Target matching method, device and equipment and storage medium

Abstract

The present disclosure provides a target matching method and device, equipment and storage medium, relates to the technical field of artificial intelligence, in particular to the technical field of image processing, computer vision and deep learning, and especially to target detection, intelligent transportation and the like. The specific implementation scheme is as follows: obtaining target detection data of each to-be-tracked target in different groups of to-be-tracked targets; wherein different groups of to-be-tracked targets correspond to different target detection devices, and the detection regions of different target detection devices are the same; determining matching weights between corresponding to-be-tracked targets according to the target detection data of to-be-tracked targets in different groups; and performing target matching on to-be-tracked targets in different groups according to the matching weights. According to the technology of the present disclosure, the accuracy of the target matching result is improved.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of artificial intelligence, in particular to the technical field of image processing, computer vision and deep learning, and more particularly to target detection and intelligent transportation scenarios. BACKGROUND

[0002] Multi-target tracking technology is widely used in smart city, intelligent transportation, robot navigation, intelligent monitoring video, industrial detection, aerospace and autonomous driving fields. At present, most of the multi-target tracking methods only maintain the target trajectory under the current sensor (such as a camera), and cannot achieve fusion tracking under multiple sensors. SUMMARY

[0003] The present disclosure provides a target matching method, device, equipment and storage medium to improve target matching accuracy and realize fusion tracking of targets.

[0004] According to an aspect of the present disclosure, a target matching method is provided, comprising:

[0005] Obtaining target detection data of each target to be tracked in different groups of targets to be tracked; wherein different target detection devices correspond to different groups of targets to be tracked, and the detection regions of different target detection devices are the same;

[0006] Determining matching weights between corresponding targets to be tracked according to the target detection data of the targets to be tracked in different groups;

[0007] Performing target matching on the targets to be tracked in different groups according to the matching weights.

[0008] According to another aspect of the present disclosure, an electronic device is also provided, comprising:

[0009] At least one processor; and

[0010] A memory connected in communication with the at least one processor; wherein

[0011] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform any one of the target matching methods provided by the embodiments of the present disclosure.

[0012] According to another aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is also provided, wherein the computer instructions are used to enable a computer to perform any one of the target matching methods provided by the embodiments of the present disclosure.

[0013] According to another aspect of the present disclosure, a computer program product is also provided, which comprises a computer program, and the computer program, when executed by a processor, implements any one of the target matching methods provided by the embodiments of the present disclosure.

[0014] According to the technology of the present disclosure, the accuracy of the target matching result is improved.

[0015] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are used to better understand the present scheme, and do not limit the present disclosure. Among them:

[0017] Figure 1 is a schematic diagram of a target matching method according to an embodiment of the present disclosure;

[0018] Figure 2 is a schematic diagram of another target matching method according to an embodiment of the present disclosure;

[0019] Figure 3 is a schematic diagram of another target matching method according to an embodiment of the present disclosure;

[0020] Figure 4A is a schematic diagram of another target matching method according to an embodiment of the present disclosure;

[0021] Figure 4B is a schematic diagram of a target fusion and target tracking of a target to be tracked according to an embodiment of the present disclosure;

[0022] Figure 4C is a schematic diagram of a multi-target detection device fusion tracking framework according to an embodiment of the present disclosure;

[0023] Figure 5 is a structural diagram of a target matching device according to an embodiment of the present disclosure;

[0024] Figure 6 is a block diagram of an electronic device for implementing the target matching method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0025] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, which are cited as illustrative examples. Various details of the embodiments of the present disclosure are described herein in order to provide a thorough understanding of the embodiments. It will be understood by those of ordinary skill in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Also, for the sake of brevity and clarity, descriptions of well-known functions and constructions are omitted herein.

[0026] The target matching method and the target matching device provided by the embodiments of the present disclosure are suitable for the application scenarios of matching, further fusing and tracking the targets collected by multiple sensors under the same point position region. The target matching method provided by the embodiments of the present disclosure can be executed by a target matching device, which can be implemented by software and / or hardware and specifically configured in an electronic device. The electronic device can be a target matching device or other computing device associated with the target matching device.

[0027] For the sake of convenience, the target matching method provided by the embodiments of the present disclosure is described in detail first.

[0028] Referring to FIG. 1, a target matching method is shown, which comprises the following steps. Figure 1

[0029] In S110, target detection data of each target to be tracked in different target to be tracked groups is acquired; wherein the different target to be tracked groups correspond to different target detection devices, and the detection regions of the different target detection devices are the same.

[0030] It should be noted that in the intelligent transportation system, multiple target detection devices can be deployed under the same detection region to collect the target to be tracked in the detection region from different directions. The target detection device can be a gun camera (for example, a long-focus gun camera or a short-focus gun camera), a fisheye camera or a millimeter wave radar, etc. The target to be tracked can be a pedestrian or a vehicle of different types passing through the detection region, etc.

[0031] The target to be tracked group can be a target group corresponding to the image collected by any target detection device in the same detection region at the same time. Different target to be tracked groups can correspond to different target detection devices, and the detection regions of the different target detection devices are the same. Since the target detection devices in the same detection region are deployed at different positions, the targets to be tracked collected by the different target detection devices are at least partially the same, but the target detection data corresponding to the targets to be tracked collected by the different target detection devices is different.

[0032] ​The target detection data can include at least one of position detection data, category detection data, and size detection data. The position detection data can be a relative position of the target to be tracked in a detection area. The category detection data can include a category to which the target to be tracked belongs and a category confidence score. The category of the target to be tracked can include at least one of a car, a truck, a bus, a trolleybus, a motorcycle, a bicycle, and a pedestrian. The size detection data can be an actual size (e.g., length, width, and height) of the target to be tracked in a world coordinate system.

[0033] The target detection data of each target to be tracked can be determined by a convolutional neural network. For example, a pre-constructed convolutional neural network model can be a PPyolov2 model. The target detection data can be determined by training the pre-constructed convolutional neural network model using a sample training set including target detection images of the target to be tracked in a historical period. The target detection images can be single-frame RGB (Red Green Blue) images. The target detection images can be images pre-labeled with position labels and category labels. The position labels can be 2D (dimensional) bounding box positions (coordinates of length, width, and height) of each target in the target detection images. The category labels can include a category to which each target in the target detection images belongs and a confidence score of the category.

[0034] For example, in the process of obtaining the target detection data, the detection images obtained by the target detection devices corresponding to the target to be tracked can be input into the target detection model to obtain the target detection data of each target to be tracked in the detection images.

[0035] It should be noted that, since the deployment positions of the target detection devices in the same detection area are different, in order to ensure the consistency of the position detection data of the same target to be detected in the target detection data collected by each target detection device, the 2D coordinates in the image coordinate system can be converted into 3D coordinates in the world coordinate system according to the calibration parameters of the target detection device and the ground equation.

[0036] For any target detection device, the manner of tracking the to-be-tracked target can be to continuously optimize and update the target detection data of the to-be-tracked target by using a DeepSORT (Deep SORT) multi-target tracking algorithm. For example, a bounding box can be set for each to-be-tracked target, and a small image enclosed by the bounding box of each to-be-tracked target is input into a convolutional neural network to obtain a ReID (Re-identification, re-identification) feature vector of the to-be-tracked target. The feature vector can represent the appearance information of the to-be-tracked target. The ReID feature vector and the position information are combined to associate and match the detection result of the current frame with the tracking result of the previous frame, thereby updating the target detection data of the to-be-tracked target.

[0037] It can be understood that, due to the characteristics of easy occlusion, short distance, few dimensions, and large noise of the target detection device, the target detection data determined by a single target detection device can have certain errors, and the errors of different target detection devices can be different. After matching and fusing the target detection data of different target detection devices corresponding to the same detection region, the tracking accuracy of the to-be-tracked target can be improved to a certain extent.

[0038] S120, determining a matching weight between the to-be-tracked targets in the different groups according to the target detection data of the to-be-tracked targets.

[0039] The matching weight between the to-be-tracked targets corresponding to different groups is used to represent the probability that the two to-be-tracked targets matched are the same target.

[0040] Determining the matching weight between the to-be-tracked targets can include: constructing a connection edge between the to-be-tracked targets according to the target detection data of each to-be-tracked target, and taking the weight corresponding to the connection edge as the matching weight between the to-be-tracked targets.

[0041] The matching weight between the to-be-tracked targets can be determined by a related technical person according to actual experience values or test values; or a pre-trained weight determination network model can be used to determine. For example, the target detection data of the corresponding to-be-tracked targets can be input into the weight determination network model, and the weight result output by the model is taken as the matching weight between the corresponding to-be-tracked targets. The weight determination network model can be obtained by unsupervised training by a related technical person according to the target detection data of the to-be-tracked targets in the historical period.

[0042] It can be understood that there is at least one to-be-tracked target in each of the different to-be-tracked target groups, and therefore the matching weight of each to-be-tracked target in the different groups needs to be determined.

[0043] Exemplarily, if there are two groups of to-be-tracked targets, which are to-be-tracked target group A and to-be-tracked target group B. The to-be-tracked target group A includes to-be-tracked target A1 and to-be-tracked target A2, and the to-be-tracked target group B includes to-be-tracked target B1 and to-be-tracked target B2. Correspondingly, the matching weight M11 between the to-be-tracked target A1 and the to-be-tracked target B1 can be determined according to the target detection data of the to-be-tracked target A1 and the target detection data of the to-be-tracked target B1; the matching weight M12 between the to-be-tracked target A1 and the to-be-tracked target B2 can be determined according to the target detection data of the to-be-tracked target A1 and the target detection data of the to-be-tracked target B2; the matching weight M21 between the to-be-tracked target A2 and the to-be-tracked target B1 can be determined according to the target detection data of the to-be-tracked target A2 and the target detection data of the to-be-tracked target B1; and the matching weight M22 between the to-be-tracked target A2 and the to-be-tracked target B2 can be determined according to the target detection data of the to-be-tracked target A2 and the target detection data of the to-be-tracked target B2. Therefore, according to the target detection data of the to-be-tracked targets in the to-be-tracked target group A and the to-be-tracked target group B, the matching weight M11, the matching weight M12, the matching weight M21 and the matching weight M22 between the corresponding to-be-tracked targets can be determined.

[0044] In S130, target matching is performed on the to-be-tracked targets in different groups according to the matching weights.

[0045] The target matching can be used to represent whether two to-be-tracked targets can be matched, i.e., whether the two to-be-tracked targets are the same target in the same detection region.

[0046] Exemplarily, the to-be-tracked targets in different groups can be matched based on a preset target matching algorithm according to the matching weights. The target matching algorithm can be preset by a person skilled in the art. For example, the target matching algorithm can be the Hungarian Algorithm or the Kuhn-Munkres Algorithm.

[0047] It can be understood that there can be at least three target detection devices in the same detection region, i.e., corresponding to three groups of to-be-tracked targets. If there are multiple (more than three) groups of to-be-tracked targets, two groups of to-be-tracked targets can be matched first to obtain a virtual group of to-be-tracked targets after target matching; then the virtual group of to-be-tracked targets after target matching and any one of the other groups of to-be-tracked targets that have not been matched can be matched, and so on, so as to realize target matching on the to-be-tracked targets in multiple groups of to-be-tracked targets.

[0048] For example, if there are three groups of to-be-tracked targets, namely, a to-be-tracked target group A, a to-be-tracked target group B, and a to-be-tracked target group C, any two of the to-be-tracked target groups can be matched first, for example, the to-be-tracked target group A and the to-be-tracked target group B can be matched to obtain a matched virtual to-be-tracked target group D; and the virtual to-be-tracked target group D and the to-be-tracked target group C can be matched. The matching weights between the to-be-tracked targets in the virtual to-be-tracked target group D and the to-be-tracked target group C are determined, and the to-be-tracked targets are matched according to the matching weights.

[0049] It should be noted that the to-be-tracked target group can include matched to-be-tracked targets and unmatched to-be-tracked targets. The matched to-be-tracked targets are to-be-tracked targets that have been matched. The unmatched to-be-tracked targets can be the results detected by the target detection device corresponding to the to-be-tracked target group itself, and the unmatched to-be-tracked targets do not match any to-be-tracked target in other to-be-tracked target groups.

[0050] The technical scheme of the embodiment of the present disclosure determines the matching weights between the to-be-tracked targets in different to-be-tracked target groups by obtaining the target detection data of the to-be-tracked targets in different to-be-tracked target groups, and determines the matching weights between the to-be-tracked targets according to the target detection data of the to-be-tracked targets in different groups, thereby determining the probability that the to-be-tracked targets are the same target, increasing the matching success rate of the to-be-tracked targets, matching the to-be-tracked targets in different groups according to the matching weights, improving the accuracy of the target matching result, and providing data support for the fusion tracking of the target in the multi-target detection device.

[0051] On the basis of the above technical schemes, the present disclosure further provides an optional embodiment, in which the target detection data includes position detection data, and the operation of “determining the matching weights between the to-be-tracked targets according to the target detection data of the to-be-tracked targets in different groups” is further refined as “generating a candidate matching target pair including the to-be-tracked targets in different groups; determining a distance weight according to the position detection data of the to-be-tracked targets in the candidate matching target pair; and determining the matching weight of the candidate matching target pair according to the distance weight”, so as to improve the determination method of the matching weight. It should be noted that the parts not described in detail in the embodiment of the present disclosure can be referred to the related descriptions in other embodiments, which will not be described here.

[0052] Referring to Figure 2 The target matching method shown in FIG. 2 includes the following steps.

[0053] In S210, target detection data of to-be-tracked targets in different to-be-tracked target groups is obtained. Different to-be-tracked target groups correspond to different target detection devices, and the detection regions of different target detection devices are the same. The target detection data includes position detection data.

[0054] S220, generate a candidate matching target pair including different groups of to-be-tracked targets.

[0055] The candidate matching target pair can be two to-be-tracked targets to be matched, and the two to-be-tracked targets are from different groups respectively.

[0056] The candidate matching target pair can be generated manually by a skilled person, or can be randomly generated according to each to-be-tracked target in different groups.

[0057] For example, if there are to-be-tracked target group A and to-be-tracked target group B, and to-be-tracked target group A has to-be-tracked target A1 and to-be-tracked target A2, and to-be-tracked target group B has to-be-tracked target B1 and to-be-tracked target B2. Correspondingly, the candidate matching target pair including to-be-tracked targets of to-be-tracked target group A and to-be-tracked target group B can include: candidate matching target pair {A1, B1}, candidate matching target pair {A1, B2}, candidate matching target pair {A2, B1} and candidate matching target pair {A2, B2}.

[0058] S230, determine a distance weight according to position detection data of each to-be-tracked target in the candidate matching target pair.

[0059] The position detection data can include position coordinates of each to-be-tracked target in a world coordinate system.

[0060] For example, the position distance between each to-be-tracked target can be determined according to the corresponding position coordinates of each to-be-tracked target. The greater the position distance, the greater the distance weight; the smaller the position distance, the smaller the distance weight. The specific correspondence between the position distance and the distance weight can be pre-set by a skilled person.

[0061] For example, a network model for determining the distance weight can also be pre-trained. Specifically, the position detection data of each to-be-tracked target in the candidate matching target pair in the historical period can be used to train a pre-constructed network model to obtain a distance weight network model. When determining the distance weight, the position detection data of each to-be-tracked target in the candidate matching target pair can be input into the distance weight network model to obtain the corresponding distance weight. The pre-constructed network model can be obtained by combining at least one machine learning model in the prior art, and the specific network structure of the model is not limited in the present disclosure.

[0062] S240, determine a matching weight of the candidate matching target pair according to the distance weight.

[0063] Exemplarily, the distance weight can be taken as the matching weight of the candidate matching target pair, or the distance weight multiplied by a coefficient can be taken as the matching weight of the candidate matching target pair, where the coefficient can be preset by a person skilled in the art according to actual requirements, or the distance weight can be processed by using an existing data processing manner, and the processed weight result can be determined as the matching weight of the candidate matching target pair.

[0064] S250, target matching is performed on different groups of the to-be-tracked targets according to the matching weights.

[0065] The embodiment of the present disclosure generates the candidate matching target pair including different groups of the to-be-tracked targets, determines the distance weight according to the position detection data of each to-be-tracked target in the candidate matching target pair, and determines the matching weight of the candidate matching target pair according to the distance weight, thereby improving the determination accuracy of the matching weight of the candidate matching target pair, and improving the target matching accuracy of the to-be-tracked targets, which facilitates more accurate fusion tracking of the to-be-tracked targets.

[0066] On the basis of the above technical solutions, the present disclosure further provides an optional embodiment, in which the operation of “determining the distance weight according to the position detection data of each to-be-tracked target in the candidate matching target pair” is further refined as “determining the distance data of the candidate matching target pair according to the position detection data of each to-be-tracked target in the candidate matching target pair; determining the distance weight of the candidate matching target pair according to the distance data”, so as to improve the determination manner of the distance weight. It should be noted that the parts not described in detail in the embodiment of the present disclosure can refer to the related descriptions in other embodiments, which will not be described here.

[0067] Referring to Figure 3 A target matching method is shown in the figure, which comprises:

[0068] S310, target detection data of each to-be-tracked target in different groups of to-be-tracked targets is acquired; different groups of to-be-tracked targets correspond to different target detection devices, and the detection regions of different target detection devices are the same; the target detection data includes position detection data.

[0069] S320, a candidate matching target pair including different groups of to-be-tracked targets is generated.

[0070] S330, distance data of the candidate matching target pair is determined according to the position detection data of each to-be-tracked target in the candidate matching target pair.

[0071] The position detection data can include a detected position of the to-be-tracked target, and the detected position can be a three-dimensional coordinate of the to-be-tracked target in a world coordinate system, which can include coordinates corresponding to a length, a width and a height of the to-be-tracked target. The distance data can be Euclidean distance data or Mahalanobis distance data.

[0072] Exemplarily, the distance data of the candidate matching target pair can be determined based on a Euclidean distance determination algorithm according to the detected positions of the to-be-tracked targets in the candidate matching target pair.

[0073] It can be understood that, if the Euclidean distance is used as the distance data of the candidate matching target pair, it can be considered that the length, the width and the height of the to-be-tracked target are independent of each other and have no direct correlation. However, on an actual road, since the road has a direction and the road length and width have requirements, the target on the road is limited by the road length and width and has correlation with the length and width in the direction of the road extension. Therefore, in order to consider the influence of the correlation, the numerical difference and the size inequality between the detected position coordinates of the to-be-tracked target on the distance, the Mahalanobis distance can be used as the distance data of the candidate matching target pair.

[0074] In an optional embodiment, the position detection data includes a detected position and a position covariance; and the distance data of the candidate matching target pair is determined according to the position detection data of the to-be-tracked targets in the candidate matching target pair, including: fusing the position covariances of the to-be-tracked targets in the candidate matching target pair to obtain a fused position covariance; and determining the distance data of the candidate matching target pair according to the fused position covariance and the detected positions of the to-be-tracked targets in the candidate matching target pair.

[0075] The position covariance is used to represent the uncertainty of the length, the width and the height coordinates in the detected position of the to-be-tracked target, and the influence relationship of different position dimensions.

[0076] The detected position and the position covariance of the to-be-tracked target can be obtained according to the prediction and observation results of the Kalman filter output in the tracking process of the to-be-tracked target by the target detection device.

[0077] The fused position covariance can be obtained by adding the position covariances of the to-be-tracked targets after reversible transformation, and then performing reversible transformation on the added result again.

[0078] Exemplarily, if the to-be-tracked targets in the candidate matching target pair are to-be-tracked target A and to-be-tracked target B, the position covariance of the to-be-tracked target A can be represented as ∑ A , and the position covariance of the to-be-tracked target B can be represented as ∑ B . Correspondingly, the fused position covariance between the to-be-tracked target A and the to-be-tracked target B can be determined in the following manner:

[0079] ∑ C =(∑ A -1 +∑ B -1 ) -1 ;

[0080] Where, ∑ C This represents the fused position covariance between target A and target B to be tracked.

[0081] For example, the distance data of the candidate matching target pair can be determined based on the fused position covariance and the detection position of each target to be tracked in the candidate matching target pair, using the Mahalanobis distance determination method.

[0082] This optional embodiment achieves accurate determination of the distance data for candidate matching target pairs by fusing the position covariances of each target to be tracked in the candidate matching target pair. This is achieved by fusing the position covariances and then determining the distance data of the candidate matching target pair based on the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair. By considering the influence of correlation, numerical differences, and unevenness in the magnitude of the detection position coordinates of the targets to be tracked on the distance, the above scheme introduces a fused position covariance in the process of determining the distance data of the candidate matching target pair, thus achieving accurate determination of the distance data of the candidate matching target pair.

[0083] To further improve the efficiency and accuracy of determining the distance data of candidate matching target pairs, the following methods can be used to determine the distance data of candidate matching target pairs.

[0084] In an optional embodiment, determining the distance data of the candidate matching target pair based on the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair includes: determining the first Mahalanobis distance between each target to be tracked in the candidate matching target pair based on the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair; and using the first Mahalanobis distance as the distance data of the candidate matching target pair.

[0085] The detection position of the target to be tracked can be represented by P = (x, y, z), where P represents the detection position of the target in the world coordinate system. Here, x can be the length coordinate of the target, y can be the width coordinate, and z can be the height coordinate. The position covariance of the target can be represented by ∑. The position covariance ∑ characterizes the uncertainty of the target's x, y, and z position coordinates and the influence relationship between different dimensions, reflecting the correlation between x, y, and z.

[0086] For example, if the target to be tracked in the candidate matching target pair is target A and target B, the detection position of target A can be P.A =(x A y A , z A The location covariance can be ∑ A The detection location of the target B to be tracked can be P. B =(x B y B , z B The location covariance can be ∑ B Based on the location covariance ∑ A and location covariance ∑ B The fusion position covariance ∑ between target A and target B can be determined. C The first Mahalanobis distance between target A and target B can be determined as follows:

[0087]

[0088] Among them, D AB Let be the first Mahalanobis distance between target A and target B to be tracked.

[0089] For example, the determined first Mahalanobis distance can be used as the distance data for the candidate matching target pair; alternatively, the first Mahalanobis distance multiplied by a preset weighting coefficient can be used as the distance data for the candidate matching target pair. The preset weighting coefficient can be pre-set by relevant technical personnel according to actual needs, and this embodiment does not impose any restrictions on it.

[0090] This optional embodiment reduces the complexity of distance data determination and improves the efficiency of distance data determination by determining the distance data between each target to be tracked in the candidate matching target pair based on the fused position covariance and the detection position of each target to be tracked in the candidate matching target pair.

[0091] In another optional embodiment, determining the distance data of the candidate matching target pair based on the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair includes: determining the fused detection position based on the fused position covariance and the position detection data of each target to be tracked in the candidate matching target pair; determining the second Mahalanobis distance of the corresponding target to be tracked based on the fused detection position and the position detection data of the target to be tracked in the candidate matching target pair; and determining the distance data of the candidate matching target pair based on the second Mahalanobis distance of each target to be tracked in the candidate matching target pair.

[0092] The location detection data may include the detected location and location covariance of the target to be tracked. The fused detection location may be the merged detection location of each target to be tracked in the candidate matching target pair.

[0093] Exemplarily, if there are to-be-tracked target A and to-be-tracked target B in the candidate matching target pair, the following determination manner can be adopted to determine the fused detection position of each to-be-tracked target in the candidate matching target pair after fusion:

[0094]

[0095] wherein, P C is the fused detection position of to-be-tracked target A and to-be-tracked target B in the candidate matching target pair after fusion; ∑ C is the fusion position covariance between to-be-tracked target A and to-be-tracked target B; ∑ A is the position covariance of to-be-tracked target A, P A is the detection position of to-be-tracked target A; ∑ B is the position covariance of to-be-tracked target B, P B is the detection position of to-be-tracked target B.

[0096] Continuing the previous example, the determination manner of the second Mahalanobis distance of the corresponding to-be-tracked target can be as follows:

[0097]

[0098]

[0099] wherein, P C is the fused detection position of to-be-tracked target A and to-be-tracked target B after fusion. D AC may represent the second Mahalanobis distance between the detection position of to-be-tracked target A and the fused detection position; D BC may represent the second Mahalanobis distance between the detection position of to-be-tracked target B and the fused detection position.

[0100] Correspondingly, according to the second Mahalanobis distance corresponding to to-be-tracked target A and to-be-tracked target B in the candidate matching target pair respectively, the determination manner of the distance data of the candidate matching target pair can be as follows:

[0101]

[0102] wherein, D AB may be the second Mahalanobis distance between to-be-tracked target A and to-be-tracked target B.

[0103] Exemplarily, the determined second Mahalanobis distance can be taken as the distance data of the candidate matching target pair, or the second Mahalanobis distance multiplied by a weight coefficient can be taken as the distance data of the candidate matching target pair. The weight coefficient can be pre-set by a relevant technical person according to actual requirements, and the embodiment does not limit this.

[0104] The optional embodiment improves the determination accuracy of the distance data of the candidate matching target pair by determining the second Mahalanobis distance of the corresponding to-be-tracked target according to the position detection data of the to-be-tracked target in the candidate matching target pair, and determining the distance data of the candidate matching target pair according to the corresponding second Mahalanobis distance.

[0105] In S340, the distance weight of the candidate matching target pair is determined according to the distance data.

[0106] For example, the distance weight of the candidate matching target pair can be determined in the following manner:

[0107] W dis =max(0,D max_dist -D AB );

[0108] wherein D max_dist is a maximum matching distance parameter, which can be pre-set by a person skilled in the art according to actual experience value or test value; D AB is the distance data of the candidate matching target pair; and W dis is the distance weight of the candidate matching target pair.

[0109] Generally, D max_dist is not less than D AB . When D max_dist is less than D AB , it indicates that the distance data between the to-be-tracked targets in the candidate matching target pair is abnormal, i.e., the to-be-tracked targets cannot be matched. The smaller the value of D AB , the greater the distance weight of the candidate matching target pair, which indicates that the probability of matching the to-be-tracked targets is greater; and the greater the value of D AB , the smaller the distance weight of the candidate matching target pair, which indicates that the probability of matching the to-be-tracked targets is smaller.

[0110] In S350, the matching weight of the candidate matching target pair is determined according to the distance weight.

[0111] In S360, target matching is performed on different groups of to-be-tracked targets according to the matching weights.

[0112] The embodiment of the application realizes accurate determination of the distance weight of the candidate matching target pair by determining the distance data of the candidate matching target pair according to the position detection data of each to-be-tracked target in the candidate matching target pair, and determining the distance weight of the candidate matching target pair according to the distance data, thereby improving the determination accuracy of the matching weight of the candidate matching target pair, and further improving the matching accuracy of target matching performed on different groups of to-be-tracked targets.

[0113] It can be understood that in the determination of the matching weight of the candidate matching target pair, the matching weight can also be updated by using the category detection data in the target detection data, so that the determined matching weight is more accurate.

[0114] In an optional embodiment, the target detection data further comprises category detection data; a category weight of the candidate matching target pair is determined according to the category detection data of each to-be-tracked target in the candidate matching target pair; and the matching weight of the candidate matching target pair is updated according to the category weight.

[0115] The category detection data can comprise a detection category of the to-be-tracked target, for example, the detection category can comprise at least one of a car, a truck, a bus, a trolleybus, a motorcycle, a bicycle and a pedestrian.

[0116] For example, whether the to-be-tracked targets belong to the same detection category can be determined according to the detection categories corresponding to the to-be-tracked targets. If yes, the category weight can be set to 1; if no, the category weight can be set to 0.

[0117] For example, the category detection data can further comprise a category confidence, which is used to represent the accuracy of determining that the to-be-corrected target belongs to the corresponding detection category. Optionally, a network model for determining the category weight can be pre-trained according to the detection categories and the category confidences corresponding to the to-be-tracked targets in the candidate matching target pairs in the historical period, and the category weight of the candidate matching target pair can be determined according to the trained category weight determination network model. The network model can be obtained by combining at least one existing machine learning model, and the specific network structure of the network model is not limited in the present disclosure.

[0118] For example, the matching weight of the candidate matching target pair can be updated by using the weighted average of the category weight and the distance weight, or the product of the category weight and the distance weight. For example, the matching weight of the candidate matching target pair can be updated in the following manner:

[0119] W=W dis *W type ;

[0120] W=W dis *W type ;

[0121] The optional embodiment determines the category weight of the candidate matching target pair according to the category detection data of each to-be-tracked target in the candidate matching target pair, and updates the matching weight of the candidate matching target pair according to the category weight, so that the matching weight is updated from multiple angles such as distance factor and category factor, and the determination accuracy of the matching weight of the candidate matching target pair is improved.

[0122] To further improve the determination accuracy of the category weight, the category weight of the candidate matching target pair can be determined according to the detection category and the category confidence in the category detection data.

[0123] In an optional embodiment, the category detection data includes a detection category and a category confidence; and the determination of the category weight of the candidate matching target pair according to the category detection data of each to-be-tracked target in the candidate matching target pair includes: determining the category weight of the candidate matching target pair according to the detection category consistency between each to-be-tracked target in the candidate matching target pair and the category confidence of each to-be-tracked target in the candidate matching target pair.

[0124] The detection category and the category confidence in the category detection data can be obtained according to the Kalman filtering result of the target detection device in tracking the to-be-tracked target.

[0125] The detection category consistency can be used to represent whether the detection categories of each to-be-tracked target in the candidate matching target pair are consistent.

[0126] The determination of the category weight is different when the detection categories between each to-be-tracked target are consistent and when the detection categories between each to-be-tracked target are inconsistent. For example, if the to-be-tracked targets in the candidate matching target pair are to-be-tracked target A and to-be-tracked target B, the category weight of the candidate matching target pair can be determined as follows:

[0127]

[0128] Wherein, S A may be the category confidence of the to-be-tracked target A, S B may be the category confidence of the to-be-tracked target B; cls A may be the detection category of the to-be-tracked target A, cls B may be the detection category of the to-be-tracked target B. W type may be the category weight of the candidate matching target pair.

[0129] It should be noted that S A may represent the probability that the detection categories of the to-be-tracked target A and the to-be-tracked target B are both correct; (1-S B may represent the probability that the detection categories of the to-be-tracked target A and the to-be-tracked target B are not correct. A B ​) can represent the probability of the detection categories of the to-be-tracked target A and the to-be-tracked target B being both wrong; and the sum of the probability of the detection categories being both correct and the probability of the detection categories being both wrong is taken as the category weight when the detection categories of the to-be-tracked targets are consistent.

[0130] S A *(1-S B ) can represent the probability of the detection category of the to-be-tracked target A being correct and the detection category of the to-be-tracked target B being wrong; (1-S A )*S B can represent the probability of the detection category of the to-be-tracked target A being wrong and the detection category of the to-be-tracked target B being correct; and the sum of the probability of one detection category being wrong and one detection category being correct is taken as the category weight when the detection categories of the to-be-tracked targets are inconsistent.

[0131] The optional embodiment determines the category weight of the candidate matching target pair according to the consistency of the detection categories of the to-be-tracked targets in the candidate matching target pair and the category confidence of the to-be-tracked targets in the candidate matching target pair. By considering the influencing factor of the consistency of the detection categories of the to-be-tracked targets in the process of determining the category weight, the determination accuracy of the category weight of the candidate matching target pair is improved.

[0132] It can be understood that the size detection data in the target detection data can also be used to update the matching weight when the matching weight of the candidate matching target pair is determined, so that the determined matching weight is more accurate.

[0133] In an optional embodiment, the target detection data further includes size detection data. The size weight of the candidate matching target pair is determined according to the size detection data of the to-be-tracked targets in the candidate matching target pair; and the matching weight of the candidate matching target pair is updated according to the size weight.

[0134] The size detection data can include the length, width and height of the to-be-tracked target. The size detection data can reflect the difference in size between the two to-be-tracked targets.

[0135] For example, the size of the length, width and height of the to-be-tracked targets in the candidate matching target pair can be compared. If the size difference is greater than a preset difference threshold, the size weight can be set to 0; if the size difference is not greater than the preset difference threshold, the size weight can be set to 1. The difference threshold can be preset by a relevant technical person. For example, the size difference can be set to 0.2 meters.

[0136] According to the size weight, the manner of updating the matching weight of the candidate matching target pair can be to take a weighted average of the size weight and the distance weight as the matching weight of the candidate matching target pair. Alternatively, the product of the size weight and the distance weight can also be taken as the matching weight of the candidate matching target pair. For example, the matching weight of the candidate matching target pair can be updated in the following manner:

[0137] W = W dis * W size ;

[0138] wherein W is the matching weight of the candidate matching target pair; W dis is the distance weight; and W size is the size weight.

[0139] Alternatively, the product of the size weight, the category weight and the distance weight can also be taken as the matching weight of the candidate matching target pair. For example, the matching weight of the candidate matching target pair can be updated in the following manner:

[0140] W = W dis * W type * W size ;

[0141] wherein W is the matching weight of the candidate matching target pair; W dis is the distance weight; W type is the category weight; and W size is the size weight. Among them, the distance weight W dis is primary, and the category weight W type and the size weight W size are secondary.

[0142] The optional embodiment realizes the updating of the matching weight from multiple angles such as the size weight, the distance weight and the category weight, and improves the determination accuracy of the matching weight of the candidate matching target pair, by determining the size weight of the candidate matching target pair according to the size detection data of each to-be-tracked target in the candidate matching target pair, and updating the matching weight of the candidate matching target pair according to the size weight.

[0143] To further improve the determination accuracy of the size weight, the size weight of the candidate matching target pair can also be determined by determining size difference data.

[0144] In an optional embodiment, determining the size weight of the candidate matching target pair according to the size detection data of each to-be-tracked target in the candidate matching target pair comprises: determining size difference data of different size dimensions between the corresponding to-be-tracked targets according to the size detection data of each to-be-tracked target in the candidate matching target pair; and determining the size weight of the candidate matching target pair according to the size difference data of the different size dimensions.

[0145] The size difference data can be a size difference value between the to-be-tracked targets. The size detection data of the to-be-tracked targets can include different size dimension data, for example, the size dimensions can include length, width, and height. Correspondingly, the size difference data of the to-be-tracked targets can include length difference data, width difference data, and height difference data. For example, if the length of the to-be-tracked target A is l A , and the length of the to-be-tracked target B is l B , the length difference data of the to-be-tracked target A and the to-be-tracked target B can be |l A -l B |.

[0146] For example, if the to-be-tracked target A and the to-be-tracked target B are included in the candidate matching target pair; the size detection data corresponding to the to-be-tracked target A includes length l A , width w A , and height h A ; the size detection data corresponding to the to-be-tracked target B includes length l B , width w B , and height h B . The size difference data can include length difference data, width difference data, and height difference data. The length difference data between the to-be-tracked target A and the to-be-tracked target B can be |l A -l B |; the width difference data between the to-be-tracked target A and the to-be-tracked target B can be |w A -w B |; and the height difference data between the to-be-tracked target A and the to-be-tracked target B can be |h A -h B |. The size weight of the candidate matching target pair can be determined as follows:

[0147]

[0148]

[0149] The size weight of the candidate matching target pair can be determined as follows: The length difference data of the to-be-tracked target can be normalized. The width difference data of the to-be-tracked target can be normalized. The height difference data of the to-be-tracked target can be normalized. Wherein, D size can reflect the size difference between the to-be-tracked targets in different size dimensions.

[0150] Wherein, D max_size is a maximum size distance parameter, which can be pre-set by a person skilled in the art. If D size is greater than Dmax_size , it indicates that the size difference between each to-be-tracked target is too large, and the two targets do not need to be matched; if D size is not greater than D max_size , it indicates that the size difference between each to-be-tracked target is small, and the two targets can be matched.

[0151] Optionally, in the process of determining D size In the process of selecting difference data of different size dimensions or normalizing difference data, the max function can also be replaced by a min function or an avg function. It should be noted that, in order to maximize the difference between the targets detected by different target detection devices, in a preferred embodiment, the max function is used in the process of selecting difference data of different size dimensions and normalizing difference data.

[0152] The optional embodiment determines the size difference data of different size dimensions between the corresponding to-be-tracked targets according to the size detection data of each to-be-tracked target in the candidate matching target pair, and determines the size weight of the candidate matching target pair according to the size difference data of different size dimensions. The above scheme improves the determination accuracy of the size weight of the candidate matching target pair by considering the size difference data in the process of determining the size dimension.

[0153] On the basis of the above technical solutions, the present disclosure further provides an optional embodiment, in which, after the operation of "performing target matching on different groups of to-be-tracked targets according to the matching weights", the operation of "fusing target detection data of matched to-be-tracked targets and performing target tracking according to the fusion result" is added to improve the fusion tracking of the matched to-be-tracked targets. It should be noted that, the parts not described in detail in the embodiments of the present disclosure can be referred to the related descriptions in other embodiments, which will not be described here.

[0154] Referring to the target matching method shown in Figure 4A , the method comprises the following steps:

[0155] S410, obtaining target detection data of each to-be-tracked target in different groups of to-be-tracked targets; wherein, different groups of to-be-tracked targets correspond to different target detection devices, and the detection regions of different target detection devices are the same.

[0156] S420, determining matching weights between corresponding to-be-tracked targets according to the target detection data of different groups of to-be-tracked targets.

[0157] S430, performing target matching on different groups of to-be-tracked targets according to the matching weights.

[0158] S440. The target detection data that has been matched with the target to be tracked are fused, and the target is tracked based on the fusion result.

[0159] Among them, the matched target to be tracked can be the target to be tracked that has been matched successfully; the target to be tracked that has not been matched or has been matched but has not been successfully matched will not be fused.

[0160] The fusion of target detection data that have matched targets to be tracked can include fusing the detection categories of the matched targets to be tracked, as well as fusing the detection locations of the matched targets to be tracked.

[0161] For example, such as Figure 4B The diagram illustrates target fusion and target tracking for a target to be tracked. Wherein, Figure 4B The left image shows the position results of each target to be tracked within the same detection area detected by two target detection devices. The targets detected by the two devices are matched, and the successfully matched targets are then fused for tracking, resulting in the image shown. Figure 4B The right figure shows the result after fusion tracking.

[0162] For example, target tracking based on the fusion results can be performed by using a Kalman filter to predict and update the target detection data for the matched target to be tracked.

[0163] In one optional embodiment, target detection data that have matched targets to be tracked are fused, and target tracking is performed based on the fusion result, including: fusing the detection categories of the matched targets to be tracked to obtain a fused detection category; fusing the detection locations of the matched targets to be tracked to obtain a fused detection location; and performing target tracking based on the fused detection category and / or the fused detection location.

[0164] The method for determining the fusion detection category can be to obtain the detection category and category confidence of the matched target to be tracked; and to take the detection category corresponding to the target to be tracked with a high category confidence as the fusion detection category.

[0165] For example, suppose the targets to be tracked are target A and target B. Target A is detected as a car with a category confidence of 0.9; target B is detected as a bus with a category confidence of 0.1. Based on the category confidences of target A and target B, since the category confidence of target A is greater than that of target B, the detection category of target A is used as the fused detection category. Therefore, the fused detection category for the matched targets A and B is "car".

[0166] The determination manner of the fusion detection position can be: obtaining the fusion covariance corresponding to the matched to-be-tracked target and position detection data of the to-be-tracked target. The position detection data comprises the detection position and position covariance of the to-be-tracked target. According to the fusion covariance and the position detection data, the fusion detection position corresponding to the matched to-be-tracked target is determined.

[0167] For example, if the matched to-be-tracked target is to-be-tracked target A and to-be-tracked target B, the detection position of to-be-tracked target A is P A =(x A , y A , z A ), and the position covariance is∑ A . The detection position of to-be-tracked target B is P B =(x B , y B , z B ), and the position covariance is∑ B . The fusion covariance corresponding to to-be-tracked target A and to-be-tracked target B is∑ C . The determination manner of the fusion detection position corresponding to the matched to-be-tracked target A and to-be-tracked target B can be as follows:

[0168]

[0169] According to the fusion detection category and / or the fusion detection position, target tracking is performed based on a Kalman filter. For example, during target tracking, target tracking can be performed according to the fusion detection category, target tracking can be performed according to the fusion detection position, or target tracking can be performed according to both the fusion detection category and the fusion detection position. The present embodiment does not limit this.

[0170] The present optional embodiment fuses the detection category and the detection position of the matched to-be-tracked target respectively to obtain the fusion detection category and the fusion detection position, thereby achieving comprehensive fusion of the matched to-be-tracked target. According to the fusion detection category and / or the fusion detection position, target tracking is performed, thereby achieving accurate tracking of the fused target.

[0171] To further improve the determination accuracy of the fusion detection category, the fusion detection category can be determined by introducing a reference probability as follows.

[0172] In an optional embodiment, the matched detection categories of the to-be-tracked targets are fused to obtain a fused detection category, including: determining reference probabilities of each detection category of the matched to-be-tracked targets according to a historical passing proportion of the detection categories of the matched to-be-tracked targets in the road to which the detection region belongs and a category confidence of the matched to-be-tracked targets; and determining the fused detection category according to the reference probabilities of each detection category of the matched to-be-tracked targets.

[0173] The historical passing proportion can be pre-set by a related technical person according to an actual experience value or a test value. Alternatively, historical data of the road to which the detection region belongs can be obtained, and the historical passing proportion can be determined according to the number of the detection categories of the matched to-be-tracked targets on the road in the historical data.

[0174] The reference probability can be used to represent the possibility of the detection category being true. The greater the reference probability corresponding to the detection category, the greater the possibility that the fused detection category is the detection category; and the smaller the reference probability corresponding to the detection category, the smaller the possibility that the fused detection category is the detection category.

[0175] For example, if the matched to-be-tracked targets are a to-be-tracked target A and a to-be-tracked target B, the corresponding detection categories are a car and a bus respectively, and the historical data of the road to which the detection region belongs in a historical period is obtained, where the historical data includes the number of cars passing on the road and the number of buses passing on the road. The historical passing proportion of the to-be-tracked target A and the to-be-tracked target B is determined according to the number of cars passing and the number of buses passing. For example, if the number of cars passing is 30 and the number of buses passing is 5, the historical passing proportion of the to-be-tracked target A and the to-be-tracked target B can be determined as 6:1.

[0176] The reference probability of the corresponding detection category is determined according to the historical passing proportion of the matched to-be-tracked targets and the category confidence corresponding to the to-be-tracked target. For example, a neural network model pre-constructed can be trained according to the historical passing proportion of the detection categories of the matched to-be-tracked targets in the road to which the detection region belongs in a historical period and the category confidence of the matched to-be-tracked targets, to obtain a probability determination network model for determining the reference probability. In the process of using the network model, the historical passing proportion and the category confidence corresponding to the matched to-be-tracked target can be input into the probability determination network model, and the output result of the model can be used as the reference probability of each detection category of the matched to-be-tracked targets. The pre-constructed neural network model can be obtained by combining at least one machine learning model in the prior art, and the specific network structure of the model is not limited in the present disclosure.

[0177] According to the reference probabilities of the detection categories of the matched to-be-tracked targets, the fusion detection category is determined. For example, the detection category with a higher reference probability can be taken as the fusion detection category. For example, if the reference probability of the detection category of to-be-tracked target A is 0.8 and the reference probability of the detection category of to-be-tracked target B is 0.2, the reference probability of the detection category of to-be-tracked target A is greater than that of to-be-tracked target B, and thus the detection category of to-be-tracked target A is taken as the fusion detection category.

[0178] According to the reference probabilities of the detection categories of the matched to-be-tracked targets, the fusion detection category is determined. For example, the detection category with a higher reference probability can be taken as the fusion detection category. For example, if the reference probability of the detection category of to-be-tracked target A is 0.8 and the reference probability of the detection category of to-be-tracked target B is 0.2, the reference probability of the detection category of to-be-tracked target A is greater than that of to-be-tracked target B, and thus the detection category of to-be-tracked target A is taken as the fusion detection category.

[0179] To further improve the accuracy of the determination of the reference probability, the following determination method can also be used to determine the reference probability.

[0180] In an optional embodiment, the reference probabilities of the detection categories of the matched to-be-tracked targets are determined according to the historical passing proportion of the detection categories of the matched to-be-tracked targets in the road to which the detection region belongs and the category confidence of the matched to-be-tracked targets, and the method comprises the following steps: determining the category probability of each detection category of the matched to-be-tracked targets according to the historical passing proportion of the detection categories of the matched to-be-tracked targets in the road to which the detection region belongs; determining the conditional probability of each detection category of the matched to-be-tracked targets according to the category probability of each detection category of the matched to-be-tracked targets and the category confidence of the matched to-be-tracked targets; and determining the reference probabilities of the detection categories of the matched to-be-tracked targets according to the category probability and the conditional probability of each detection category of the matched to-be-tracked targets.

[0181] The category probability can be the probability that the detection category of the matched to-be-tracked target is correct. For example, if the matched to-be-tracked targets in the road to which the detection region belongs are to-be-tracked target A and to-be-tracked target B, and the historical passing proportion of the detection categories of to-be-tracked target A and to-be-tracked target B is 9:1, the category probability of the detection category of to-be-tracked target A is 0.9, and the category probability of the detection category of to-be-tracked target B is 0.1.

[0182] The conditional probability can be the probability of the detection category under the condition that the detection category meets a certain category confidence.

[0183] For example, if the matched to-be-tracked targets are to-be-tracked target A and to-be-tracked target B, and the detection category corresponding to to-be-tracked target A is car with a category confidence of m, and the detection category corresponding to to-be-tracked target B is bus with a category confidence of n. The conditional probability of the detection category of to-be-tracked target A can be represented as P(A: car, B: bus|car); and the conditional probability of the detection category of to-be-tracked target B can be represented as P(A: car, B: bus|bus). Wherein,

[0184] P(A: car, B: bus|car) = m * (1-n);

[0185] P(A: car, B: bus|bus) = n * (1-m);

[0186] According to the category probability and the conditional probability, the reference probability of each detection category of to-be-tracked target A and to-be-tracked target B in the matched to-be-tracked targets can be determined in the following manner:

[0187]

[0188]

[0189] Wherein, P(car|A: car, B: bus) can be the reference probability of the detection category of to-be-tracked target A; P(bus|A: car, B: bus) can be the reference probability of the detection category of to-be-tracked target B. P(A: car, B: bus|car) can be the conditional probability of to-be-tracked target A; P(A: car, B: bus|bus) can be the conditional probability of to-be-tracked target B. P(car) can be the category probability of to-be-tracked target A; and P(bus) can be the category probability of to-be-tracked target B.

[0190] If the category confidence of to-be-tracked category A is 0.7, and the category confidence of to-be-tracked target B is 0.8, the reference probability of to-be-tracked target A can be determined in the following manner:

[0191]

[0192] The reference probability of to-be-tracked target B can be determined in the following manner:

[0193]

[0194] If the category probability P(car) of to-be-tracked target A is 0.9, and the category probability P(bus) of to-be-tracked target B is 0.1, the reference probability P(car|A: car, B: bus) of to-be-tracked target A is 0.84, and the reference probability P(bus|A: car, B: bus) of to-be-tracked target B is 0.12.

[0195] This optional embodiment determines the category probability and the conditional probability of each detection category of the matched target to be tracked, and determines the reference probability of each detection category of the matched target to be tracked based on the category probability and the conditional probability. This achieves accurate determination of the reference probability of each detection category, thereby improving the accuracy of determining the fused detection category, and thus improving the accuracy of target tracking based on the fused detection category.

[0196] The present invention integrates target detection data of matched targets and performs target tracking based on the integration results, thereby achieving integrated tracking of multiple targets collected by a multi-target detection device and improving the accuracy of target tracking results.

[0197] In one specific embodiment, such as Figure 4C The diagram illustrates a multi-target detection device fusion tracking framework. Four target detection devices exist within the same detection area, acquiring images of the same region. The acquired images are labeled A, B, C, and D. It should be noted that the diagram uses four target detection devices as an example and should not be construed as a specific limitation on the number of detection devices. Each image includes the target to be tracked within the detection area. The tracking module continuously optimizes and updates the target detection data of the target to be tracked and outputs the optimized and updated target detection data. Target matching is performed on the target to be tracked corresponding to each target detection device. Specifically, the target detection data can be fused using a fusion tracking interface via asynchronous transmission, incorporating prior information (such as the aforementioned historical traffic ratio). Multi-dimensional information feature matching is performed on the target to be tracked corresponding to each target detection device. This multi-dimensional information can include location, category, and size dimensions. The target detection data of the successfully matched target to be tracked are fused, and target tracking is performed based on the fusion result.

[0198] As an implementation of the above-described target matching methods, this disclosure also provides an optional embodiment of an execution device for implementing the above-described target matching methods.

[0199] Figure 5 This is a schematic diagram of a target matching device according to an embodiment of the present disclosure. This embodiment is applicable to application scenarios where targets acquired by multiple sensors in the same location area are matched, fused, and tracked. The device can be configured in an electronic device and can implement the target matching method described in any embodiment of the present disclosure. (Reference) Figure 5 The target matching device 500 specifically includes the following:

[0200] The target detection data acquisition module 501 is configured to acquire target detection data of each target to be tracked in different groups of targets to be tracked.

[0201] The matching weight determination module 502 is configured to determine matching weights between corresponding targets to be tracked according to the target detection data of the different groups of targets to be tracked.

[0202] The target matching module 503 is configured to perform target matching on the different groups of targets to be tracked according to the matching weights.

[0203] The technical scheme of the embodiments of the present disclosure acquires target detection data of each target to be tracked in different groups of targets to be tracked, and determines matching weights between corresponding targets to be tracked according to the target detection data of the different groups of targets to be tracked, thereby realizing determination of the probability that the targets to be tracked are the same target, increasing the matching success rate of the targets to be tracked, performing target matching on the different groups of targets to be tracked according to the matching weights, improving the accuracy of the target matching result, and providing data support for target fusion tracking under multiple target detection devices.

[0204] In an optional implementation, the target detection data includes position detection data.

[0205] The matching weight determination module 502 includes:

[0206] The matching target pair generation unit is configured to generate a candidate matching target pair including different groups of targets to be tracked.

[0207] The distance weight determination unit is configured to determine a distance weight according to the position detection data of each target to be tracked in the candidate matching target pair.

[0208] The matching weight determination unit is configured to determine a matching weight of the candidate matching target pair according to the distance weight.

[0209] In an optional implementation, the distance weight determination unit includes:

[0210] The distance data determination subunit is configured to determine distance data of the candidate matching target pair according to the position detection data of each target to be tracked in the candidate matching target pair.

[0211] The distance weight determination subunit is configured to determine a distance weight of the candidate matching target pair according to the distance data.

[0212] In an optional implementation, the position detection data includes a detection position and a position covariance.

[0213] The distance data determining subunit comprises:

[0214] A position covariance determining subunit is configured to fuse position covariances of each to-be-tracked target in the candidate matching target pair to obtain a fused position covariance.

[0215] A distance data determining subunit is configured to determine distance data of the candidate matching target pair according to the fused position covariance and the detected positions of each to-be-tracked target in the candidate matching target pair.

[0216] In an optional implementation, the distance data determining subunit comprises:

[0217] A first Mahalanobis distance determining subunit is configured to determine a first Mahalanobis distance between each to-be-tracked target in the candidate matching target pair according to the fused position covariance and the detected positions of each to-be-tracked target in the candidate matching target pair.

[0218] A first distance data determining subunit is configured to take the first Mahalanobis distance as the distance data of the candidate matching target pair.

[0219] In an optional implementation, the distance data determining subunit comprises:

[0220] A detected position determining subunit is configured to determine a fused detected position according to the fused position covariance and the position detection data of each to-be-tracked target in the candidate matching target pair.

[0221] A second Mahalanobis distance determining subunit is configured to determine a second Mahalanobis distance of a corresponding to-be-tracked target according to the fused detected position and the position detection data of the to-be-tracked target in the candidate matching target pair.

[0222] A second distance data determining subunit is configured to determine distance data of the candidate matching target pair according to the second Mahalanobis distance of each to-be-tracked target in the candidate matching target pair.

[0223] In an optional implementation, the target detection data further comprises category detection data, and the apparatus further comprises:

[0224] A category weight determining module is configured to determine a category weight of the candidate matching target pair according to the category detection data of each to-be-tracked target in the candidate matching target pair.

[0225] A first matching weight updating module is configured to update the matching weight of the candidate matching target pair according to the category weight.

[0226] In an optional implementation, the category detection data comprises a detected category and a category confidence.

[0227] The category weight determination module comprises:

[0228] A category weight determination unit is configured to determine a category weight of the candidate matching target pair according to detection category consistency between each to-be-tracked target in the candidate matching target pair and category confidence of each to-be-tracked target in the candidate matching target pair.

[0229] In an optional implementation, the target detection data further comprises size detection data, and the apparatus further comprises:

[0230] A size weight determination module is configured to determine a size weight of the candidate matching target pair according to size detection data of each to-be-tracked target in the candidate matching target pair.

[0231] A second matching weight updating module is configured to update a matching weight of the candidate matching target pair according to the size weight.

[0232] In an optional implementation, the size weight determination module comprises:

[0233] A size difference data determination unit is configured to determine size difference data between different size dimensions of each to-be-tracked target in the candidate matching target pair according to size detection data of the to-be-tracked target.

[0234] A size weight determination unit is configured to determine a size weight of the candidate matching target pair according to the size difference data of different size dimensions.

[0235] In an optional implementation, the apparatus 500 further comprises:

[0236] A target tracking module is configured to fuse target detection data of matched to-be-tracked targets and perform target tracking according to a fusion result.

[0237] In an optional implementation, the target tracking module comprises:

[0238] A fused detection category determination unit is configured to fuse detection categories of the matched to-be-tracked targets to obtain a fused detection category.

[0239] A fused detection position determination unit is configured to fuse detection positions of the matched to-be-tracked targets to obtain a fused detection position.

[0240] A target tracking unit is configured to perform target tracking according to the fused detection category and / or the fused detection position.

[0241] In an optional implementation, the fused detection category determination unit comprises:

[0242] The reference probability determination subunit is configured to determine reference probabilities of each detection category in the matched-to-be-tracked target according to a historical passing proportion of the detection category of the matched-to-be-tracked target in a road to which the detection region belongs and a category confidence of the matched-to-be-tracked target.

[0243] The fusion detection category determination subunit is configured to determine the fusion detection category according to the reference probabilities of each detection category in the matched-to-be-tracked target.

[0244] In an optional implementation,

[0245] The reference probability determination subunit includes:

[0246] The category probability determination subunit is configured to determine category probabilities of each detection category of the matched-to-be-tracked target according to a historical passing proportion of the detection category of the matched-to-be-tracked target in a road to which the detection region belongs.

[0247] The conditional probability determination subunit is configured to determine conditional probabilities of each detection category of the matched-to-be-tracked target according to the category probabilities of each detection category of the matched-to-be-tracked target and a category confidence of the matched-to-be-tracked target.

[0248] The reference probability determination subunit is configured to determine reference probabilities of each detection category in the matched-to-be-tracked target according to the category probabilities of each detection category of the matched-to-be-tracked target and the conditional probabilities.

[0249] The target matching device described above can perform the target matching method provided by any embodiment of the present disclosure, and has corresponding function modules and beneficial effects of performing the target matching method.

[0250] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of the target detection data involved all comply with relevant laws and regulations and do not violate public order and good customs.

[0251] According to the embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0252] Figure 6A schematic block diagram of an example electronic device Y00 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0253] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0254] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0255] The computing unit 601 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs various methods and processes described above, such as the target matching method. For example, in some embodiments, the target matching method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded onto the RAM 603 and executed by the computing unit 601, one or more steps of the target matching method described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the target matching method by any other suitable means, such as by means of firmware.

[0256] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0257] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0258] In the context of this disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0259] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0260] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0261] The computer system can include clients and servers. This relationship can be. The servers are generally remote from the users and can be accessed via the Internet using a communication network. The relationship can be a client-server relationship over a communications network, and as such both the client and the server are typically computers, or other client and server computers. In a client-server relationship, the server is often providing functionality and data to the client. For example, the server can provide data, or functionality, to the client using any one of a number of protocols that are well known to those of ordinary skill in the art.

[0262] Artificial intelligence is a discipline that studies enabling computers to simulate some human thinking processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), both hardware and software technologies. Artificial intelligence hardware technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing, etc.; artificial intelligence software technology mainly includes computer vision technology, speech recognition technology, natural language processing technology, machine learning / deep learning technology, big data processing technology, knowledge graph technology, etc.

[0263] Cloud computing refers to accessing elastic and scalable shared physical or virtual resource pools through a network, which can include servers, operating systems, networks, software, applications, and storage devices, and can deploy and manage resources in a self-service manner as needed. Through cloud computing technology, powerful data processing capabilities can be provided for artificial intelligence, blockchain, and other technology applications and model training.

[0264] It should be understood that various forms of the flow shown above can be used to reorder, add or delete steps. For example, each step described in the present disclosure can be executed in parallel, sequentially or in a different order, as long as the desired results of the technical solutions provided by the present disclosure can be achieved, which is not limited herein.

[0265] The above detailed description does not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A target matching method, comprising: obtaining target detection data of each target to be tracked in different groups of targets to be tracked; wherein different groups of targets to be tracked correspond to different target detection devices, and detection regions of different target detection devices are the same; determining matching weights between corresponding targets to be tracked according to the target detection data of different groups of targets to be tracked, wherein the matching weights between corresponding targets to be tracked in different groups are used to represent probabilities that two targets to be tracked for matching are the same target; performing target matching on different groups of targets to be tracked according to the matching weights; fusing target detection data of matched targets to be tracked, and performing target tracking according to a fusion result.

2. The method of claim 1, wherein, The target detection data comprises position detection data. The method of determining matching weights between corresponding targets to be tracked according to the target detection data of different groups of targets to be tracked comprises: generating candidate matching target pairs comprising different groups of targets to be tracked; determining distance weights according to position detection data of each target to be tracked in the candidate matching target pairs; determining matching weights of the candidate matching target pairs according to the distance weights.

3. The method of claim 2, wherein, The method of determining distance weights according to position detection data of each target to be tracked in the candidate matching target pairs comprises: determining distance data of the candidate matching target pairs according to position detection data of each target to be tracked in the candidate matching target pairs; determining distance weights of the candidate matching target pairs according to the distance data.

4. The method of claim 3, wherein, The position detection data comprises detection positions and position covariances. The method of determining distance data of the candidate matching target pairs according to position detection data of each target to be tracked in the candidate matching target pairs comprises: fusing position covariances of each target to be tracked in the candidate matching target pairs to obtain fused position covariances; determining distance data of the candidate matching target pairs according to the fused position covariances and detection positions of each target to be tracked in the candidate matching target pairs.

5. The method of claim 4, wherein, The method of determining distance data of the candidate matching target pairs according to the fused position covariances and detection positions of each target to be tracked in the candidate matching target pairs comprises: determining first Mahalanobis distances between each target to be tracked in the candidate matching target pairs according to the fused position covariances and detection positions of each target to be tracked in the candidate matching target pairs; taking the first Mahalanobis distances as the distance data of the candidate matching target pairs.

6. The method of claim 4, wherein, The method of determining distance data of the candidate matching target pairs according to the fused position covariances and detection positions of each target to be tracked in the candidate matching target pairs comprises: determining fused detection positions according to the fused position covariances and position detection data of each target to be tracked in the candidate matching target pairs; determining second Mahalanobis distances of corresponding targets to be tracked according to the fused detection positions and position detection data of targets to be tracked in the candidate matching target pairs; determining distance data of the candidate matching target pairs according to the second Mahalanobis distances of each target to be tracked in the candidate matching target pairs.

7. The method according to any one of claims 2-6, wherein, The target detection data further comprises category detection data, and the method further comprises: determine a category weight of the candidate matched target pair according to category detection data of each to-be-tracked target in the candidate matched target pair; update the matching weight of the candidate matched target pair according to the category weight.

8. The method of claim 7, wherein, the category detection data comprises a detection category and a category confidence; the determining the category weight of the candidate matched target pair according to the category detection data of each to-be-tracked target in the candidate matched target pair comprises: determine the category weight of the candidate matched target pair according to consistency of detection categories between each to-be-tracked target in the candidate matched target pair and the category confidence of each to-be-tracked target in the candidate matched target pair.

9. The method according to any one of claims 2-6, wherein, the target detection data further comprises size detection data, and the method further comprises: determine a size weight of the candidate matched target pair according to size detection data of each to-be-tracked target in the candidate matched target pair; update the matching weight of the candidate matched target pair according to the size weight.

10. The method of claim 9, wherein, the determining the size weight of the candidate matched target pair according to the size detection data of each to-be-tracked target in the candidate matched target pair comprises: determine size difference data of different size dimensions between corresponding to-be-tracked targets according to the size detection data of each to-be-tracked target in the candidate matched target pair; determine the size weight of the candidate matched target pair according to the size difference data of different size dimensions.

11. The method of claim 1, wherein, the fusing the target detection data of the matched to-be-tracked target and performing target tracking according to a fusion result comprises: fuse detection categories of the matched to-be-tracked target to obtain a fusion detection category; fuse detection positions of the matched to-be-tracked target to obtain a fusion detection position; perform target tracking according to the fusion detection category and / or the fusion detection position.

12. The method of claim 11, wherein, the fusing the detection categories of the matched to-be-tracked target to obtain the fusion detection category comprises: determine a reference probability of each detection category of the matched to-be-tracked target according to a historical passing proportion of the detection categories of the matched to-be-tracked target in a road to which the detection region belongs and a category confidence of the matched to-be-tracked target; determine the fusion detection category according to the reference probability of each detection category of the matched to-be-tracked target.

13. The method of claim 12, wherein, the determining the reference probability of each detection category of the matched to-be-tracked target according to the historical passing proportion of the detection categories of the matched to-be-tracked target in the road to which the detection region belongs and the category confidence of the matched to-be-tracked target comprises: determine a category probability of each detection category of the matched to-be-tracked target according to the historical passing proportion of the detection categories of the matched to-be-tracked target in the road to which the detection region belongs; determine a conditional probability of each detection category of the matched to-be-tracked target according to the category probability of each detection category of the matched to-be-tracked target and the category confidence of the matched to-be-tracked target; determine the reference probability of each detection category of the matched to-be-tracked target according to the category probability of each detection category of the matched to-be-tracked target and the conditional probability.

14. An apparatus for target matching, comprising: The target detection data acquisition module is configured to acquire target detection data of each target to be tracked in different groups of targets to be tracked, wherein the different groups of targets to be tracked correspond to different target detection devices, and the detection regions of the different target detection devices are the same; The matching weight determination module is configured to determine matching weights between corresponding targets to be tracked according to the target detection data of the targets to be tracked in different groups, wherein the matching weights between the targets to be tracked in different groups correspond to probabilities that two targets to be tracked for matching are the same target; The target matching module is configured to perform target matching on the targets to be tracked in different groups according to the matching weights. The target tracking module is configured to fuse the target detection data of the matched targets to be tracked, and perform target tracking according to a fusion result.

15. The apparatus of claim 14, wherein, The target detection data includes position detection data. The matching weight determination module includes: The matching target pair generation unit is configured to generate a candidate matching target pair including the targets to be tracked in different groups; The distance weight determination unit is configured to determine a distance weight according to the position detection data of each target to be tracked in the candidate matching target pair; The matching weight determination unit is configured to determine a matching weight of the candidate matching target pair according to the distance weight.

16. The apparatus of claim 15, wherein, The distance weight determination unit includes: The distance data determination subunit is configured to determine distance data of the candidate matching target pair according to the position detection data of each target to be tracked in the candidate matching target pair; The distance weight determination subunit is configured to determine a distance weight of the candidate matching target pair according to the distance data.

17. The apparatus of claim 16, wherein, The position detection data includes a detection position and a position covariance; The distance data determination subunit includes: The position covariance determination subunit is configured to fuse the position covariances of each target to be tracked in the candidate matching target pair to obtain a fused position covariance; The distance data determination subunit is configured to determine distance data of the candidate matching target pair according to the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair.

18. The apparatus of claim 17, wherein, The distance data determination subunit includes: The first Mahalanobis distance determination subunit is configured to determine a first Mahalanobis distance between each target to be tracked in the candidate matching target pair according to the fused position covariance and the detection positions of each target to be tracked in the candidate matching target pair; The first distance data determination subunit is configured to take the first Mahalanobis distance as the distance data of the candidate matching target pair.

19. The apparatus of claim 17, wherein, The distance data determination subunit includes: The detection position determination subunit is configured to determine a fused detection position according to the fused position covariance and the position detection data of each target to be tracked in the candidate matching target pair; The second Mahalanobis distance determination subunit is configured to determine a second Mahalanobis distance of a corresponding target to be tracked according to the fused detection position and the position detection data of the target to be tracked in the candidate matching target pair; The second distance data determination subunit is configured to determine distance data of the candidate matching target pair according to the second Mahalanobis distances of each target to be tracked in the candidate matching target pair.

20. The apparatus of any of claims 15-19, wherein, The target detection data further comprises category detection data, and the device further comprises: a category weight determination module, configured to determine a category weight of the candidate matching target pair according to the category detection data of each to-be-tracked target in the candidate matching target pair; a first matching weight updating module, configured to update the matching weight of the candidate matching target pair according to the category weight.

21. The apparatus of claim 20, wherein, The category detection data comprises a detection category and a category confidence; The category weight determination module comprises: a category weight determination unit, configured to determine the category weight of the candidate matching target pair according to the detection category consistency between each to-be-tracked target in the candidate matching target pair and the category confidence of each to-be-tracked target in the candidate matching target pair.

22. The apparatus of any one of claims 15-19, wherein, The target detection data further comprises size detection data, and the device further comprises: a size weight determination module, configured to determine a size weight of the candidate matching target pair according to the size detection data of each to-be-tracked target in the candidate matching target pair; a second matching weight updating module, configured to update the matching weight of the candidate matching target pair according to the size weight.

23. The apparatus of claim 22, wherein, The size weight determination module comprises: a size difference data determination unit, configured to determine size difference data of different size dimensions between corresponding to-be-tracked targets according to the size detection data of each to-be-tracked target in the candidate matching target pair; a size weight determination unit, configured to determine the size weight of the candidate matching target pair according to the size difference data of different size dimensions.

24. The apparatus of claim 14, wherein, The target tracking module comprises: a fused detection category determination unit, configured to fuse the detection categories of the matched to-be-tracked targets to obtain a fused detection category; a fused detection position determination unit, configured to fuse the detection positions of the matched to-be-tracked targets to obtain a fused detection position; a target tracking unit, configured to perform target tracking according to the fused detection category and / or the fused detection position.

25. The apparatus of claim 24, wherein, The fused detection category determination unit comprises: a reference probability determination subunit, configured to determine a reference probability of each detection category of the matched to-be-tracked targets according to a historical passing proportion of the detection categories of the matched to-be-tracked targets in a road to which the detection region belongs and a category confidence of the matched to-be-tracked targets; a fused detection category determination subunit, configured to determine the fused detection category according to the reference probability of each detection category of the matched to-be-tracked targets.

26. The apparatus of claim 25, wherein, The reference probability determination subunit comprises: a category probability determination subunit, configured to determine a category probability of each detection category of the matched to-be-tracked targets according to the historical passing proportion of the detection categories of the matched to-be-tracked targets in the road to which the detection region belongs; a conditional probability determination subunit, configured to determine a conditional probability of each detection category of the matched to-be-tracked targets according to the category probability of each detection category of the matched to-be-tracked targets and the category confidence of the matched to-be-tracked targets; a reference probability determination subunit, configured to determine the reference probability of each detection category of the matched to-be-tracked targets according to the category probability of each detection category of the matched to-be-tracked targets and the conditional probability.

27. An electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the target matching method of any one of claims 1-13.

28. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, the computer instructions are for causing a computer to perform the target matching method of any one of claims 1-13.

29. A computer program product comprising computer programs / instructions that, when executed by a processor, implement the steps of the target matching method of any one of claims 1-13.

Images