CN112615604A - Filtering method and device of intelligent driving perception system and electronic equipment - Google Patents

Abstract

The present invention provides a filtering method, device and electronic device for an intelligent driving perception system. Wherein, the method includes: determining the target motion model set of the obstacle target based on the motion model in the prior knowledge base, and judging whether the target motion model set is consistent with the model set of the VSIMM filter at the previous moment; if not, based on the target motion model set The motion model set and the model set of the VSIMM filter at the previous moment, determine the target model set of the VSIMM filter at the current moment; obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs the measurement according to the measurement information and the target model set. Filter the state estimation to obtain the total estimated output information of the VSIMM filter at the current moment; track the obstacle target according to the total estimated output information, thereby improving the tracking accuracy of the obstacle target through the VSIMM filter, thereby improving the performance of the intelligent driving perception system. Has better practical value.

Description

Filtering method, device and electronic device for intelligent driving perception system

technical field

The present invention relates to the technical field of intelligent driving environment perception, in particular to a filtering method, device and electronic device of an intelligent driving perception system.

Background technique

In the intelligent driving perception system, to improve its performance, it is mainly to improve the ability of obstacle detection and obstacle target tracking, of which target tracking is a very important link. The filter is used in the target tracking process to calculate the position, speed, trajectory, quantity, type and characteristics of the surrounding obstacle target. In the field of intelligent driving, commonly used sensors are generally lidar, camera, millimeter-wave radar and ultrasonic radar.

The computational optimization problems faced by target tracking and filtering mainly have two points: one is that the obstacles in the intelligent driving environment are highly maneuverable and the difference is huge, and it is difficult to predict; the other is that there is a lot of noise in the complex environment of the vehicle and the sensor itself also has a certain error noise. It is difficult to fit, which leads to problems such as low accuracy of filter state estimation and even filter divergence in the process of target tracking. The existing method mainly uses adaptive filtering algorithm for filtering. Although this method can improve the tracking accuracy of obstacle targets to a certain extent, due to the strong mobility and randomness of obstacle targets, it is difficult to determine the motion model of obstacle targets. , thereby affecting the filtering effect, resulting in a decrease in the tracking accuracy of the obstacle target, and even the situation of losing the obstacle target, which affects the performance of the intelligent driving perception system.

SUMMARY OF THE INVENTION

In view of this, the purpose of the present invention is to provide a filtering method, device and electronic device for an intelligent driving perception system to alleviate the above problems.

In a first aspect, an embodiment of the present invention provides a filtering method for an intelligent driving perception system, the method is applied to a server of an intelligent driving perception system, wherein the server is provided with a prior knowledge base and a variable-structure interactive multi-model VSIMM filter , the method includes: determining the target motion model set of the obstacle target based on the motion model in the prior knowledge base, wherein the motion model includes at least one of the following: a uniform velocity CV model, a uniform acceleration CA model, a uniform rotational speed CT model, a current Current model and curve motion CM model; determine whether the target motion model set is consistent with the model set of the VSIMM filter at the last moment; wherein, the model set of the VSIMM filter at the last moment includes the corresponding value of each filter in the VSIMM filter at the last moment. If not, then based on the target motion model set and the model set of the VSIMM filter at the previous moment, determine the target model set of the VSIMM filter at the current moment; obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter The device performs filter state estimation according to the measurement information and the target model set, and obtains the total estimated output information of the VSIMM filter at the current moment; and tracks the obstacle target according to the total estimated output information.

In conjunction with the first aspect, the embodiment of the present invention provides a first possible implementation manner of the first aspect, wherein the above step of judging whether the target motion model set is consistent with the model set at the previous moment of the VSIMM filter includes: judging the target motion model set Whether each target motion model in the motion model set is exactly the same as the motion model corresponding to each filter in the VSIMM filter at the last moment; if so, then determine that the target motion model set is consistent with the model set at the previous moment of the VSIMM filter; If there is any difference, it is determined that the target motion model set is inconsistent with the model set of the VSIMM filter at the previous moment.

In conjunction with the first possible implementation manner of the first aspect, the embodiment of the present invention provides the second possible implementation manner of the first aspect, wherein the above-mentioned model set based on the target motion model set and the model set at the previous moment of the VSIMM filter, The step of determining the target model set at the current moment of the VSIMM filter, comprising: obtaining the model set at the current moment of the VSIMM filter according to the union of the target motion model set and the model set at the previous moment of the VSIMM filter; Obtain the model set of the target motion model set The first model probability and the second model probability of the model set at the previous moment of the VSIMM filter; according to the ratio of the first model probability and the second model probability, the target model set of the VSIMM filter at the current moment is determined.

With reference to the second possible implementation manner of the first aspect, the embodiment of the present invention provides the third possible implementation manner of the first aspect, wherein the VSIMM filter is determined according to the ratio of the first model probability and the second model probability The step of the target model set at the current moment includes: judging whether the ratio is greater than the first threshold, and if so, determining that the target model set of the VSIMM filter at the current moment is the target motion model set; if not, judging whether the ratio is less than the second threshold, If yes, determine that the target model set of the VSIMM filter at the current moment is the model set of the VSIMM filter at the previous moment; wherein, the second threshold is smaller than the first threshold.

With reference to the third possible implementation manner of the first aspect, the embodiment of the present invention provides the fourth possible implementation manner of the first aspect, wherein the VSIMM filter performs the step of estimating the filtering state according to the measurement information and the target model set , including: determine the target filter in the corresponding VSIMM filter according to each target model in the target model set; carry out filtering state estimation by each target filter according to the measurement information and the corresponding target model, and obtain the VSIMM filter at the current moment. Total estimated output information; wherein, the total estimated output information includes the total state estimate and total error covariance, the total state estimate includes the state estimate of each target filter at the current moment, and the total error covariance includes each target filter The error covariance at the current instant.

With reference to the fourth possible implementation manner of the first aspect, the embodiment of the present invention provides the fifth possible implementation manner of the first aspect, wherein each target model is further configured with a model probability, and the method further includes: based on each target model The likelihood function of each target model is updated, and the model probability of each target model is updated; wherein, the likelihood function is calculated by the following formula:

Among them, Λ _j (k) represents the likelihood function of the j-th target model at time k, S _j (k) represents the measurement covariance of the j-th target model at time k, and v _j (k) represents the j-th Kalman filter residuals of the target model at time k.

In conjunction with the first aspect, the embodiment of the present invention provides a sixth possible implementation manner of the first aspect, wherein the above-mentioned step of determining the target motion model set of the obstacle target based on the motion model in the prior knowledge base includes: Obtain the collection information of the obstacle target; wherein, the collection information includes type information, position information and parameter information; determine the first motion model set of the obstacle target based on the type information; determine the second motion model set of the obstacle target based on the position information; A third motion model set of the obstacle target is determined based on the parameter information; wherein the parameter information includes at least one of the following: acquisition device information, obstacle target contour information and weather information; according to the first motion model set, the second motion model set and The third motion model set is to determine the target motion model set of the obstacle target.

In a second aspect, an embodiment of the present invention further provides a filtering device for an intelligent driving perception system, which is applied to a server of an intelligent driving perception system, wherein the server is provided with a prior knowledge base and a variable-structure interactive multi-model VSIMM filter , the device includes: a first determination module for determining the target motion model set of the obstacle target based on the motion model in the prior knowledge base, wherein the motion model includes at least one of the following: a uniform velocity CV model, a uniform acceleration CA model , CT model of uniform rotation speed, current current model and CM model of curvilinear motion; judgment module, used to judge whether the target motion model set is consistent with the model set of the VSIMM filter at the last moment; wherein, the model set of the VSIMM filter at the last moment includes: The motion model corresponding to each filter in the VSIMM filter at the last moment; the second determination module, for if not, then based on the target motion model set and the model set at the previous moment of the VSIMM filter, determine the current moment of the VSIMM filter. Target model set; filter estimation module, used to obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs filtering state estimation according to the measurement information and the target model set, and obtains the total estimated output information of the VSIMM filter at the current moment. ; Tracking module, used to track the obstacle target according to the total estimated output information.

In a third aspect, an embodiment of the present invention further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor implements the intelligent driving of the first aspect when the processor executes the computer program Steps of a filtering method for a perceptual system.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and the computer program is executed by a processor to execute the filtering method of the intelligent driving perception system of the first aspect above. step.

The embodiments of the present invention have brought the following beneficial effects:

The embodiments of the present invention provide a filtering method, device and electronic device for an intelligent driving perception system, dynamically determine the target motion model of the obstacle target based on the motion model in the prior knowledge base, and determine the current value of the VSIMM filter according to the target motion model A set of target models at all times for filtering processing, so that the VSIMM filter can quickly converge when the obstacle target is maneuvering, which improves the tracking accuracy of the obstacle target, thereby improving the performance of the intelligent driving perception system, and has good practicality. value.

Other features and advantages of the present invention will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the description and drawings.

In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below, and are described in detail as follows in conjunction with the accompanying drawings.

Description of drawings

In order to illustrate the specific embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the specific embodiments or the prior art. Obviously, the accompanying drawings in the following description The drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

1 is a schematic diagram of a filtering method for an intelligent driving perception system provided by an embodiment of the present invention;

2 is a flowchart of a filtering method for an intelligent driving perception system provided by an embodiment of the present invention;

3 is a schematic diagram of a road terrain according to an embodiment of the present invention;

4 is a schematic diagram of a target motion model set for constructing an obstacle target according to an embodiment of the present invention;

5 is a schematic diagram of determining a target model set according to an embodiment of the present invention;

FIG. 6 is a working principle diagram of a VSIMM filter provided by an embodiment of the present invention;

7 is a schematic diagram of a filtering device of an intelligent driving perception system according to an embodiment of the present invention;

FIG. 8 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of them. example. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Aiming at the problem that the motion model of the existing obstacle target is difficult to determine, thereby affecting the filtering effect, the embodiments of the present invention provide a filtering method, device and electronic device for an intelligent driving perception system, which improve the tracking accuracy of the obstacle target, thereby improving the It improves the performance of the intelligent driving perception system and has good practical value.

In order to facilitate the understanding of this embodiment, a filtering method of an intelligent driving perception system provided by an embodiment of the present invention is first introduced in detail below.

An embodiment of the present invention provides a filtering method for an intelligent driving perception system, and the method is applied to an intelligent driving perception system. Among them, the intelligent driving perception system includes a server, and a detection unit connected to the server, where the detection unit includes a plurality of detection elements such as sensors, etc., in order to obtain the environmental perception information of the intelligent driving perception system, where the environmental perception information includes sensor information, electronic map information and other environmental information, etc.

As shown in Figure 1, the server builds a prior knowledge base according to the acquired environment perception information, and the prior knowledge base outputs the corresponding filter motion model set, and according to the filter motion model set and VSIMM (Variable Structure Interacting Multiple Model, variable structure The model probability calculated by the interactive multi-model) filter adjusts the model set of the VSIMM filter, and performs filtering state estimation based on the adjusted VSIMM filter according to the obstacle motion data of the input obstacle target, so as to track the obstacle target. . It should be noted that the above VSIMM filter is also referred to as a VSIMM filtering algorithm or a VSIMM algorithm.

Based on the above server, a filtering method of an intelligent driving perception system provided by an embodiment of the present invention is shown in FIG. 2 , and the method includes the following steps:

Step S202, determining the target motion model set of the obstacle target based on the motion model in the prior knowledge base;

Among them, in the intelligent driving perception system, the motion model of the obstacle target includes at least one of the following: a CV (Constant Velocity, uniform velocity) model, a CA (Constant Acceleration, uniform acceleration) model, a CT (ConstantTurn, uniform rotation speed) model, current Current model and CM (Curvilinear Motion, curve motion) model. It should be noted that the motion models of other obstacle targets may be set according to actual conditions, which are not limited in this embodiment of the present invention.

For ease of understanding, each motion model is described as follows:

(1) CV model; wherein, the CV model is calculated according to the following formula:

表示k时刻x方向障碍物目标的速度，y_k表示k时刻y方向障碍物目标的位移，

表示k时刻y方向障碍物目标的速度，x_k+1表示k+1时刻x方向障碍物目标的位移，

表示k+1时刻x方向障碍物目标的速度，y_k+1表示k+1时刻y方向障碍物目标的位移，

表示k+1时刻y方向障碍物目标的速度，T表示时间间隔，其中，该时间间隔为k时刻到k+1时刻的时间间隔。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Represents the velocity of the obstacle target in the x direction at time k, y _k represents the displacement of the obstacle target in the y direction at time k,

represents the velocity of the obstacle target in the y direction at time k, x _k+1 represents the displacement of the obstacle target in the x direction at time k+1,

Represents the velocity of the obstacle target in the x direction at the time k+1, y _k+1 represents the displacement of the obstacle target in the y direction at the time k+1,

represents the speed of the obstacle target in the y direction at time k+1, and T represents the time interval, where the time interval is the time interval from time k to time k+1.

(2) CA model; wherein, CA model is calculated according to the following formula:

表示k时刻x方向障碍物目标的速度，

表示k时刻x方向障碍物目标的加速度，y_k表示k时刻y方向障碍物目标的位移，

表示k时刻y方向障碍物目标的速度，

表示k时刻y方向障碍物目标的加速度，x_k+1表示k+1时刻x方向障碍物目标的位移，

表示k+1时刻x方向障碍物目标的速度，

表示k+1时刻x方向障碍物目标的加速度，y_k+1表示k+1时刻y方向障碍物目标的位移，

表示k+1时刻y方向障碍物目标的速度，

表示k+1时刻y方向障碍物目标的加速度，T表示时间间隔。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Indicates the velocity of the obstacle target in the x direction at time k,

represents the acceleration of the obstacle target in the x direction at time k, y _k represents the displacement of the obstacle target in the y direction at time k,

represents the velocity of the obstacle target in the y direction at time k,

represents the acceleration of the obstacle target in the y direction at time k, x _k+1 represents the displacement of the obstacle target in the x direction at time k+1,

Indicates the velocity of the obstacle target in the x direction at time k+1,

represents the acceleration of the obstacle target in the x direction at the time k+1, y _k+1 represents the displacement of the obstacle target in the y direction at the time k+1,

Represents the velocity of the obstacle target in the y direction at time k+1,

It represents the acceleration of the obstacle target in the y direction at time k+1, and T represents the time interval.

(3) CT model; wherein, the CT model is calculated according to the following formula:

表示k时刻x方向障碍物目标的速度，y_k表示k时刻y方向障碍物目标的位移，

表示k时刻y方向障碍物目标的速度，x_k+1表示k+1时刻x方向障碍物目标的位移，

表示k+1时刻x方向障碍物目标的速度，y_k+1表示k+1时刻y方向障碍物目标的位移，

表示k+1时刻y方向障碍物目标的速度，T表示时间间隔，ω表示障碍物目标的角速度。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Represents the velocity of the obstacle target in the x direction at time k, y _k represents the displacement of the obstacle target in the y direction at time k,

represents the velocity of the obstacle target in the y direction at time k, x _k+1 represents the displacement of the obstacle target in the x direction at time k+1,

Represents the velocity of the obstacle target in the x direction at the time k+1, y _k+1 represents the displacement of the obstacle target in the y direction at the time k+1,

Represents the velocity of the obstacle target in the y direction at time k+1, T represents the time interval, and ω represents the angular velocity of the obstacle target.

(4) Current model; wherein, the Current model is calculated according to the following formula:

表示k时刻x方向障碍物目标的速度，

表示k时刻x方向障碍物目标的加速度，

表示k时刻x方向障碍物目标的急动度，

表示随机加速度的均值，w(t)表示均值为零、机动加速度方差为σ_a ²的高斯白噪声，α表示机动加速度时间常数的倒数。需要说明的是，这里

在每一个采样间隔内均为常数，机动频率α通常根据经验取值，并通过实际测量确定，具体可以根据实际情况进行设置，本发明实施例对此不作限制说明。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Indicates the velocity of the obstacle target in the x direction at time k,

represents the acceleration of the obstacle target in the x direction at time k,

represents the jerkiness of the obstacle target in the x direction at time k,

is the mean value of random acceleration, w(t) is Gaussian white noise with zero mean value and variance of maneuvering acceleration σ _a ² , and α is the inverse of the time constant of maneuvering acceleration. It should be noted that here

It is a constant in each sampling interval, and the maneuvering frequency α is usually valued according to experience and determined through actual measurement, and can be specifically set according to the actual situation, which is not limited in this embodiment of the present invention.

Its discrete equation is as follows:

表示k时刻x方向障碍物目标的速度，

表示k时刻x方向障碍物目标的加速度，x_k+1表示k+1时刻x方向障碍物目标的位移，

表示k+1时刻x方向障碍物目标的速度，

表示k+1时刻x方向障碍物目标的加速度，

表示随机加速度的均值，α表示高斯白噪声与噪声方差的系数，T表示时间间隔，W(t)表示均值为零的高斯白噪声的离散值。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Indicates the velocity of the obstacle target in the x direction at time k,

represents the acceleration of the obstacle target in the x direction at time k, x _k+1 represents the displacement of the obstacle target in the x direction at the time k+1,

Represents the velocity of the obstacle target in the x direction at time k+1,

represents the acceleration of the obstacle target in the x direction at time k+1,

represents the mean value of random acceleration, α represents the coefficient of white Gaussian noise and noise variance, T represents the time interval, and W(t) represents the discrete value of white Gaussian noise with zero mean.

(5) CM model; wherein, the CM model is calculated according to the following formula:

表示k时刻x方向障碍物目标的速度，y_k表示k时刻y方向障碍物目标的位移，

表示k时刻y方向障碍物目标的速度，x_k+1表示k+1时刻x方向障碍物目标的位移，

表示k+1时刻x方向障碍物目标的速度，y_k+1表示k+1时刻y方向障碍物目标的位移，

表示k+1时刻y方向障碍物目标的速度，T表示时间间隔，Δθ表示速度方向角的变化量，a_t表示曲线运动切向加速度，Δt_k表示k+1时刻和k时刻的时间差，θ_k表示k时刻障碍物目标的速度方向角。Among them, x _k represents the displacement of the obstacle target in the x direction at time k,

Represents the velocity of the obstacle target in the x direction at time k, y _k represents the displacement of the obstacle target in the y direction at time k,

represents the velocity of the obstacle target in the y direction at time k, x _k+1 represents the displacement of the obstacle target in the x direction at time k+1,

Represents the velocity of the obstacle target in the x direction at the time k+1, y _k+1 represents the displacement of the obstacle target in the y direction at the time k+1,

Represents the speed of the obstacle target in the y direction at the time k+1, T represents the time interval, Δθ represents the change of the speed direction angle, a _t represents the tangential acceleration of the curve motion, Δt _k represents the time difference between the time k+1 and the time k, θ _k represents the velocity direction angle of the obstacle target at time k.

Among them, the change amount of the above-mentioned velocity direction angle is calculated according to the following formula:

Δθ = θ _{k+1 -} θ _k (6)

Among them, Δθ represents the change amount of the speed direction angle, θ _k+1 represents the speed direction angle of the obstacle target at time k+1, and θ _k represents the speed direction angle of the obstacle target at time k.

Therefore, according to the above-mentioned motion model pre-stored in the prior knowledge base, the target motion model set of the obstacle target can be constructed, and one of the possible target motion model sets is constructed as follows:

(1) Obtain the collection information of the obstacle target; wherein, the collection information includes but is not limited to type information, location information and parameter information; it can be set according to the actual situation, which is not limited in this embodiment of the present invention.

(2) Determine the first motion model set of the obstacle target based on the type information; specifically, determine the corresponding first motion model set according to the type information of the obstacle target. Uniform motion, uniform turning motion and curved motion, that is, when the type information of the obstacle target is a large motor vehicle, the corresponding first motion model set is {CV model, CT model, CM model}; for pedestrians, it is generally straight walking Or the uniform motion and current model of sprinting, that is, when the type information of the obstacle target is pedestrian, the corresponding first motion model set is {CV model, Current model}; for small vehicles such as bicycles or motorcycles, there will be A certain acceleration and deceleration, that is, when the type information of the obstacle target is a small motor vehicle, the corresponding first motion model set is {CV model, CA model, Current model}, etc. Therefore, the first A collection of motion models, which can be set according to the actual situation.

(3) Determine the second motion model set of the obstacle target based on the position information; specifically, obtain the position information of the obstacle target according to the high-precision electronic map and the position matching of the obstacle target in the electronic map, such as straight road, U-turn Lanes, etc., as shown in Figure 3. For motor vehicles, traffic rules must be followed. For example, on a straight lane, the corresponding second motion model set is {CV model, CA model, Current model}; on an overpass, a corner lane or a U-turn lane, the corresponding second motion model The model set is {CT model, Current model}; in the area to be turned or variable roads such as fork roads and ramps, the corresponding second motion model set is {CM model, Current model}, etc.; the specifics can be based on actual conditions Make settings.

(4) Determine the third motion model set of the obstacle target based on the parameter information; wherein, the parameter information includes at least one of the following: acquisition device information, obstacle target contour information and weather information; wherein, the acquisition device information such as sensor information, etc., Since the above parameter information is rarely considered in the existing methods, the third motion model set here is the complete set by default, that is, the third motion model set at this time is {CV model, CA model, CT model, CM model, Current model}; The rest of the non-comprehensive situations can be set according to the actual scene.

(5) Determine the target motion model set of the obstacle target according to the first motion model set, the second motion model set and the third motion model set. Specifically, the intersection of the first motion model set, the second motion model set and the third motion model set is determined, and the intersection is determined as the target motion model set of the obstacle target, so as to ensure that the target motion model set can accurately simulate the obstacle target movement.

For ease of understanding, the sensor is used as an example here. As shown in FIG. 4 , the collected information includes sensor perception information, high-precision map information and other environmental information; wherein, the sensor perception information includes sensor information and obstacle categories, and the sub-model set M ₁ of the obstacle target can be determined according to the sensor information, The sub-model set M ₂ of the obstacle target can be determined according to the obstacle category, the obstacle terrain can be determined based on the sensor perception information and the high-precision map information, and the sub-model set M ₃ of the obstacle target can be determined based on the obstacle terrain, and, The sub-model set M _n of the obstacle target can be determined based on other perception information in other environmental information, where n is the number of collected information. Finally, according to the above-mentioned sub-model set M ₁ , sub-model set M ₂ , and sub-model set M ₃ The intersection with the sub-model set _Mn determines the target motion model set of the obstacle target, that is, the target motion model set M=M ₁ ∩M ₂ ∩M ₃ ∩... _Mn .

It should be noted that the target motion model set of a specific obstacle target may be constructed according to an actual application scenario, so as to ensure that the target motion model set can accurately simulate the motion of the obstacle target, which is not limited in this embodiment of the present invention.

In practical applications, since the target maneuvering type, maneuvering intensity and maneuvering time of the obstacle target are unknown, especially the obstacle target changes its motion mode many times during the maneuvering process, such as turning, accelerating, climbing, etc. Therefore, The motion model is generally difficult to determine, that is, the target motion model of the obstacle target is difficult to match the actual motion pattern of the obstacle target, which leads to the divergence of the filter, the decrease of the tracking accuracy and even the loss of the target. However, the embodiment of the present invention uses prior knowledge such as map environment information and type information of the obstacle target to determine the target motion model set of the obstacle target, thereby alleviating the problem that the motion model of the existing obstacle target is difficult to determine, and, By constructing a prior knowledge base, the target motion model set of the obstacle target is also expanded, so that the target motion model set can better fit the actual motion of the obstacle target, thereby improving the estimation accuracy of the obstacle target state.

Step S204, determine whether the target motion model set is consistent with the model set at the previous moment of the VSIMM filter;

Wherein, the model set of the VSIMM filter at the last moment includes the motion model corresponding to each filter in the VSIMM filter at the last moment; Whether the motion models corresponding to each filter are exactly the same; if so, it is determined that the target motion model set is consistent with the model set of the VSIMM filter at the last moment, that is, there is no new filter model in the model set of the VSIMM filter at the last moment Activation; if there is any difference, it is determined that the target motion model set is inconsistent with the model set of the VSIMM filter at the previous moment, that is, the model set of the VSIMM filter at the current moment is compared with the model set at the previous moment, there is a new filter. In this case, the model set of the VSIMM filter at the current moment needs to be determined, that is, the filter model of the VSIMM filter is adjusted.

Step S206, if not, then based on the target motion model set and the model set at the previous moment of the VSIMM filter, determine the target model set at the current moment of the VSIMM filter;

In practical applications, the adaptive filtering algorithm enables the target tracking system to adjust adaptively according to the maneuvering situation of the target, which greatly improves the tracking accuracy. There are mainly three types of commonly used adaptive filtering algorithms: detection adaptive filtering algorithm, real-time identification adaptive filtering algorithm and multi-model method; among them, the detection adaptive filtering algorithm and real-time identification adaptive filtering algorithm are both in the filtering process. The noise is corrected, but there will be a general lag and a large dependence on the model, while the multi-model method such as the IMM (Interactive Multi-tude Model, interactive multi-model adaptive) filtering algorithm does not require maneuver detection, but adjusts each model by adjusting each model. The probability of switching between different modes can be achieved to achieve comprehensive self-adaptation. However, none of these algorithms essentially change the fixed structure of the IMM, and still cannot solve the problem of performance degradation caused by the interference of redundant target motion models in the IMM algorithm. In the embodiment of the present invention, the fixed structure of the IMM algorithm is changed as much as possible by using prior knowledge and external sensor information, that is, the filter performance is improved through the VSIMM algorithm, for example, the motion model of the obstacle target can be dynamically increased or decreased, or the obstacle The estimated position of the object target is corrected, so that when the obstacle target maneuvers, the tracking filter in the VSIMM filter can converge more quickly, so as to obtain higher tracking accuracy and improve operation speed.

Specifically, since each target motion model in the target motion model set is not exactly the same as the motion model corresponding to each filter in the VSIMM filter at the previous moment, the target motion determined in the prior knowledge base according to certain rules Adjust the filter model in the model set. Specifically, first, according to the union of the target motion model set and the model set of the VSIMM filter at the previous moment, the model set of the VSIMM filter at the current moment is obtained; then, the first model probability of the target motion model set and the VSIMM filter are obtained. The second model probability of the model set at a moment, and the ratio of the first model probability to the second model probability is obtained, and it is judged whether the ratio is greater than the first threshold. If so, the target model set of the VSIMM filter at the current moment is determined as the target. Motion model set; if the ratio is not greater than the first threshold, determine whether the ratio is less than the second threshold, if so, determine the target model set of the VSIMM filter at the current moment is the model set of the VSIMM filter at the previous moment. It should be noted that the specific values of the first threshold and the second threshold may be set according to actual conditions, as long as the second threshold is less than the first threshold.

和VSIMM滤波器k时刻的模型集合M_k的第二模型概率

其中，根据下式计算模型集合的模型概率：For ease of understanding, the model set M _k and the target motion model set M _m at the previous moment of the VSIMM filter, that is, moment k, are used as examples for illustration. As shown in Figure 5, according to M _k and M _m , it is judged whether the model set at the current moment of the VSIMM filter, that is, the model set at time k+1, is updated; if not, the model set at the time k+1 of the VSIMM filter is M _k+1 , And, M _k+1 =M _k , if there is an update, the model set M _k+1 =M at the time of VSIMM filter k+1 is determined according to the target motion model set M _m and the model set M _k at the time of VSIMM filter k _k ∪M _m , and obtain the first model probability of the target motion model set M _m

and the second model probability of the model set M _k at time k of the VSIMM filter

Among them, the model probability of the model set is calculated according to the following formula:

Among them, u _M (k) represents the model probability of the model set M, m _i represents the number of models in the model set M, and u _i (k) represents the probability of each model in the model set.

和第二模型概率

并根据下式计算得到第一模型概率和第二模型概率的比值：Therefore, the first model probability can be calculated according to the above formula

and the second model probability

And calculate the ratio of the first model probability and the second model probability according to the following formula:

表示k时刻目标运动模型集合M_m的第一模型概率，

表示VSIMM滤波器k时刻的模型集合M_k的第二模型概率。Among them, t represents the ratio of the first model probability and the second model probability,

represents the first model probability of the target motion model set M _m at time k,

represents the second model probability of the model set M _k at time k of the VSIMM filter.

After calculating the above ratio, determine whether the ratio t is greater than the first threshold t ₁ , and if so, determine that the target model set at the time of VSIMM filter k+1 is the target motion model set, that is, M _k+1 =M _m ; if not, Determine whether the ratio t is less than the second threshold t ₂ , and if so, determine that the target model set at the time of VSIMM filter k+1 is the model set at the time of VSIMM filter k, that is, M _k+1 =M _k , so that according to the target motion The model set M _m and the model set u _Mm (k) at the time of the VSIMM filter k, determine the target model set M _{k+1 at the time of the VSIMM filter k+1} , if the ratio t is greater than or equal to the second threshold t ₂ and less than or is equal to the first threshold t ₁ , then continue to determine the target model set M _k+ 1 at the time of VSIMM filter k+1 according to M _k+1 =M _k ∪M _m .

Step S208, obtaining the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs filtering state estimation according to the measurement information and the target model set, and obtains the total estimated output information of the VSIMM filter at the current moment;

For the VSIMM filter, the basic idea is to first establish a general model set, which is composed of multiple independent and compatible model set sequences. At each moment in the target tracking process, according to prior knowledge and The estimated state of the target selects a model set that is most consistent with the target motion state in the total model set; among them, each model set can be converted to each other according to certain rules, that is, each filter model in the model set is independent and can interact with each other. converted. Therefore, for the target model set determined above, the VSIMM filter will also allocate a corresponding target model to each filter in the VSIMM filter according to a certain allocation rule, so that each filter performs filtering according to the allocated target model.

Specifically, the target filter in the corresponding VSIMM filter is determined according to each target model in the target model set; then, the filtering state estimation is performed by each target filter according to the measurement information and the corresponding target model, and the VSIMM filter at the current moment is obtained. The total estimated output information of the filter; the total estimated output information includes the total state estimate and total error covariance, the total state estimate includes the state estimate of each target filter at the current moment, and the total error covariance includes each target filter. The error covariance at the current instant. It should be noted that the measurement information of the obstacle target here includes but is not limited to the position information and speed information of the obstacle target, etc., which may be specifically set according to the actual situation.

For ease of understanding, the model set M _k and the target motion model set M _m at the previous moment of the VSIMM filter, that is, moment k, are used as examples for illustration. If it is determined that the filter model of the VSIMM filter needs to be adjusted at time k+1, then for the determined target model set M _k+ 1 at time k+1 of the VSIMM filter, the target model corresponding to each target filter can be assigned an initial The model probability is initialized, and the model probability of each target filter can also be adjusted according to the transition probability matrix, where the transition probability matrix includes multiple transition probabilities, where the transition probability includes the transition probabilities corresponding to multiple target filters. The transition probability corresponding to the target filter also includes a plurality of transition probabilities between different target models corresponding to the target filter.

For the target model set M _k+ 1 at time k+1 of the VSIMM filter, the input of the target filters corresponding to the r target models at time k+1 includes: mixed state estimation and mixed error covariance output at time k. Among them, the mixed state estimation at time k can be calculated according to the following formula:

表示k时刻的混合状态估计，

表示目标模型i在k时刻的总状态估计值，u_i|j(k)表示k时刻的目标模型i到目标模型j的混合概率，其中，目标模型i为VSIMM滤波器k时刻的模型集合M_k的目标模型，目标模型j为VSIMM滤波器k+1时刻的模型集合M_k+1的目标模型。in,

represents the mixed state estimate at time k,

Represents the total state estimate value of target model i at time k, u _i|j (k) represents the mixture probability of target model i to target model j at time k, where target model i is the model set M of the VSIMM filter at time k The target model of _k , and the target model j is the target model of the model set M _k+ 1 at the moment k+1 of the VSIMM filter.

It should be noted that the above mixed probability is based on the model probability of target model i and target model j, and the transition probability from target model i to target model j at time k+1 determined by the initial transition probability of target model i and target model j. , so that by adjusting the transition probability between the models at each moment, the uncertainty of the original transition probability matrix due to the uncertainty of the obstacle target maneuver in the existing method is alleviated, so that the real movement of the obstacle target cannot be well reflected. Model switching issues, improved tracking performance.

And, the mixed error covariance at time k is calculated according to the following formula:

表示目标模型i在k时刻的总状态估计值，

表示k时刻的混合状态估计，u_i|j(k)表示k时刻的目标模型i到目标模型j的混合概率，其中，目标模型i为VSIMM滤波器k时刻的模型集合M_k的目标模型，目标模型j为VSIMM滤波器k+1时刻的模型集合M_k+1的目标模型。Among them, P _0j (k) represents the mixed error covariance at time k, P _i (k) represents the total error covariance of the target model i at time k,

represents the total state estimate of the target model i at time k,

Represents the mixed state estimation at time k, u _i|j (k) represents the mixing probability of target model i to target model j at time k, where target model i is the target model of the model set M _k of the VSIMM filter at time k, The target model j is the target model of the model set M _k+ 1 at the time of VSIMM filter k+1.

Among them, the mixing probability can be calculated according to the following formula:

表示目标模型j的预测概率。Among them, u _i|j (k) represents the mixture probability from target model i to target model j at time k, p _ij represents the transition probability of target model i to target model j, and u _i (k) represents the target model i at time k The model probability of ,

represents the predicted probability of the target model j.

Calculate the predicted probability of the target model j according to the following formula:

表示目标模型j的预测概率，p_ij表示目标模型i转变为目标模型j的混合概率，u_i(k)表示k时刻目标模型i的模型概率，目标模型i为VSIMM滤波器k时刻的模型集合M_k的目标模型，目标模型j为VSIMM滤波器k+1时刻的模型集合M_k+1的目标模型。in,

Represents the prediction probability of target model j, p _ij represents the mixture probability of target model i transforming into target model j, u _i (k) represents the model probability of target model i at time k, and target model i is the model set of VSIMM filter at time k The target model of M _k , and the target model j is the target model of the model set M _k+ 1 at the moment of VSIMM filter k+1.

和总误差协方差P_j(k+1)。以及，基于目标模型j的似然函数对目标模型j的模型概率进行更新，其中，根据下式计算目标模型j的似然函数：At this time, filter the target filters in the target model set M _k+1 according to the mixed state estimation and mixed error covariance at time k above to obtain the total estimated output information of each target filter at time k+1, that is, Total state estimates for each target filter at time k+1

and total error covariance P _j (k+1). And, the model probability of the target model j is updated based on the likelihood function of the target model j, wherein the likelihood function of the target model j is calculated according to the following formula:

Among them, Λ _j (k+1) represents the j-th target model, that is, the likelihood function of target model j at time k+1, and S _j (k+1) represents the measurement covariance of target model j at time k+1 , v _j (k+1) represents the Kalman filter residual of the target model j at time k+1.

The model probability of the target model j at time k+1 is updated according to the following formula:

表示目标模型j的预测概率，C表示模型概率归一化常数。Among them, u _j (k+1) represents the model probability of target model j at time k+1, Λ _j (k+1) represents the likelihood function of target model j at time k+1,

represents the predicted probability of the target model j, and C represents the model probability normalization constant.

Among them, the model probability normalization constant is calculated according to the following formula:

表示目标模型j的预测概率，目标模型j为VSIMM滤波器k+1时刻的模型集合M_k+1的目标模型。Among them, C represents the model probability normalization constant, Λ _j (k+1) represents the likelihood function of the target model j at time k+1,

represents the prediction probability of the target model j, and the target model j is the target model of the model set M _k+ 1 at the time of the VSIMM filter k+1.

According to the above formula, the total estimated output information of the target model set M _k+ 1 at time k+1 of the VSIMM filter can be calculated. where the total state estimate is calculated according to

表示目标模型集合M_k+1的总状态估计值，

表示目标模型j在k+1时刻的总状态估计值，u_j(k+1)表示k+1时刻目标模型j的模型概率。in,

represents the total state estimate of the target model set M _k+1 ,

Represents the total state estimate of target model j at time k+1, and u _j (k+1) represents the model probability of target model j at time k+1.

And, the total error covariance is calculated according to:

表示目标模型j在k+1时刻的总状态估计值，

表示目标模型集合M_k+1的总状态估计值。Among them, P(k+1) represents the total error covariance of the target model set M _k+1 , u _j (k+1) represents the model probability of the target model j at k+1 time, P _j (k+1) represents the target The total error covariance of model j at time k+1,

represents the total state estimate of the target model j at time k+1,

represents the total state estimate of the target model set _Mk+1 .

For ease of understanding, examples are provided here. As shown in FIG. 6 , based on the measurement information at time k+1, the first target filter output corresponding to the first target model in the target model set M _k+1 of the VSIMM filter at time k+1 has the first estimated output information and the probability of the first model, the output of the second target filter corresponding to the second target model has the second estimated output information and the probability of the second model, and so on, the output of the Nth target filter corresponding to the Nth target model has the Nth output The estimated output information and the Nth model probability can be obtained according to the first estimated output information, the second estimated output information to the Nth estimated output information, and the total estimated output information corresponding to the target model set M _k+1 . In addition, the probability of the first model, the probability of the second model and the probability of the Nth model can also be updated based on the likelihood function, so as to improve the accuracy of the model probability corresponding to each target filter at each moment, thereby improving the target Track performance.

Therefore, the transition probability is adjusted by the mixing probability of the VSIMM filter, and the model probability of each target filter is updated by the likelihood function, thereby alleviating the maneuvering uncertainty of the obstacle target, which leads to the original VSIMM filtering. The transition probability matrix setting method in the detector cannot reflect the switching of the real motion model of the target very well; and, due to the interference of many external environmental information, the model probability of the obstacle target in different environments has also changed a lot. , the original setting method is difficult to obtain a more accurate model probability, which leads to the problem of degraded tracking performance. At the same time, it can also alleviate the problem that the measurement model of the sensor is difficult to determine and the noise variance of the measurement is constantly changing, so real-time estimation is necessary to ensure The problem of tracking accuracy, thereby improving the target tracking performance.

Step S210, track the obstacle target according to the total estimated output information.

Therefore, the above filtering method of the intelligent driving perception system makes full use of various prior knowledge of the intelligent driving perception system to constrain the motion model of the obstacle target and eliminate the wrong motion model, so as to determine the target motion model of the obstacle target, and the maximum It maximizes the utilization of prior knowledge information, reduces the parallel operation of a large number of motion models, and ensures that the target motion model of the obstacle target is more in line with the actual motion of the obstacle target.

In addition, the VSIMM algorithm also avoids the problem of tracking failure caused by the divergence of the traditional tracking filter, and alleviates the problem that the target model set of the filter in the general VSIMM algorithm is difficult to determine, which leads to the decrease of tracking accuracy and the increase of calculation time. Determine the target model set of the VSIMM filter at each moment, reduce the calculation time of the VSIMM filter, improve the state estimation accuracy of the VSIMM filter tracking filtering, thereby improving the tracking accuracy of the obstacle target, thereby improving the intelligent driving perception system. performance, has good practical value.

On the basis of the above embodiments, the embodiments of the present invention also provide a filtering device for an intelligent driving perception system, which is applied to a server of the intelligent driving perception system, wherein the server provides a prior knowledge base and variable structure interactive Multi-model VSIMM filter. As shown in FIG. 7 , the device includes a first determination module 71, a judgment module 72, a second determination module 73, a filter estimation module 74 and a tracking module 75 connected in sequence; wherein, the functions of each module are as follows:

The first determination module 71 is used to determine the target motion model set of the obstacle target based on the motion model in the prior knowledge base, wherein the motion model includes at least one of the following: a uniform velocity CV model, a uniform acceleration CA model, and a uniform rotational speed CT Model, Current Current Model and Curve Motion CM Model;

Judging module 72, for judging whether the target motion model set is consistent with the model set at the last moment of the VSIMM filter; wherein, the model set at the last moment of the VSIMM filter includes the motion corresponding to each filter in the VSIMM filter at the last moment Model;

The second determination module 73 is used to determine the target model set of the VSIMM filter at the current moment based on the target motion model set and the model set at the previous moment of the VSIMM filter if not;

The filter estimation module 74 is used to obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs filtering state estimation according to the measurement information and the target model set, and obtains the total estimated output information of the VSIMM filter at the current moment;

The tracking module 75 is used to track the obstacle target according to the total estimated output information.

The filtering device of the intelligent driving perception system provided by the embodiment of the present invention dynamically determines the target motion model of the obstacle target based on the motion model in the prior knowledge base, and determines the target model set of the VSIMM filter at the current moment according to the target motion model , so as to perform filtering processing, so that the VSIMM filter can quickly converge when the obstacle target is maneuvering, which improves the tracking accuracy of the obstacle target, thereby improving the performance of the intelligent driving perception system, and has good practical value.

In one possible embodiment, the above-mentioned judgment module 72 is also used for: judging whether each target motion model in the target motion model set is exactly the same as the motion model corresponding to each filter in the VSIMM filter at the previous moment; if If yes, it is determined that the target motion model set is consistent with the model set of the VSIMM filter at the previous moment; if there is any difference, it is determined that the target motion model set is inconsistent with the model set of the VSIMM filter at the previous moment.

In another possible embodiment, the above-mentioned second determination module 73 is further configured to: obtain the model set of the VSIMM filter at the current moment according to the union of the target motion model set and the model set of the VSIMM filter at the previous moment; obtain The first model probability of the target motion model set and the second model probability of the VSIMM filter model set at the previous moment; according to the ratio of the first model probability and the second model probability, the target model set of the VSIMM filter at the current moment is determined.

In another possible embodiment, the above-mentioned second determination module 73 is also used to: determine whether the ratio is greater than the first threshold, and if so, determine that the target model set at the current moment of the VSIMM filter is the target motion model set; if no , determine whether the ratio is less than the second threshold, and if so, determine that the target model set of the VSIMM filter at the current moment is the model set of the VSIMM filter at the previous moment; wherein, the second threshold is smaller than the first threshold.

In another possible embodiment, the above-mentioned filter estimation module 74 is further configured to: determine a corresponding target filter in the VSIMM filter according to each target model in the target model set; The corresponding target model performs filtering state estimation to obtain the total estimated output information of the VSIMM filter at the current moment; wherein, the total estimated output information includes the total state estimated value and the total error covariance, and the total state estimated value includes the current state of each target filter. The state estimate at the moment, and the total error covariance includes the error covariance of each target filter at the current moment.

In another possible embodiment, each of the above target models is further configured with a model probability, and the apparatus further includes: updating the model probability of each target model based on the likelihood function of each target model; wherein, by The likelihood function is calculated as follows:

Among them, Λ _j (k) represents the likelihood function of the j-th target model at time k, S _j (k) represents the measurement covariance of the j-th target model at time k, and v _j (k) represents the j-th Kalman filter residuals of the target model at time k.

In another possible embodiment, the above-mentioned first determining module 71 is further configured to: acquire the collection information of the obstacle target; wherein the collection information includes type information, position information and parameter information; determine the obstacle target based on the type information a first motion model set; a second motion model set for determining an obstacle target based on position information; a third motion model set for determining an obstacle target based on parameter information; wherein the parameter information includes at least one of the following: collection device information, obstacle target contour information and weather information; according to the first motion model set, the second motion model set and the third motion model set, determine the target motion model set of the obstacle target.

The filtering device of the intelligent driving perception system provided by the embodiment of the present invention has the same technical features as the filtering method of the intelligent driving perception system provided by the above-mentioned embodiment, so it can also solve the same technical problem and achieve the same technical effect.

Embodiments of the present invention further provide an electronic device, including a processor and a memory, where the memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above filtering method for an intelligent driving perception system.

8 , the electronic device includes a processor 80 and a memory 81, where the memory 81 stores machine-executable instructions that can be executed by the processor 80, and the processor 80 executes the machine-executable instructions to implement the above-mentioned intelligent driving perception system filtering method.

Further, the electronic device shown in FIG. 8 also includes a bus 82 and a communication interface 83 , and the processor 80 , the communication interface 83 and the memory 81 are connected through the bus 82 .

The memory 81 may include a high-speed random access memory (RAM, Random Access Memory), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory. The communication connection between the network element of the system and at least one other network element is realized through at least one communication interface 83 (which may be wired or wireless), which may use the Internet, a wide area network, a local network, a metropolitan area network, and the like. The bus 82 may be an ISA (Industrial Standard Architecture, industry standard architecture bus) bus, a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus, or an EISA (Enhanced Industry Standard Architecture, extended industry standard architecture) bus, or the like. The above-mentioned bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of presentation, only one bidirectional arrow is used in FIG. 8, but it does not mean that there is only one bus or one type of bus.

The processor 80 may be an integrated circuit chip with signal processing capability. In the implementation process, each step of the above-mentioned method can be completed by a hardware integrated logic circuit in the processor 80 or an instruction in the form of software. The above-mentioned processor 80 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU for short), a network processor (Network Processor, NP for short), etc.; it may also be a digital signal processor (Digital Signal Processor, referred to as DSP) , Application Specific Integrated Circuit (ASIC for short), Field-Programmable Gate Array (FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components. Various methods, steps, and logical block diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in conjunction with the embodiments of the present invention may be directly embodied as executed by a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory 81, and the processor 80 reads the information in the memory 81, and completes the steps of the methods of the foregoing embodiments in combination with its hardware.

This embodiment also provides a machine-readable storage medium, where the machine-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are called and executed by the processor, the machine-executable instructions cause the processor to realize the above-mentioned intelligent driving perception System filtering method.

The filtering method and device for an intelligent driving perception system, and the computer program product of the electronic device provided by the embodiments of the present invention include a computer-readable storage medium storing program codes, and the instructions included in the program codes can be used to execute the foregoing method embodiments. For the specific implementation of the method described in , please refer to the method embodiment, which will not be repeated here.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific working process of the system and device described above, reference may be made to the corresponding process in the foregoing method embodiments, which will not be repeated here.

In addition, in the description of the embodiments of the present invention, unless otherwise expressly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it may be a fixed connection or a detachable connection , or integrally connected; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or the internal communication between the two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood in specific situations.

The functions, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a processor-executable non-volatile computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product in essence, or the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including Several instructions are used to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the indicated device or element must have a specific orientation or a specific orientation. construction and operation, and therefore should not be construed as limiting the invention. Furthermore, the terms "first", "second", and "third" are used for descriptive purposes only and should not be construed to indicate or imply relative importance.

Finally, it should be noted that the above-mentioned embodiments are only specific implementations of the present invention, and are used to illustrate the technical solutions of the present invention, but not to limit them. The protection scope of the present invention is not limited thereto, although referring to the foregoing The embodiment has been described in detail the present invention, those of ordinary skill in the art should understand: any person skilled in the art who is familiar with the technical field within the technical scope disclosed by the present invention can still modify the technical solutions described in the foregoing embodiments. Or can easily think of changes, or equivalently replace some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be covered in the present invention. within the scope of protection. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims (10)

1. A filtering method for an intelligent driving perception system, wherein the method is applied to a server of the intelligent driving perception system, wherein the server is provided with a prior knowledge base and variable structure interactive multi-model VSIMM filtering device, the method includes: Determine the target motion model set of the obstacle target based on the motion model in the prior knowledge base, wherein the motion model includes at least one of the following: a uniform velocity CV model, a uniform acceleration CA model, a uniform rotational speed CT model, a current Current models and curvilinear motion CM models; Judging whether the target motion model set is consistent with the model set at the previous moment of the VSIMM filter; wherein, the model set at the last moment of the VSIMM filter includes the VSIMM filter corresponding to each filter at the last moment the motion model; If not, then based on the target motion model set and the model set at the previous moment of the VSIMM filter, determine the target model set of the VSIMM filter at the current moment; Obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs filtering state estimation according to the measurement information and the target model set, and obtains the total estimated output information of the VSIMM filter at the current moment ; The obstacle target is tracked according to the total estimated output information. 2. The filtering method of the intelligent driving perception system according to claim 1, wherein the step of judging whether the target motion model set is consistent with the model set at the last moment of the VSIMM filter, comprising: Judging whether each target motion model in the target motion model set is exactly the same as the motion model corresponding to each filter in the VSIMM filter described in the previous moment; If yes, then determine that the target motion model set is consistent with the model set at the previous moment of the VSIMM filter; If there is any difference, it is determined that the target motion model set is inconsistent with the model set of the VSIMM filter at the previous moment. 3. the filtering method of the intelligent driving perception system according to claim 2, is characterized in that, based on described target motion model set and the model set of the last moment of described VSIMM filter, determine the current moment of described VSIMM filter. The steps of the target model collection, including: According to the union of the target motion model set and the model set of the VSIMM filter at the previous moment, the model set of the VSIMM filter at the current moment is obtained; Obtain the first model probability of the target motion model set and the second model probability of the model set at the previous moment of the VSIMM filter; According to the ratio of the first model probability and the second model probability, the target model set of the VSIMM filter at the current moment is determined. 4. The filtering method of the intelligent driving perception system according to claim 3, wherein the target model set of the VSIMM filter at the current moment is determined according to the ratio of the first model probability and the second model probability steps, including: Determine whether the ratio is greater than the first threshold, and if so, determine that the target model set at the current moment of the VSIMM filter is the target motion model set; If no, judge whether the ratio is less than a second threshold, and if so, determine that the target model set of the VSIMM filter at the current moment is the model set of the VSIMM filter at the previous moment; wherein, the second threshold is less than the first threshold. 5. The filtering method of the intelligent driving perception system according to claim 4, wherein the step of the VSIMM filter performing filtering state estimation according to the measurement information and the target model set, comprising: Determine the corresponding target filter in the VSIMM filter according to each target model in the target model set; The filtering state estimation is performed by each of the target filters according to the measurement information and the corresponding target model, so as to obtain the total estimated output information of the VSIMM filter at the current moment; wherein, the total estimated output information includes the total estimated output information of the VSIMM filter. A state estimate and a total error covariance, the total state estimate includes the state estimate of each of the target filters at the current moment, and the total error covariance includes the error of each of the target filters at the current moment Covariance. 6. The filtering method of the intelligent driving perception system according to claim 5, wherein each of the target models is further configured with a model probability, and the method further comprises: Based on the likelihood function of each target model, the model probability of each target model is updated; wherein, the likelihood function is calculated by the following formula:

Among them, Λ _j (k) represents the likelihood function of the j-th target model at time k, S _j (k) represents the measurement covariance of the j-th target model at time k, and v _j (k) represents the j-th Kalman filter residuals of the target model at time k. 7. The filtering method of the intelligent driving perception system according to claim 1, wherein, based on the motion model in the prior knowledge base, the step of determining the target motion model set of the obstacle target, comprising: Obtain the collection information of the obstacle target; wherein, the collection information includes type information, position information and parameter information; determining a first set of motion models of the obstacle target based on the type information; determining a second set of motion models of the obstacle target based on the location information; A third motion model set of the obstacle target is determined based on the parameter information; wherein the parameter information includes at least one of the following: acquisition device information, obstacle target contour information, and weather information; A target motion model set of the obstacle target is determined according to the first motion model set, the second motion model set and the third motion model set. 8. A filtering device for an intelligent driving perception system, wherein the device is applied to a server of the intelligent driving perception system, wherein the server is provided with a prior knowledge base and a variable-structure interactive multi-model VSIMM filter device, the device includes: The first determination module is used to determine the target motion model set of the obstacle target based on the motion model in the prior knowledge base, wherein the motion model includes at least one of the following: a uniform velocity CV model, a uniform acceleration CA model, Constant speed CT model, current Current model and curve motion CM model; The judgment module is used to judge whether the target motion model set is consistent with the model set of the VSIMM filter at the last moment; wherein, the model set of the VSIMM filter at the last moment includes the VSIMM filter at the last moment. The motion model corresponding to each filter; The second determination module is configured to, if not, determine the target model set of the VSIMM filter at the current moment based on the target motion model set and the model set at the previous moment of the VSIMM filter; A filter estimation module, used to obtain the measurement information of the obstacle target at the current moment, so that the VSIMM filter performs filtering state estimation according to the measurement information and the target model set to obtain the VSIMM filter at the current moment The total estimated output information of the device; A tracking module, configured to track the obstacle target according to the total estimated output information. 9. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements the above claims when executing the computer program Steps of the filtering method for the intelligent driving perception system according to any one of 1-7. 10. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is run by a processor, the intelligent computer program according to any one of the preceding claims 1-7 is executed. Steps of a filtering method for a driving perception system.

Images