Track association method of vehicle

Abstract

The invention relates to a track association method of a vehicle. The method comprises the following steps: acquiring current track data of the vehicle collected by at least two sensors respectively; performing interpolation association processing on the current track data according to a preset public time point to obtain to-be-associated track data corresponding to the current track; constructing topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determining the topological polygon similarity between the topological polygons corresponding to the two different sensors at the same common time point; constructing a similar matrix corresponding to the to-be-associated track data according to the topological polygon similarity, and performing weighting processing on the similar matrix of the to-be-associated track data at each common time point to obtain a total similar matrix of the to-be-associated track data; and screening target topological polygon similarity from the total similarity matrix, and performing association processing on the to-be-associated track data corresponding to the target topological polygon similarity to obtain target associated track data corresponding to the vehicle.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology of automobiles, and in particular to a vehicle track association method. Background Art

[0002] In recent years, with the increasing number and types of sensors on vehicles, more and more information is available for intelligent driving applications. How to fuse and utilize information from different sensors has become an extremely important issue. Due to the influence of factors such as sensor measurement random errors, target distribution, target motion laws, and data processing methods, it is often difficult for the fusion center to accurately associate the tracks of each target. When there are systematic errors between different sensors, the complexity of track association processing is further increased. Systematic errors are mainly composed of two parts: ranging system errors and angular system errors. The ranging system errors usually cause the target track to shift as a whole, while the azimuth system errors mainly cause the target track to rotate as a whole.

[0003] Traditional track correlation algorithms usually assume that the error is zero-mean Gaussian white noise, and by constructing correlation statistics and then using statistical theory methods to make judgments, they often ignore the existence of systematic errors. Since the systematic error is unknown, it is impossible to obtain the threshold of the correlation statistics using statistical methods, which will result in a large number of false associations and missed associations in actual applications, which cannot meet the needs. Summary of the invention

[0004] One of the purposes of the present invention is to provide a vehicle track association method to solve the problem that the track association algorithm in the prior art assumes that the error is Gaussian white noise with zero mean, ignores the influence of system error, and the correct track association rate of the existing algorithm is relatively low in actual engineering applications.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] Acquire current track data of the vehicle collected by at least two sensors respectively;

[0007] Performing interpolation and association processing on each of the current track data according to a preset common time point to obtain track data to be associated corresponding to the current track;

[0008] Constructing the topological polygons corresponding to the sensors at the respective common time points according to the track data to be associated, and determining the similarity of the topological polygons between the topological polygons corresponding to the same common time points of two different sensors;

[0009] Constructing a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and performing weighted processing on the similarity matrix of the track data to be associated at each of the common time points to obtain a total similarity matrix of the track data to be associated;

[0010] The target topology polygon similarity is screened out from the total similarity matrix, and the to-be-associated track data corresponding to the target topology polygon similarity is associated to obtain the target associated track data corresponding to the vehicle.

[0011] Optionally, the obtaining of the current track data of the vehicle respectively collected by at least two sensors includes:

[0012] Acquire original track data of the vehicle collected by at least two sensors respectively, wherein the original track data includes a current position of the vehicle and a time point corresponding to the current position;

[0013] Taking the current position as the center of the circle, the original track data is divided into a plurality of target clusters according to a preset radius;

[0014] Filtering the original track data, marking the original track data that does not belong to the target cluster as noise, and removing the original track data with the noise mark;

[0015] Obtaining a standard lane width and a safe following distance, and setting a rectangular associated wave gate in combination with the standard lane width and the safe following distance;

[0016] The target cluster is framed according to the rectangular correlation gate, and the framed target cluster is used as an actual effective observation target;

[0017] Calculate the equivalent innovation vector corresponding to the actual effective observation target according to the joint probability;

[0018] Obtaining a first filtered value of the actual effective observation target at the previous moment;

[0019] The first filter value is predicted and updated using the equivalent new information vector to obtain current track data of the vehicle.

[0020] Optionally, the using the equivalent innovation vector to perform prediction and update processing on the first filtered value includes:

[0021] Inputting the first filter value into a preset target state transfer equation to obtain a predicted value of the actual effective observed target at the current moment output by the target state transfer equation;

[0022] Obtaining the observation value of the actual effective observation target at the current moment;

[0023] The Kalman gain is calculated based on the predicted value and the observed value;

[0024] Inputting the Kalman gain and the equivalent innovation vector into a preset target state update equation to obtain a second filtered value of the actual effective observation target at the current moment;

[0025] Acquiring historical track data of the vehicle;

[0026] The second filtered value is associated with the historical track data to obtain current track data of the vehicle.

[0027] Optionally, the step of setting a rectangular associated wave gate in combination with the standard lane width and the safe following distance includes:

[0028] The width and height of the rectangular association gate are set according to the standard lane width and the safe following distance, wherein the width is greater than or equal to the standard lane width, and the height is determined according to the safe following distance and the distribution state of the target cluster.

[0029] Optionally, the interpolation association processing is performed on each of the current track data according to a preset common time point, including:

[0030] Performing interpolation processing on the current track data using a preset common time point to obtain an interpolation position corresponding to the common time point on the current track data;

[0031] Calculating the observed distance between the interpolation positions corresponding to each of the two current track data at the same common time point, and comparing the observed distance with a preset maximum observed distance;

[0032] When the observation distance is greater than or equal to the maximum observation distance, dividing the two interpolation positions corresponding to the observation distance into different observation point sets;

[0033] In the case that the observation distance is less than the maximum observation distance, two current track distances corresponding to the observation distance are used as track data to be associated, and the track data to be associated are roughly associated.

[0034] Optionally, the observation point set includes a first observation point set and a second observation point set, and constructing the topological polygon corresponding to the sensor at each of the common time points according to the track data to be associated includes:

[0035] Acquire a first observation point in the first observation point set and a second observation point in the second observation point set, wherein the first observation point and the second observation point are the track data to be associated collected by the two sensors at the same common time point;

[0036] Calculating an actual distance between the first observation point and the second observation point, and calculating a matching degree between the first observation point and the second observation point according to the actual distance;

[0037] Calculating the cumulative matching degree of the first observation point set and the second observation point set according to the matching degree, and determining whether the cumulative matching degree has a maximum value;

[0038] When the cumulative matching degree has the maximum value, taking the observation point in the observation point set as the target observation point;

[0039] Constructing a topological polygon corresponding to each of the common time points of the sensor according to the target observation point;

[0040] The determining of the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point comprises:

[0041] The side length similarity and angle similarity between the topological polygons corresponding to the two different sensors at the same common time point are calculated to obtain the topological polygon similarity of the at least two sensors at the same common time point.

[0042] Optionally, the step of selecting target topology polygon similarity from the total similarity matrix and performing association processing on the to-be-associated track data corresponding to the target topology polygon similarity includes:

[0043] Extracting a maximum similarity value from the total similarity matrix, and determining row data and column data corresponding to the maximum similarity value;

[0044] The to-be-associated track data corresponding to the row data are associated, and the to-be-associated track data corresponding to the column data are associated, so as to obtain target associated track data.

[0045] The present invention discloses a vehicle track association device, the device comprising:

[0046] A track data acquisition module, used to acquire the current track data of the vehicle collected by at least two sensors respectively;

[0047] A track rough association module, used for performing interpolation association processing on each of the current track data according to a preset common time point to obtain track data to be associated corresponding to the current track;

[0048] A similarity calculation module, used to construct the topological polygons corresponding to the sensors at each of the common time points according to the track data to be associated, and determine the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point;

[0049] A similarity matrix generation module, used for constructing a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and performing weighted processing on the similarity matrix of the track data to be associated at each of the common time points to obtain a total similarity matrix of the track data to be associated;

[0050] The track association acquisition module is used to screen out the target topological polygon similarity from the total similarity matrix, and perform association processing on the track data to be associated corresponding to the target topological polygon similarity to obtain the target associated track data corresponding to the vehicle.

[0051] An embodiment of the present invention further discloses an electronic device, comprising at least one processor and a memory communicatively connected to the at least one processor;

[0052] The memory is used to store computer programs;

[0053] The processor is used to implement the method described in the embodiment of the present invention when executing the program stored in the memory.

[0054] The embodiment of the present invention further discloses a computer-readable storage medium on which a computer program is stored. When executed by one or more processors, the processors are enabled to execute the method described in the embodiment of the present invention.

[0055] Beneficial effects of the present invention:

[0056] The embodiment of the present invention obtains the current track data of the vehicle collected by at least two sensors respectively; performs interpolation association processing on each current track data according to a preset common time point to obtain the track data to be associated corresponding to the current track; constructs the topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determines the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point; constructs the similarity matrix corresponding to the track data to be associated according to the topological polygon similarity, and performs weighted processing on the similarity matrix of the track data to be associated at each common time point to obtain the total similarity matrix of the track data to be associated; selects the target topological polygon similarity from the total similarity matrix, and performs association processing on the track data to be associated corresponding to the target topological polygon similarity to obtain the target associated track data corresponding to the vehicle. The present invention makes full use of the topological information of radar measurement and cross-location points, can well suppress the influence of system errors on track association, improve the correct association rate, reduce the false association rate and the missed association rate, and effectively reduces the amount of calculation and saves the operation time through the local association based on the distance information, so as to ensure the real-time performance of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flowchart of a vehicle track association method provided in an embodiment of the present invention;

[0058] FIG. 2( a) is a topological polygon structure diagram constructed in an embodiment of the present invention;

[0059] FIG. 2( b ) is a topological polygon structure diagram constructed in an embodiment of the present invention;

[0060] FIG. 2( c ) is a topological polygonal structure diagram constructed in an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of the structure of a vehicle track association device provided in an embodiment of the present invention;

[0062] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0063] The following will describe the embodiments of the present invention with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0064] In modern intelligent vehicles, sensors are key devices for obtaining environmental information. However, due to various factors, when there are multiple sensors in a vehicle, the data collected by each sensor may have a certain degree of error. For example, different types of sensors, such as lidar, cameras, ultrasonic sensors, etc., have limited measurement accuracy. Lighting, weather, road conditions, etc. will affect the measurement results of the sensor. For example, rainy days will affect the ranging accuracy of lidar. Similarly, the installation position and angle of the sensor on the vehicle will also affect its perception range and accuracy.

[0065] The advantages of millimeter-wave radar are its high resolution, high accuracy, and strong ability to resist active interference. In addition, millimeter-wave radar has a long detection distance, can work all day and all day, and has a strong ability to penetrate fog and dust. However, millimeter-wave radar has difficulties in detecting pedestrians because the reflected waves of pedestrians are easily buried by the reflected waves of other objects, making it difficult to distinguish.

[0066] LiDAR can obtain a three-dimensional model of the environment around the car through laser scanning, and detect, classify and track obstacles through algorithms. In addition, LiDAR also performs well in enhancing positioning, and can synchronously create map information through the point cloud data obtained through scanning. However, the disadvantage of LiDAR is that it is too expensive and its accuracy will decrease in bad weather such as rain, snow, fog and haze. At the same time, LiDAR is also easily affected by other light, such as sunlight and LiDAR of other vehicles.

[0067] Cameras are low-cost and widely used. Through image processing technology, cameras can detect and identify traffic lights, lane detection, pedestrian recognition, road sign recognition and other tasks. However, cameras are highly dependent on lighting conditions and the environment. Changes in lighting conditions will directly affect image quality, which in turn affects the accuracy of object recognition and positioning. In addition, cameras may not be as effective as radar sensors in perceiving distant objects.

[0068] From the above analysis, it can be seen that different sensors such as millimeter wave radar, lidar, camera, etc. have their own characteristics. At the same time, because vehicles may encounter various scenarios during travel, most smart cars with intelligent driving applications now have multiple different sensors, and the number of each sensor is not unique. At this time, after each sensor detects the target, it tracks the target to obtain the track. Therefore, in order to improve the credibility of the data, the embodiment of the present invention associates the track data collected by different sensors to obtain accurate track information.

[0069] Reference Figure 1 , shows a flow chart of a vehicle track association method provided in an embodiment of the present invention, which specifically includes the following steps:

[0070] Step 101, obtaining current track data of the vehicle collected by at least two sensors respectively;

[0071] In the embodiment of the present invention, the current track data of the vehicle collected by at least two sensors are first obtained. Since the sensors may already have inaccuracies and other problems at this time, the track data collected by the sensors cannot be directly used as the current track data. Therefore, accurate current track data can be obtained through the following steps:

[0072] The original track data of the vehicle collected by at least two sensors are obtained, and the original track data includes the current position of the vehicle and the time point corresponding to the current position. For example, the original track data of the vehicle is obtained from at least two sensors (such as millimeter wave radar, laser radar, camera, etc.), and the original track data obtained by each sensor includes the position of the vehicle at this moment and the corresponding timestamp.

[0073] With the current position as the center, the original track data is divided into several target clusters according to the preset time radius or distance radius. The target cluster can be used to identify and distinguish the track data of different targets.

[0074] The original track data is filtered, the original track data that does not belong to the target cluster is marked as noise, and the original track data with noise marks is eliminated. The original track data is filtered, the data that does not belong to the target cluster is marked as noise, and the noise data is eliminated. For example, the DBSCAN algorithm (Density-Based Spatial Clustering of Applications with Noise) can be used to cluster each frame of track data and preliminarily remove noise points. First, the target position information is used as the input target center, and a clustering radius is set. The target center is clustered according to the set clustering radius to obtain a target cluster corresponding to the target center. For some track data, it does not belong to any target cluster. Such track data is marked as noise and used as noise points. All noise points are removed. These noise points are usually isolated and low-density points. The clustering radius can be a time radius or a distance radius. The present invention does not make specific restrictions on this. If the clustering radius is a time radius, then the track data collected within the time radius around the target center are divided into the same category. If the clustering radius is a distance radius, then the track data collected within the distance radius around the target center are divided into the same category.

[0075] Obtain the standard lane width and safe following distance, and set a rectangular association gate based on the standard lane width and safe following distance. The gate defines an area, and only the target clusters falling within this area are considered valid observation targets. Select the target clusters based on the rectangular association gate, and use the selected target clusters as actual valid observation targets. Specifically, according to the standard lane width and safe following distance, set the width and height of the rectangular association gate, where the width is greater than or equal to the standard lane width, and determine the height based on the safe following distance and the distribution status of the target clusters.

[0076] The equivalent new information vector corresponding to the actual effective observation target is calculated based on the joint probability. The new information vector reflects the difference between the observed data and the predicted data. The joint probability refers to the probability of combining the actual effective observation target with other effective data. The joint probability can be calculated by Bayes' theorem. The present invention does not specifically limit the method of obtaining the joint probability.

[0077] The first filtered value of the actual effective observed target at the previous moment is obtained. The filtered value can be obtained based on historical data and model prediction. The present invention does not specifically limit the method for obtaining the filtered value.

[0078] The first filter value is predicted and updated using the equivalent innovation vector to obtain the current track data of the vehicle, which specifically includes the following steps:

[0079] The first filter value is input into the preset target state transfer equation to obtain the predicted value of the actual effective observed target at the current moment output by the target state transfer equation. The target state equation is as follows:

[0080] X(k+1)=F(k)X(k)+G(k)w(k)

[0081] Among them, X(k+1) is the state vector of the target at time (k+1), which contains information such as position and velocity; F(k) is the state transfer matrix, which describes how the state is transferred from time k-1 to time k; x(k) is the state vector of the target at time k; G(k) is the control input matrix, which describes the influence of external control input (such as acceleration) on the state; w(k) is the process noise vector, which represents the random noise introduced in the state transfer process.

[0082] On the other hand, the predicted value is compared with the actual value using the measurement equation, which is as follows:

[0083]

[0084] Among them, Z(k) is the measurement vector of the sensor at time k, including measurement values ​​such as position and velocity; H(k) is the measurement matrix, which describes which components of the state vector X(k) can be directly measured by the sensor; X(k) is the state vector of the target at time k, including information such as position and velocity; v(k) is the measurement noise vector, which represents the random noise introduced during the sensor measurement process.

[0085] Obtain the actual effective observation value of the target at the current moment. For example, obtain the actual effective observation value Z(k) of the target at the current moment from the sensor. This observation value directly reflects the current state of the target. The Kalman gain is calculated based on the predicted value and the observed value.

[0086] The Kalman gain and the equivalent new information vector are input into the preset target state update equation to obtain the second filtered value of the actual effective observed target at the current moment.

[0087] The historical track data of the vehicle is acquired, and the second filter value is associated with the historical track data to obtain the current track data of the vehicle.

[0088] The embodiment of the present invention effectively combines the current observation value and the historical prediction value through Kalman filtering, gradually corrects the state estimation, and obtains more accurate current track data.

[0089] Step 102, performing interpolation association processing on each current track data according to a preset common time point to obtain track data to be associated corresponding to the current track, specifically includes the following steps:

[0090] The current track data is interpolated using the preset common time point to obtain the interpolation position corresponding to the common time point on the current track data, so that the two current track data have the same sampling point in the subsequent calculation and processing.

[0091] Calculate the observed distance between the interpolation positions corresponding to each of the two current track data at the same common time point, and compare the observed distance with the preset maximum observation distance.

[0092] When the observation distance is greater than or equal to the maximum observation distance, the two interpolation positions corresponding to the observation distance are divided into different observation point sets.

[0093] When the observation distance is less than the maximum observation distance, the two current track distances corresponding to the observation distance are used as track data to be associated, and the track data to be associated are roughly associated.

[0094] Step 103 , constructing topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determining the topological polygon similarity between topological polygons corresponding to two different sensors at the same common time point.

[0095] After data preprocessing and rough association, the next step is to perform track association based on the topological structure. In order to ensure that the number of vertices of the constructed topological polygons is the same, it is necessary to screen the observation points around the two targets to be associated at the same time, remove the observations caused by clutter and false alarms, and then construct polygons with the same vertices according to the rules, and then calculate the similarity of the two topological polygons of the targets to be associated.

[0096] The observation point set includes a first observation point set and a second observation point set. The first observation point set refers to the track data to be associated collected by the first sensor, and the second observation point set refers to the track data to be associated collected by the second sensor. The first sensor and the second sensor are different sensors. Specifically, they can be two sensors of different types or two sensors of the same type. The present invention does not make any specific limitation to this.

[0097] The first observation point in the first observation point set and the second observation point in the second observation point set are obtained, wherein the first observation point and the second observation point are the track data to be associated collected by two sensors at the same common time point. For example, the first observation point A1 in the first observation point set A and the second observation point B1 in the second observation point set B, wherein A1 and B1 are the track data to be associated collected by two sensors at the same time.

[0098] The actual distance between the first observation point and the second observation point is calculated, and the matching degree between the first observation point and the second observation point is calculated based on the actual distance. i With each observation point B in the observation point set B j The actual distance between them, assuming point A i and point B j The distance between them is D ij , and calculate the matching degree between each pair of observation points based on the actual distance. The higher the matching degree, the more matching the two points are.

[0099] The cumulative matching degree of the first observation point set and the second observation point set is calculated based on the matching degree, and it is determined whether the cumulative matching degree has a maximum value. The matching degrees calculated for each pair of points are added together, and the total value obtained is the cumulative matching degree of the first observation point set and the second observation point set. In one embodiment, the Hungarian algorithm can be used to solve the maximum matching degree of this cumulative matching degree. In the case where the cumulative matching degree has a maximum value, all observation points in the observation point set are retained, and the observation points in the observation point set are used as target observation points. The topological polygons corresponding to the sensor at each common time point are constructed based on the target observation point.

[0100] At the tth sampling time, after the rough correlation, the possible associated tracks of the two sensors are screened out. Within the same rectangular correlation gate, T i (t) and T j (t) are the observation points observed by sensor A and sensor B at time t. i (t) N is observed around i A goal, remember to do Sensor B is at target T j (t) N is observed around j A goal, remember to do

[0101] If N i =N j , we can directly construct the topological polygons corresponding to the sensors at each common time point.

[0102] If N i ≠N j In a feasible embodiment of the present invention, due to the missed alarm of sensor B, the number of targets observed by the two sensors may be inconsistent. At this time, all target observation points of sensor B will be observed by sensor A, that is, N i >N j .

[0103] At this time, it is necessary to remove the target observation points that do not exist in observation point set B from observation point set A to keep the number of target observation points in both sets consistent.

[0104] Calculate all observation points A in point set A separately m and all observation points B in point set B n The actual distance D mn :

[0105]

[0106] Among them, x m ,y m is the observation point A in the point set A m Coordinates in the Cartesian coordinate system, x n ,y n is the observation point B in point set B m Coordinates in the Cartesian coordinate system.

[0107] Calculate the matching degree γ between each pair of points mn :

[0108]

[0109] The target observation points in the two observation point sets are matched based on the matching degree. For each pair of matched target observation points, the cumulative matching degree Ψ is calculated:

[0110]

[0111] The variable λ mn Indicates the target observation point A m and B n Match:

[0112]

[0113] Among them, assuming that H1 is A m and B n Match, assuming H2 is A m and B n No match.

[0114] In the embodiment of the present invention, for each target observation point, there is only one corresponding target observation point in another sensor observation point set that matches it, so the matrix λ=[λ mn ]There is only one element in each row and column that is 1, and the rest are 0.

[0115] In another embodiment, when Ψ is the largest, the current matching result is selected as the optimal matching result, and the target observation point of the current matching is retained to construct a topological polygon, and the similarity of the topological polygon is calculated.

[0116] As shown in Figure 2(a), when constructing a topological polygon, i Taking the topological polygon of (t) as an example, the specific construction process is as follows:

[0117] From T i (t) Find the highest point among the surrounding targets as the first point of the polygon (point P1 in the figure), connect all target observation points except the first point (points P2, P3, P4 and P5 in the figure) with the first point (point P1 in the figure), take the north coordinate axis as the starting point, and calculate the angle between each remaining point and the first point in the clockwise direction.

[0118] As shown in Figure 2(b), for T i (t) Sort the angles of the surrounding targets (other points except point P1 in the figure), connect the points in the order of the angle from small to large, and then use T i (t) (point P2 in the figure) is the starting point. The first edge encountered in clockwise rotation is marked as 1. Each edge is numbered in turn in the clockwise direction. Then, as shown in Figure 2(c), the vertex is T i The inner angles of (t) are numbered clockwise from the starting point to complete the topological polygon.

[0119] In an embodiment of the present invention, determining the topological polygon similarity between topological polygons corresponding to two different sensors at the same common time point includes:

[0120] The side length similarity and angle similarity between the topological polygons corresponding to two different sensors at the same common time point are calculated to obtain the topological polygon similarity of at least two sensors at the same common time point.

[0121] Define the target T i (t) and T j (t) Edge length similarity coefficient at time t and angular similarity coefficient

[0122]

[0123] in, and They are topological polygons A t The length of the kth side and the angle of the kth interior angle. Similarly, and They are topological polygons B t The length of the kth side and the angle of the kth interior angle, n is the topological polygon A t The number of interior angles.

[0124] In another embodiment of the present invention, in order to avoid a large difference in the magnitude of the side length and the angle, the relative difference Δl between the length and the angle is calculated. k and Take the relative difference of side lengths as an example:

[0125]

[0126] The mean of the relative differences in their side lengths is l mean and variance for:

[0127]

[0128] The above formula is still applicable to angles, so for T i (t) and T j (t) The topological polygon similarity F(i,j,t) at time t is:

[0129]

[0130] in, is the variance of the angle.

[0131] Step 104 , constructing a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and performing weighted processing on the similarity matrix of the track data to be associated at each common time point to obtain a total similarity matrix of the track data to be associated.

[0132] Two different sensors have M and N tracks in the same correlation gate. After calculating the similarity of the topological polygons between each pair of tracks at time t, an M×N similarity matrix can be obtained. If there are T sampling times, the final total similarity matrix is ​​obtained by weighted averaging the T similarity matrices:

[0133]

[0134] Step 105, screening out the target topology polygon similarity from the total similarity matrix, and performing association processing on the to-be-associated track data corresponding to the target topology polygon similarity to obtain the target associated track data corresponding to the vehicle, specifically includes the following steps:

[0135] The maximum similarity value is extracted from the total similarity matrix, and the row data and column data corresponding to the maximum similarity value are determined.

[0136] The to-be-associated track data corresponding to the row data are associated, and the to-be-associated track data corresponding to the column data are associated, so as to obtain the target associated track data.

[0137] F(i, j) is the similarity between the i-th track observed by sensor A and the j-th track observed by sensor B. Set the association threshold θ, traverse all elements of the matrix, find the maximum value in the element, and compare it with the association threshold θ. If the value of the element is greater than or equal to the threshold θ, the i-th track of sensor A and the j-th track of sensor B are associated with each other. Because a track can only be uniquely matched with a track of other sensors, after associating the i-th row and j-th column, set all elements in the i-th row and j-th column to 0, find the new maximum value in the matrix at this time, and compare the new maximum value with the association threshold. If the new maximum value is greater than or equal to the association threshold, associate the row and column where the new maximum value is located and set it to 0. Repeat the cycle until all element values ​​in the matrix are less than the association threshold, that is, there is no qualified association data, and the track association is completed at this time.

[0138] After the vehicle leaves the factory, as the vehicle is used daily, the sensor position may shift slightly, causing offset and rotation to the tracks of all targets. Therefore, before online calibration, the embodiment of the present invention associates the tracks of multiple sensors in the vehicle to avoid the impact of system errors and improve the track accuracy after fusion. At the same time, it can increase the application scenarios of real-vehicle intelligent driving, enable it to cope with more complex weather conditions, and obtain more target information for the same target, providing more usable information for subsequent planning and decision-making.

[0139] The embodiment of the present invention obtains the current track data of the vehicle collected by at least two sensors respectively; performs interpolation association processing on each current track data according to a preset common time point to obtain the track data to be associated corresponding to the current track; constructs the topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determines the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point; constructs the similarity matrix corresponding to the track data to be associated according to the topological polygon similarity, and performs weighted processing on the similarity matrix of the track data to be associated at each common time point to obtain the total similarity matrix of the track data to be associated; selects the target topological polygon similarity from the total similarity matrix, and performs association processing on the track data to be associated corresponding to the target topological polygon similarity to obtain the target associated track data corresponding to the vehicle. The present invention makes full use of the topological information of radar measurement and cross-location points, can well suppress the influence of system errors on track association, improve the correct association rate, reduce the false association rate and the missed association rate, and effectively reduces the amount of calculation and saves the operation time through the local association based on the distance information, so as to ensure the real-time performance of the method.

[0140] like Figure 3 As shown, the present invention discloses a vehicle track association device, the device comprising:

[0141] A track data acquisition module 301 is used to acquire the current track data of the vehicle collected by at least two sensors respectively;

[0142] The track rough association module 302 is used to perform interpolation association processing on each current track data according to a preset common time point to obtain the track data to be associated corresponding to the current track;

[0143] A similarity calculation module 303 is used to construct the topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determine the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point;

[0144] The similarity matrix generation module 304 is used to construct a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and to perform weighted processing on the similarity matrix of the track data to be associated at each common time point to obtain a total similarity matrix of the track data to be associated;

[0145] The track association acquisition module 305 is used to filter out the target topology polygon similarity from the total similarity matrix, and perform association processing on the track data to be associated corresponding to the target topology polygon similarity to obtain the target associated track data corresponding to the vehicle.

[0146] The track data acquisition module 301 includes:

[0147] A raw data acquisition module, used to acquire raw track data of the vehicle collected by at least two sensors respectively, the raw track data including the current position of the vehicle and the time point corresponding to the current position;

[0148] The raw data division module is used to divide the raw track data into several target clusters according to a preset radius with the current position as the center;

[0149] The raw data filtering module is used to filter the raw track data, mark the raw track data that does not belong to the target cluster as noise, and remove the raw track data with noise marks;

[0150] A correlation gate generation module is used to obtain the standard lane width and the safe following distance, and to set a rectangular correlation gate based on the standard lane width and the safe following distance;

[0151] The frame selection module is used to frame the target cluster according to the rectangular correlation wave gate, and take the framed target cluster as the actual effective observation target;

[0152] A calculation module is used to calculate the equivalent innovation vector corresponding to the actual effective observation target based on the joint probability;

[0153] A filter value acquisition module is used to obtain the first filter value of the actual effective observation target at the previous moment;

[0154] The filter value update module is used to use the equivalent new information vector to perform predictive update processing on the first filter value to obtain the current track data of the vehicle.

[0155] Filter value update module, including:

[0156] A first prediction unit is used to input a first filter value into a preset target state transfer equation to obtain a predicted value of an actual effective observed target output by the target state transfer equation at a current moment;

[0157] A first acquisition unit is used to acquire the observation value of the actual effective observation target at the current moment;

[0158] A first calculation unit, used for calculating a Kalman gain according to a predicted value and an observed value;

[0159] A second calculation unit is used to input the Kalman gain and the equivalent new information vector into a preset target state update equation to obtain a second filtered value of the actual effective observed target at the current moment;

[0160] A second acquisition unit, used to acquire historical track data of the vehicle;

[0161] The first updating unit is used to associate the second filter value with the historical track data to obtain the current track data of the vehicle.

[0162] The associated wave gate generation module is specifically used to set the width and height of the rectangular associated wave gate according to the standard lane width and the safe following distance, wherein the width is greater than or equal to the standard lane width, and the height is determined according to the safe following distance and the distribution status of the target cluster.

[0163] The track coarse association module 302 includes:

[0164] An interpolation processing module is used to perform interpolation processing on the current track data using a preset common time point to obtain an interpolation position corresponding to the common time point on the current track data;

[0165] A distance calculation module is used to calculate the observed distance between the interpolation positions corresponding to each two current track data at the same common time point, and compare the observed distance with a preset maximum observed distance;

[0166] A point set division module, used for dividing two interpolation positions corresponding to the observation distance into different observation point sets when the observation distance is greater than or equal to the maximum observation distance;

[0167] The coarse association submodule is used to use the two current track distances corresponding to the observation distances as track data to be associated when the observation distance is less than the maximum observation distance, and to perform coarse association on the track data to be associated.

[0168] The similarity calculation module 303 includes:

[0169] An observation point acquisition module, used to acquire a first observation point in a first observation point set and a second observation point in a second observation point set, wherein the first observation point and the second observation point are track data to be associated collected by two sensors at the same common time point;

[0170] A matching degree calculation module, used to calculate the actual distance between the first observation point and the second observation point, and calculate the matching degree between the first observation point and the second observation point according to the actual distance;

[0171] A maximum value determination module, used to calculate the cumulative matching degree of the first observation point set and the second observation point set according to the matching degree, and determine whether the cumulative matching degree has a maximum value;

[0172] A target observation point determination module is used to select an observation point in the observation point set as a target observation point when the cumulative matching degree has a maximum value;

[0173] The topological polygon construction module is used to construct the topological polygons corresponding to the sensors at each common time point based on the target observation points;

[0174] The similarity calculation module 303 further includes:

[0175] The similarity calculation unit is used to calculate the side length similarity and angle similarity between the topological polygons corresponding to two different sensors at the same common time point, and obtain the topological polygon similarity of at least two sensors at the same common time point.

[0176] The track association acquisition module 305 includes:

[0177] An extraction module is used to extract the maximum similarity value from the total similarity matrix and determine the row data and column data corresponding to the maximum similarity value;

[0178] The association module is used to perform association processing on the track data to be associated corresponding to the row data, and to perform association processing on the track data to be associated corresponding to the column data, so as to obtain target associated track data.

[0179] The embodiment of the present invention obtains the current track data of the vehicle collected by at least two sensors respectively; performs interpolation association processing on each current track data according to a preset common time point to obtain the track data to be associated corresponding to the current track; constructs the topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determines the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point; constructs the similarity matrix corresponding to the track data to be associated according to the topological polygon similarity, and performs weighted processing on the similarity matrix of the track data to be associated at each common time point to obtain the total similarity matrix of the track data to be associated; selects the target topological polygon similarity from the total similarity matrix, and performs association processing on the track data to be associated corresponding to the target topological polygon similarity to obtain the target associated track data corresponding to the vehicle. The present invention makes full use of the topological information of radar measurement and cross-location points, can well suppress the influence of system errors on track association, improve the correct association rate, reduce the false association rate and the missed association rate, and effectively reduces the amount of calculation and saves the operation time through the local association based on the distance information, so as to ensure the real-time performance of the method.

[0180] The embodiment of the present invention further provides an electronic device, such as Figure 4 As shown, it includes a processor 401, a device interface 402, a memory 403 and a bus 404;

[0181] The memory 403 is used to store computer programs.

[0182] The processor 401 is used to execute the program stored in the memory 403 to implement the following steps:

[0183] Acquire current track data of the vehicle collected by at least two sensors respectively;

[0184] Performing interpolation and association processing on each current track data according to a preset common time point to obtain track data to be associated corresponding to the current track;

[0185] Constructing the topological polygons corresponding to the sensors at each common time point according to the track data to be associated, and determining the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point;

[0186] A similarity matrix corresponding to the track data to be associated is constructed based on the similarity of topological polygons, and the similarity matrix of the track data to be associated at each common time point is weighted to obtain the total similarity matrix of the track data to be associated;

[0187] The target topology polygon similarity is screened out from the total similarity matrix, and the to-be-associated track data corresponding to the target topology polygon similarity is associated to obtain the target associated track data corresponding to the vehicle.

[0188] The bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0189] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0190] The above-mentioned processor can be a general-purpose processor, including a central processing unit, a network processor, etc.; it can also be a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component.

[0191] The present invention also provides a storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the vehicle track association method of the aforementioned embodiment.

[0192] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0193] The algorithm and display provided herein are not inherently related to any particular computer, virtual device or other equipment. According to the above description, it is obvious that the structure required for constructing this type of device is. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific language is to disclose the best mode of the present invention.

[0194] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0195] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the following intention: that the claimed invention requires more features than those expressly recited in each claim. Rather, as reflected in the claims below, inventive aspects lie in less than all of the features of the individual embodiments disclosed above. Therefore, the claims that follow the detailed description are hereby expressly incorporated into the detailed description, with each claim itself serving as a separate embodiment of the present invention.

[0196] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0197] The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor may be used in practice to implement some or all of the functions of some or all of the components in the sorting device according to the present invention. The present invention may also be implemented as a device or apparatus program for executing part or all of the methods described herein. Such a program for implementing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0198] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc., does not indicate any order. These words may be interpreted as names.

[0199] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices, apparatuses and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0200] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0201] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0202] It should be noted that the various data-related processes in the embodiments of the present invention are all carried out in compliance with the corresponding data protection laws and policies of the country where the device is located and with the authorization given by the corresponding device owner.

Claims

1. A vehicle track association method, characterized in that: The method comprises: Acquire current track data of the vehicle collected by at least two sensors respectively; Performing interpolation and association processing on each of the current track data according to a preset common time point to obtain track data to be associated corresponding to the current track; Constructing the topological polygons corresponding to the sensors at the respective common time points according to the track data to be associated, and determining the similarity of the topological polygons between the topological polygons corresponding to the same common time points of two different sensors; Constructing a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and performing weighted processing on the similarity matrix of the track data to be associated at each of the common time points to obtain a total similarity matrix of the track data to be associated; The target topology polygon similarity is screened out from the total similarity matrix, and the to-be-associated track data corresponding to the target topology polygon similarity is associated to obtain the target associated track data corresponding to the vehicle.

2. The method according to claim 1, characterized in that The step of obtaining the current track data of the vehicle respectively collected by at least two sensors includes: Acquire original track data of the vehicle collected by at least two sensors respectively, wherein the original track data includes a current position of the vehicle and a time point corresponding to the current position; Taking the current position as the center of the circle, the original track data is divided into a plurality of target clusters according to a preset radius; Filtering the original track data, marking the original track data that does not belong to the target cluster as noise, and removing the original track data with the noise mark; Obtaining a standard lane width and a safe following distance, and setting a rectangular correlation gate in combination with the standard lane width and the safe following distance; selecting the target cluster according to the rectangular correlation gate, and using the selected target cluster as an actual effective observation target; Calculate the equivalent innovation vector corresponding to the actual effective observation target according to the joint probability; Obtaining a first filtered value of the actual effective observation target at the previous moment; The first filter value is predicted and updated using the equivalent new information vector to obtain current track data of the vehicle.

3. The method according to claim 2, characterized in that The using the equivalent innovation vector to perform prediction and update processing on the first filter value includes: Inputting the first filter value into a preset target state transfer equation to obtain a predicted value of the actual effective observed target at the current moment output by the target state transfer equation; Obtaining the observation value of the actual effective observation target at the current moment; The Kalman gain is calculated based on the predicted value and the observed value; Inputting the Kalman gain and the equivalent innovation vector into a preset target state update equation to obtain a second filtered value of the actual effective observation target at the current moment; Acquiring historical track data of the vehicle; The second filtered value is associated with the historical track data to obtain current track data of the vehicle.

4. The method according to claim 2, characterized in that: The step of setting a rectangular associated wave gate in combination with the standard lane width and the safe following distance includes: The width and height of the rectangular association gate are set according to the standard lane width and the safe following distance, wherein the width is greater than or equal to the standard lane width, and the height is determined according to the safe following distance and the distribution state of the target cluster.

5. The method according to claim 1, characterized in that The interpolation association processing is performed on each of the current track data according to the preset common time point, including: Performing interpolation processing on the current track data using a preset common time point to obtain an interpolation position corresponding to the common time point on the current track data; Calculating the observed distance between the interpolation positions corresponding to each of the two current track data at the same common time point, and comparing the observed distance with a preset maximum observed distance; When the observation distance is greater than or equal to the maximum observation distance, dividing the two interpolation positions corresponding to the observation distance into different observation point sets; In the case that the observation distance is less than the maximum observation distance, two current track distances corresponding to the observation distance are used as track data to be associated, and the track data to be associated are roughly associated.

6. The method according to claim 5, characterized in that The observation point set includes a first observation point set and a second observation point set, and constructing the topological polygon corresponding to the sensor at each of the common time points according to the track data to be associated includes: Acquire a first observation point in the first observation point set and a second observation point in the second observation point set, wherein the first observation point and the second observation point are the track data to be associated collected by the two sensors at the same common time point; Calculating an actual distance between the first observation point and the second observation point, and calculating a matching degree between the first observation point and the second observation point according to the actual distance; Calculating the cumulative matching degree of the first observation point set and the second observation point set according to the matching degree, and determining whether the cumulative matching degree has a maximum value; When the cumulative matching degree has the maximum value, taking the observation point in the observation point set as the target observation point; Constructing a topological polygon corresponding to each of the common time points of the sensor according to the target observation point; The determining of the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point comprises: The side length similarity and angle similarity between the topological polygons corresponding to the two different sensors at the same common time point are calculated to obtain the topological polygon similarity of the at least two sensors at the same common time point.

7. The method according to claim 1, characterized in that The step of selecting target topology polygon similarity from the total similarity matrix and performing association processing on the to-be-associated track data corresponding to the target topology polygon similarity includes: Extracting a maximum similarity value from the total similarity matrix, and determining row data and column data corresponding to the maximum similarity value; The to-be-associated track data corresponding to the row data are associated, and the to-be-associated track data corresponding to the column data are associated, so as to obtain target associated track data.

8. A vehicle track association device, characterized in that: The device comprises: A track data acquisition module, used to acquire the current track data of the vehicle collected by at least two sensors respectively; A track rough association module, used for performing interpolation association processing on each of the current track data according to a preset common time point to obtain track data to be associated corresponding to the current track; A similarity calculation module, used to construct the topological polygons corresponding to the sensors at each of the common time points according to the track data to be associated, and determine the topological polygon similarity between the topological polygons corresponding to two different sensors at the same common time point; A similarity matrix generation module, used for constructing a similarity matrix corresponding to the track data to be associated according to the similarity of the topological polygons, and performing weighted processing on the similarity matrix of the track data to be associated at each of the common time points to obtain a total similarity matrix of the track data to be associated; The track association acquisition module is used to screen out the target topological polygon similarity from the total similarity matrix, and perform association processing on the track data to be associated corresponding to the target topological polygon similarity to obtain the target associated track data corresponding to the vehicle.

9. An electronic device, characterized in that: include: at least one processor; a memory communicatively coupled to the at least one processor; The memory has a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the vehicle track association method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the vehicle track association method according to any one of claims 1 to 7 is implemented.

Images