also comprises:

Described n sensor after rejecting outlier is carried out time-space relation at current time to measuring value X (k) of obstructing objects;

To measuring value X (k) of obstructing objects, association process is carried out at current time to described n sensor after time-space relation, obtains the measuring value belonging to target object;

The described measuring value belonging to target object is carried out merging to export.

3. method according to claim 2, is characterized in that, describedly carries out association process to described n sensor after time-space relation at measuring value X (k) of current time, comprising:

Obtain described n sensor after described time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

According to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object;

If there is obstructing objects is described target object, from described n sensor after described time-space relation current time to measuring value X (k) of obstructing objects obtain and belong to the measuring value of described target object.

4. method according to claim 3, is characterized in that, described according to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object, comprising:

Obtain each sensor in a described n sensor in the measuring value to obstructing objects of current time and estimated value;

Determine the estimated value probability of described each sensor in the estimated value to obstructing objects of current time, described estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, for described i-th sensor is in the estimated value to obstructing objects of current time, A _ifor the estimated value probability of described i-th sensor;

Determine the measuring value probability of described each sensor at the measuring value to obstructing objects of current time, described measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _i(k) for described i-th sensor is at the measuring value to obstructing objects of current time, B _ifor the measuring value probability of described i-th sensor;

Estimated value probability based on described each sensor determines that each obstructing objects of a described n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_j}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that described i-th sensor exists the measuring value to described each obstructing objects at current time, represent that described i-th sensor does not exist the measuring value to described each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to described each obstructing objects, then if there is not the measuring value to described each obstructing objects at current time in i-th sensor, then

Be the probability m (u of target object by described each obstructing objects ₁) and described each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m;

Judge whether described interactive calculation value m is greater than the second preset value ε;

If described interactive calculation value m is greater than described second preset value ε, judge in a described n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ;

If the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ, then there is obstructing objects is described target object, and described target object is the barrier that the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ.

5. method according to claim 1, is characterized in that, the described upper estimated value of a moment to target obstacle according to each sensor in a described n sensor obtains the predicted value of current time comprise:

Obtained a upper moment of described each sensor to the estimated value of target obstacle;

Based on upper moment of described each sensor to the described predicted value of the estimated value determination current time of target obstacle described predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&mu;}}_i{\hat{X}}_i(k-1);$

Wherein, for a upper moment of described i-th sensor is to the estimated value of target obstacle, μ _ifor the predicted value coefficient of described i-th sensor, and

6. method according to claim 2, is characterized in that, described the described measuring value belonging to target object is carried out merging export, comprising:

Obtain the mean value belonging to the measuring value of described target object;

Described mean value is exported.

7. a unruly-value rejecting device, is characterized in that, described device comprises:

First acquiring unit, for obtaining n sensor that vehicle is arranged at measuring value X (k) of current time to obstructing objects, a described n sensor comprises i-th sensor at the measuring value X of current time to obstructing objects at measuring value X (k) of current time to obstructing objects _i(k), described i-th sensor is any one in n the sensor that vehicle is arranged, and i, for being greater than 0, is less than or equal to the integer of n;

Filter unit, for the current time of described i-th sensor to the measuring value X of obstructing objects _ik () is carried out filtering and is obtained the current time of described i-th sensor to the estimated value of obstructing objects

Second acquisition unit, for obtaining the predicted value of current time according to the upper estimated value of a moment to target obstacle of each sensor in a described n sensor a described upper moment differs t with described current time, and described t is greater than 0, and described target obstacle is the target object that described each sensor was determined in a upper moment;

First determining unit, for determining filtering error e ₁(k) and predicated error e ₂the residual error β (k) of (k) and covariance W (k), described filtering error e ₁k () is for the current time of described i-th sensor is to the estimated value of obstructing objects with the current time of described i-th sensor to the measuring value X of obstructing objects _ithe difference of (k), described predicated error e ₂the current time of k predicted value X ~ (k) that () is described current time and described i-th sensor is to the measuring value X of obstructing objects _ithe difference of (k);

Second determining unit, for according to described filtering error e ₁(k) and described predicated error e ₂k residual error β (k) and the covariance W (k) of () determine test value λ (k), described test value λ (k) is:

λ(k)＝β ^T(k)W ^-1(k)β(k)；

Wherein, β ^tk () represents the transposition of β (k);

Judging unit, for judging whether described test value λ (k) is less than preset value T _d;

3rd determining unit, for being not less than described preset value T described test value λ (k) _dtime, determine that the current time of i-th sensor that described test value λ (k) is corresponding is to the measuring value X of obstructing objects _ik () is outlier;

Culling unit, for from a described n sensor in measuring value X (k) of current time to obstructing objects by described unruly-value rejecting.

8. device according to claim 7, is characterized in that, described device also comprises:

Registration unit, for carrying out time-space relation at current time to measuring value X (k) of obstructing objects by described n sensor after rejecting outlier;

Association process unit, for carrying out association process at current time to measuring value X (k) of obstructing objects to described n sensor after time-space relation, obtains the measuring value belonging to target object;

Output unit, exports for the described measuring value belonging to target object being carried out merging.

9. device according to claim 8, is characterized in that, described association process unit comprises:

First acquisition module, for obtaining described n sensor after described time-space relation in measuring value X (k) to obstructing objects of current time and estimated value

Judge module, for according to a described n sensor in measuring value X (k) to obstructing objects of current time and estimated value judge whether that there is obstructing objects is described target object;

Second acquisition module, for when there is obstructing objects and being described target object, from described n sensor after described time-space relation current time to measuring value X (k) of obstructing objects obtain and belong to the measuring value of described target object.

10. device according to claim 9, is characterized in that, described judge module also for:

Obtain each sensor in a described n sensor in the measuring value to obstructing objects of current time and estimated value;

Determine the estimated value probability of described each sensor in the estimated value to obstructing objects of current time, described estimated value probability is:

$A_i=\frac{\left|{\hat{X}}_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{\hat{X}}_i\left(k\right)};$

Wherein, for described i-th sensor is in the estimated value to obstructing objects of current time, A _ifor the estimated value probability of described i-th sensor;

Determine the measuring value probability of described each sensor at the measuring value to obstructing objects of current time, described measuring value probability is:

$B_i=\frac{\left|X_i\left(k\right)\right|}{{\mathrm{\&Sigma;}}_{i=1}^n{X}_i\left(k\right)};$

Wherein, X _i(k) for described i-th sensor is at the measuring value to obstructing objects of current time, B _ifor the measuring value probability of described i-th sensor;

Estimated value probability based on described each sensor determines that each obstructing objects of a described n sensor measurement is the probability m (u of target object ₁) and each obstructing objects be not the probability m (u of target object ₂), wherein, j=1 or 2, m (u _j) be:

$m\left(u_j\right)=\frac{\underset{u_k^1\&cap;...\&cap;u_k^i...\&cap;u_k^n=u_j}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)}{1-\underset{u_k^1\&cap;...\&cap;u_k^i...u_k^n=\mathrm{\&phi;}}{\mathrm{\&Sigma;}}{m}_1\left(u_k^1\right)m_2\left(u_k^2\right)...m_i\left(u_k^i\right)...m_n\left(u_k^n\right)};$

Wherein, ∩ represents and gets common factor, and φ represents empty set, represent that described i-th sensor exists the measuring value to described each obstructing objects at current time, represent that described i-th sensor does not exist the measuring value to described each obstructing objects at current time, if i-th sensor is at the measuring value of current time existence to described each obstructing objects, then if there is not the measuring value to described each obstructing objects at current time in i-th sensor, then

Be the probability m (u of target object by described each obstructing objects ₁) and described each obstructing objects be not the probability m (u of target object ₂) difference as interactive calculation value m;

Judge whether described interactive calculation value m is greater than the second preset value ε;

If described interactive calculation value m is greater than described second preset value ε, judge in a described n sensor, whether the estimated value probability of each sensor and the difference of measuring value probability are less than the 3rd preset value τ;

If the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ, then there is obstructing objects is described target object, and described target object is the barrier that the estimated value probability of described each sensor and the difference of measuring value probability are all less than described 3rd preset value τ.

11. devices according to claim 7, is characterized in that, described second acquisition unit comprises:

3rd acquisition module, for obtaining a upper moment of described each sensor to the estimated value of target obstacle;

Determination module, for the upper moment based on described each sensor to the described predicted value of the estimated value determination current time of target obstacle described predicted value for:

$\tilde{X}\left(k\right)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&mu;}}_i{\hat{X}}_i(k-1);$

Wherein, for a upper moment of described i-th sensor is to the estimated value of target obstacle, μ _ifor the predicted value coefficient of described i-th sensor, and

12. devices according to claim 8, is characterized in that, described output unit comprises:

4th acquisition module, for obtaining the mean value of the measuring value belonging to described target object;

Output module, for exporting described mean value.