Target tracking method, electronic device and computer readable storage medium

Abstract

The application discloses a target tracking method, an electronic device and a computer readable storage medium. The method comprises: acquiring a first to-be-matched set containing a plurality of to-be-matched targets, the plurality of to-be-matched targets being respectively acquired by a plurality of cameras; dividing the first to-be-matched set into a plurality of to-be-matched subsets based on appearance feature similarities between the plurality of to-be-matched targets, the appearance feature similarities of to-be-matched targets contained in different to-be-matched subsets being less than the appearance feature similarities of to-be-matched targets contained in the same to-be-matched subset; performing deduplication on to-be-matched targets in a to-be-deduplicated to-be-matched subset, the to-be-deduplicated to-be-matched subset including at least one to-be-matched subset; and tracking the to-be-matched targets in the deduplicated to-be-matched subsets. In the foregoing manner, the accuracy of the target tracking result can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a target tracking method, an electronic device and a computer readable storage medium. BACKGROUND

[0002] Target tracking technology is widely used in the field of intelligent security. Target tracking is divided into single-camera target tracking and cross-camera target tracking, and the cross-camera target tracking is also referred to as multi-camera multi-target tracking. Single-camera target tracking refers to target tracking based on video data collected by a single camera, and multi-camera target tracking refers to target tracking based on video data collected by multiple cameras.

[0003] Single-camera target tracking and cross-camera target tracking can be used in different monitoring scenes. However, compared with single-camera target tracking, the existing cross-camera target tracking method has low accuracy of target tracking result. SUMMARY

[0004] The present application provides a target tracking method, an electronic device and a computer readable storage medium, which can solve the problem of low accuracy of target tracking result obtained by the existing cross-camera target tracking method.

[0005] To solve the above technical problems, one technical solution adopted by the present application is to provide a target tracking method. The method comprises: acquiring a first set of to-be-matched targets containing a plurality of to-be-matched targets, the plurality of to-be-matched targets being respectively collected by a plurality of cameras; dividing the first set of to-be-matched targets into a plurality of to-be-matched subsets based on the appearance feature similarity between the plurality of to-be-matched targets, the appearance feature similarity of to-be-matched targets contained in different to-be-matched subsets being less than the appearance feature similarity of to-be-matched targets contained in the same to-be-matched subset; de-duplicating to-be-matched targets in a to-be-de-duplicated to-be-matched subset, the to-be-de-duplicated to-be-matched subset including at least one to-be-matched subset; and tracking to-be-matched targets in the de-duplicated to-be-matched subset.

[0006] To solve the above technical problems, another technical solution adopted by the present application is to provide a target tracking device, which comprises a processor and a memory connected to the processor, wherein the memory stores program instructions; the processor is configured to execute the program instructions stored in the memory to implement the above method.

[0007] To solve the above technical problems, still another technical solution adopted by the present application is to provide a computer readable storage medium, which stores program instructions, and the program instructions can implement the above method when executed.

[0008] By the above manner, the first to-be-matched set is divided into a plurality of to-be-matched subsets based on the appearance feature similarity, the to-be-matched targets in the to-be-matched subset to be deduplicated are deduplicated, and the to-be-matched targets in the deduplicated to-be-matched subset are tracked. On the one hand, the present application tracks after deduplication, so as to avoid the situation of missing matching or false matching, and improve the accuracy of the target tracking result. On the other hand, since the appearance feature similarity of the to-be-matched targets contained in different to-be-matched subsets is less than the appearance feature similarity of the to-be-matched targets contained in the same to-be-matched subset, each to-be-matched target in the to-be-matched subset can be regarded as belonging to the same target, so that the present application divides the to-be-matched targets belonging to the same target in the to-be-matched subset into the same to-be-matched subset, and deduplicates in units of to-be-matched subset, so that the deduplication result obtained is more accurate, and the tracking based on the deduplication result can further improve the accuracy of the target tracking result. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a flowchart of an embodiment of the target tracking method of the present application;

[0010] Figure 2 is a schematic diagram of the fields of view of two cameras B1 and B2;

[0011] Figure 3 is a schematic diagram of a bipartite graph;

[0012] Figure 4 is a flowchart of another embodiment of the target tracking method of the present application;

[0013] Figure 5 is a flowchart of S21 in Figure 4

[0014] Figure 6 is a flowchart of a specific example of the target tracking method of the present application;

[0015] Figure 7 is a structural schematic diagram of an embodiment of the target tracking device of the present application;

[0016] Figure 8 is a structural schematic diagram of an embodiment of the computer readable storage medium of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0018] ​The terms "first", "second", "third", etc., in the present application are only for descriptive purpose, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the technical features indicated. Therefore, the features defined with "first", "second", "third" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise explicitly specified.

[0019] Reference to "embodiments" herein means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily independent or alternative embodiments to each other. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments without conflict.

[0020] Before introducing the target tracking method provided by the present application, the technical principles related to cross-camera target tracking are described:

[0021] The process of cross-camera target tracking includes: obtaining a plurality of targets to be matched acquired by a plurality of cameras; matching the plurality of targets to be matched with a plurality of matched targets to obtain a matching result, thereby realizing target tracking. In the process of matching the plurality of targets to be matched with the plurality of matched targets, at most one matched target is matched with one target to be matched. For the matched target to be matched, the target to be matched and the matched target matched therewith are regarded as the same target, and the same identity (a number, a string of characters, etc.) as the matched target matched therewith is assigned to the target to be matched. For the unmatched target to be matched, it is regarded as a new matched target, and a new identity is assigned to the target to be matched for subsequent target tracking.

[0022] The plurality of targets to be matched are acquired by a plurality of cameras at the same or similar time (allowing a certain tolerance). However, at the same or similar time, different cameras can acquire the same target, so that there can be different targets to be matched belonging to the same target in the plurality of targets to be matched. At most one matched target is matched with one target to be matched, that is, in the target tracking algorithm, different targets to be matched are considered to belong to different targets to be matched with the plurality of matched targets, which can cause missed matching or incorrect matching. For example, actually, the target to be matched A1 and the target to be matched A2 both belong to target 1, and the matched target M belongs to target 1, the target to be matched A1 and A2 should both be matched with M, but in the target tracking algorithm, the target to be matched A1 and A2 are considered to belong to different targets, if A1 is matched with M, then A2 will fail to match or match incorrectly.

[0023] To reduce or even avoid the occurrence of the situation of missing matching or false matching, and improve the accuracy of the target matching result, the target tracking method provided in the present application is as follows:

[0024] Figure 1 is a flowchart of an embodiment of the target tracking method of the present application. It should be noted that the embodiment is not limited to the flow order shown in Figure 1 . As shown in Figure 1 , the embodiment can include:

[0025] S11: Obtain a first set of to-be-matched targets containing a plurality of to-be-matched targets.

[0026] The plurality of to-be-matched targets are respectively acquired by a plurality of cameras.

[0027] The execution subject of the method embodiment of the present application is a target tracking device, which can be a camera, an electronic device (such as a mobile phone or a computer) in communication connection with the camera, or a server, etc. The target (to-be-matched target, matched target) mentioned in the present application can be any object with tracking requirements, such as a person, an animal (dog, cat), a vehicle, etc.

[0028] The plurality of cameras can be arranged at different positions in the same monitoring range, and the fields of view of different cameras can partially overlap or can not overlap, but at least the fields of view of two cameras partially overlap. In the case that the fields of view of different cameras partially overlap, different cameras can acquire the same target at the same or similar time. For example, the monitoring area is a supermarket, and two cameras with non-overlapping fields of view are arranged at the entrance area and the exit area of the supermarket, respectively. For another example, the monitoring area is a road, and two cameras B1 and B2 with partially overlapping fields of view are arranged at the first road segment and the second road segment of the road, respectively, Figure 2 is a schematic diagram of the fields of view of the two cameras B1 and B2, as shown in Figure 2 , the field of view of the camera B1 arranged at the first road segment is b1, the field of view of the camera B2 arranged at the second road segment is b2, b1 includes sub-areas b11 and b12, b2 includes sub-areas b12 and b21, b12 is the overlapping area of the fields of view of B1 and B2, and in the process of the target moving from b1 to b2, the target first appears in b11 and is acquired by B1, then appears in b12 and is acquired by B1 and B2, and then appears in b21 and is acquired by B2.

[0029] The plurality of to-be-matched targets are extracted from to-be-matched video frames acquired by the plurality of cameras at the same or similar time (with a certain tolerance). Specifically, a plurality of to-be-matched video frames acquired by the plurality of cameras at the same or similar time can be obtained, target detection is performed on the plurality of to-be-matched video frames respectively, and a plurality of to-be-matched targets are obtained.

[0030] The to-be-matched target has identity, corresponding camera identity, position, quality score, appearance feature, etc. The camera identity corresponding to the to-be-matched target refers to the identity of the camera that collects the to-be-matched target. The position of the to-be-matched target refers to the position of the to-be-matched target in the field of view of the camera corresponding to the to-be-matched target. The quality score can reflect the quality of the to-be-matched target, and the appearance feature can include texture, color, contour, etc. The appearance feature can be extracted by a feature extraction network, for example, the feature extraction network is a Reid neural network, and the appearance feature consists of 1024 numbers. The quality score can be obtained by using a quality evaluation algorithm, for example, using a NIQE algorithm, and the smaller the quality score, the higher the image quality.

[0031] S12: Based on the appearance feature similarity between the plurality of to-be-matched targets, the first to-be-matched set is divided into a plurality of to-be-matched subsets.

[0032] The appearance feature similarity of the to-be-matched targets included in different to-be-matched subsets is less than the appearance feature similarity of the to-be-matched targets included in the same to-be-matched subset.

[0033] The appearance feature similarity is the similarity between the appearance features. The similarity can be Euclidean distance, cosine distance, etc. For example, the appearance feature similarity of two to-be-matched targets is calculated according to the following formula:

[0034]

[0035] Wherein, F1 and F2 represent the appearance features of two to-be-matched targets respectively. F1·F2 represents the dot product of the appearance features of the two to-be-matched targets. ||F|| represents the two-norm of the appearance feature of the to-be-matched target. The value range of the appearance feature similarity S 12 is 0-1, and the greater the value, the more similar the appearance features of the two to-be-matched targets.

[0036] The appearance feature similarity between the to-be-matched targets in the same to-be-matched subset is greater than a similarity threshold (for example, 0.6), and the appearance feature similarity between the to-be-matched targets in different to-be-matched subsets is not greater than the similarity threshold.

[0037] It can be understood that the plurality of to-be-matched targets in the first to-be-matched set belong to a plurality of targets. Ideally, based on the appearance feature similarity, the to-be-matched subset division can be regarded as dividing the first to-be-matched set belonging to a plurality of targets into a plurality of first to-be-matched subsets each belonging to one target, and different first to-be-matched subsets belong to different targets. For example, the first to-be-matched set belongs to targets 1-3, and through division, a first to-be-matched subset belonging to target 1, a first to-be-matched subset belonging to target 2, and a first to-be-matched subset belonging to target 3 can be obtained.

[0038] In some embodiments, the appearance feature similarity of different to-be-matched targets can also be adaptively adjusted according to the quality scores of the different to-be-matched targets. For example, the appearance feature similarity between to-be-matched targets A1 and A2 is S12, and at least one of to-be-matched targets A1 and A2 is poor in quality (for example, greater than a quality score threshold), then S12 is adjusted to be larger by a preset ratio.

[0039] S13: De-duplication is performed on the to-be-matched targets in the to-be-de-duplicated to-be-matched subsets.

[0040] The to-be-de-duplicated to-be-matched subsets include at least one to-be-matched subset.

[0041] De-duplication refers to removing only the to-be-matched targets of the same target in the to-be-de-duplicated to-be-matched subsets. That is, one to-be-matched target of the same target in the to-be-de-duplicated to-be-matched subsets is retained, and the other to-be-matched targets of the same target are filtered out. For example, the to-be-matched subset includes to-be-matched targets A1-A4 of target 1, one is selected from to-be-matched targets A1-A4, and the other two are filtered out from the to-be-matched subset. Thus, in subsequent tracking, the to-be-matched targets retained in each to-be-de-duplicated to-be-matched subset are used for tracking.

[0042] The manner of de-duplication in S13 can include at least the following two manners:

[0043] Manner one, for each to-be-de-duplicated to-be-matched subset, one to-be-matched target can be selected from the to-be-de-duplicated to-be-matched subset and retained in the to-be-de-duplicated to-be-matched subset, and the other to-be-matched targets in the to-be-de-duplicated to-be-matched subset are filtered out from the to-be-de-duplicated to-be-matched subset. The to-be-matched target selected from the to-be-de-duplicated to-be-matched subset can be any to-be-matched target in the to-be-de-duplicated to-be-matched subset, or can be the to-be-matched target with the best quality in the to-be-de-duplicated to-be-matched subset. The quality of the to-be-matched target depends on the degree of occlusion, the intensity of light, and the like. In this manner, the to-be-de-duplicated to-be-matched subset after filtering and retaining is the to-be-de-duplicated to-be-matched subset mentioned later.

[0044] Manner two, one to-be-matched target can be selected from each to-be-de-duplicated to-be-matched subset and entered into the second to-be-matched set. The to-be-matched target selected from the to-be-de-duplicated to-be-matched subset can be any to-be-matched target in the to-be-de-duplicated to-be-matched subset, or can be the to-be-matched target with the best quality in the to-be-de-duplicated to-be-matched subset. In this manner, each to-be-matched target in the second to-be-matched set is used for subsequent tracking.

[0045] S14: Tracking is performed on the to-be-matched targets in the to-be-de-duplicated to-be-matched subsets.

[0046] In the case of the first mode of deduplication in S13, in S14, the deduplicated to-be-matched subset is each to-be-deduplicated to-be-matched subset after filtering and reservation. Each to-be-matched target in each deduplicated to-be-matched subset can be matched with a plurality of matched targets to obtain a matching result of each to-be-matched target in each deduplicated to-be-matched subset. Furthermore, for each to-be-deduplicated to-be-matched subset, the matching result of the to-be-matched target belonging to the to-be-deduplicated to-be-matched subset and belonging to the deduplicated to-be-matched subset is determined as the matching result of all to-be-matched targets in the to-be-deduplicated to-be-matched subset.

[0047] For example, the to-be-deduplicated to-be-matched subset includes to-be-matched targets A1-A4, to-be-matched targets A2-A4 are filtered, and to-be-matched target A1 is reserved to form a deduplicated to-be-matched subset. To-be-matched target A1 is matched with a plurality of matched targets M to obtain a matching result of to-be-matched target A1. The matching result of to-be-matched target A1 is determined as the matching result of to-be-matched targets A2-A4.

[0048] In the case of the second mode of deduplication in S13, in S14, the deduplicated to-be-matched subset is the second to-be-matched set. Each to-be-matched target in the second to-be-matched set can be matched with a plurality of matched targets to obtain a matching result of each to-be-matched target in the second to-be-matched set. Furthermore, for each to-be-deduplicated to-be-matched subset, the matching result of the to-be-matched target belonging to the to-be-deduplicated to-be-matched subset and being classified into the second to-be-matched set is determined as the matching result of all to-be-matched targets in the to-be-deduplicated to-be-matched subset.

[0049] For example, the to-be-deduplicated to-be-matched subset includes to-be-matched targets A1-A4, to-be-matched target A1 is selected to be classified into the second to-be-matched set, to-be-matched target A1 is matched with a plurality of matched targets M to obtain a matching result of to-be-matched target A1, and the matching result of to-be-matched target A1 is determined as the matching result of to-be-matched targets A2-A4.

[0050] The algorithm for matching the to-be-matched target with the matched target includes but is not limited to a bipartite graph matching algorithm, which can include KM, a contrast nearest neighbor algorithm, etc. In the process of matching using the bipartite graph matching algorithm, each to-be-matched target and each matched target form a bipartite graph, wherein each to-be-matched target is a first part node of the bipartite graph, each matched target is a second part node of the bipartite graph, a connection between a to-be-matched target and a matched target forms an edge of the bipartite graph, and the weight of the edge is the matching score between the to-be-matched target and the matched target forming the edge. The matching result can be obtained by solving the bipartite graph through a bipartite graph solving algorithm. Figure 3 In the bipartite graph, A1-A4 represent to-be-matched targets, and M1-M4 represent matched targets.

[0051] Through implementation of the embodiment, the application divides the first to-be-matched set into a plurality of to-be-matched subsets based on the appearance feature similarity, de-duplicates the to-be-matched targets in the to-be-de-duplicated to-be-matched subset, and tracks the to-be-matched targets in the de-duplicated to-be-matched subset. On the one hand, the application tracks after de-duplication, so as to avoid the situation of missed matching or false matching, and improve the accuracy of the target tracking result. On the other hand, since the appearance feature similarity of the to-be-matched targets contained in different to-be-matched subsets is less than the appearance feature similarity of the to-be-matched targets contained in the same to-be-matched subset, each to-be-matched target in the to-be-matched subset can be regarded as belonging to the same target, so that the application divides the to-be-matched targets belonging to the same target in the to-be-matched subset into the same to-be-matched subset, and de-duplicates in units of to-be-matched subset, so that the accuracy of the de-duplication result is higher, and the tracking based on the de-duplication result can further improve the accuracy of the target tracking result.

[0052] Further, in some embodiments, before S13, it can also be judged for each to-be-de-duplicated to-be-matched subset whether the number of to-be-matched targets in the to-be-de-duplicated to-be-matched subset is greater than a second number; if greater than the second number, S13-14 is executed. The second number can be the minimum value of the number of to-be-matched targets in the to-be-de-duplicated to-be-matched subset, for example, the second number is 1, 2 (allowing a certain fault tolerance), and the like.

[0053] It can be further understood that in actual situations, there can be noise in the to-be-matched subset obtained by dividing based on the appearance similarity in S12, that is, the to-be-matched targets in the to-be-matched subset can not only belong to the same target, but also belong to multiple targets. For example, in an ideal case, there is a to-be-matched target belonging to target 4 in the to-be-matched subset belonging to target 1. The existence of this noise can be due to the fact that the appearance feature has a poor expression ability for the to-be-matched target, or due to the fact that the algorithm for dividing has insufficient accuracy, and the like. The appearance feature can have a poor expression ability for the to-be-matched target due to external environment (for example, being blocked, light change, camera perspective difference), or due to the fact that the target detection algorithm has insufficient accuracy, and the like.

[0054] Therefore, in some embodiments, noise filtering can be performed on the to-be-de-duplicated to-be-matched subset before S13, so that the to-be-matched targets in the to-be-de-duplicated to-be-matched subset belong to the same target, and then the to-be-matched targets in the to-be-de-duplicated to-be-matched subset are de-duplicated. The noise filtering introduced below is realized under the premise of mode two.

[0055] Figure 4 is a flowchart of another embodiment of the target tracking method of the application. It should be noted that the embodiment is not limited to the flow order shown in Figure 4 the figure. For example, the steps in the figure can be performed in a different order. Figure 4As shown, the present embodiment can include:

[0056] S21: For each to-be-de-duplicated to-be-matched subset, based on the relationship between different to-be-matched targets in the to-be-de-duplicated to-be-matched subset, finding out a to-be-filtered to-be-matched target from the to-be-de-duplicated to-be-matched subset.

[0057] The relationship between different to-be-matched targets can include at least one of a relationship between appearance similarities of different to-be-matched targets, a connectivity relationship between cameras corresponding to different to-be-matched targets, and a positional relationship between different to-be-matched targets.

[0058] The connectivity relationship between different cameras can include that the fields of view of different cameras have a connectivity relationship or do not have a connectivity relationship. For example, the fields of view of camera 1 and camera 2 have a connectivity relationship, so within a preset time range after a target disappears from the field of view of camera 1, the target appears in the field of view of camera 2, or in the field of view of camera 3, or returns to the field of view of camera 1. The fields of view of camera 1 and camera 2 do not have a connectivity relationship, so within a preset time range after a target disappears from the field of view of camera 1, the target does not appear in the field of view of camera 2.

[0059] The connectivity relationship between different cameras can be manually input or obtained based on target tracking results of a plurality of matched targets. Specifically, the camera sequence corresponding to a plurality of matched targets can be determined based on the target tracking results of the plurality of matched targets, and the connectivity relationship between different cameras can be obtained by clustering the camera sequence corresponding to the plurality of matched targets. For example, the camera sequence of most of the matched targets includes the camera sub-sequence {camera 1, camera 2}, so by clustering, it can be determined with high probability that camera 1 and camera 2 have a connectivity relationship.

[0060] The positional relationship between different to-be-matched targets can include that the positions of different to-be-matched targets are both in a camera field of view overlap region, or are not both in the camera field of view overlap region, where the camera field of view overlap region refers to an overlap region of the fields of view of the cameras corresponding to different to-be-matched targets. Continue to combine the above-mentioned Figure 3 It is explained that the camera corresponding to to-be-matched target A1 is B1, and the camera corresponding to to-be-matched target A2 is B2. If to-be-matched targets A1 and A2 are both in the field of view overlap region B12 of B1 and B2, then to-be-matched targets A1 and A2 are both in the camera field of view overlap region; otherwise, they are not both in the camera field of view overlap region.

[0061] In combination with reference to Figure 5 S21 can include the following sub-steps:

[0062] S211: Determine whether each to-be-matched target in the to-be-de-duplicated to-be-matched subset meets a filtering condition.

[0063] The filtering condition can include at least one of the following:

[0064] The first filtering condition is that there is no communication relationship between the camera corresponding to the to-be-matched target and the camera corresponding to each reference target.

[0065] The second filtering condition is that there is a communication relationship between the camera corresponding to the to-be-matched target and the camera corresponding to at least one reference target, and the to-be-matched target and the at least one reference target are both in a camera field-of-view overlap region. The reference target is another to-be-matched target in the same to-be-de-duplicated to-be-matched subset to which the to-be-matched target belongs, and the camera field-of-view overlap region is a field-of-view overlap region of the cameras corresponding to the to-be-matched target and the at least one reference target.

[0066] The third filtering condition is that the appearance feature similarity between the to-be-matched target and each reference target is greater than a first similarity threshold and less than a second similarity threshold. The second similarity threshold is greater than the first similarity threshold, and the first similarity threshold is a similarity threshold when the first to-be-matched set is divided into to-be-de-duplicated to-be-matched subsets based on the appearance feature similarity, that is, a similarity threshold for being divided into the same to-be-de-duplicated to-be-matched subset.

[0067] S212: Determine the to-be-matched target that satisfies the filtering condition as a to-be-filtered to-be-matched target.

[0068] S22: Filter the to-be-filtered to-be-matched target from the to-be-de-duplicated to-be-matched subset and enter the second to-be-matched set.

[0069] It can be understood that the to-be-filtered to-be-matched target is noise relative to the to-be-de-duplicated to-be-matched subset, that is, the to-be-filtered to-be-matched target belongs to a different target relative to other to-be-matched targets in the to-be-de-duplicated to-be-matched subset, and thus it is entered into the second to-be-matched set for tracking after filtering.

[0070] Further, in the case where the filtering condition in S211 includes multiple (for example, including the first filtering condition and the second filtering condition), the first filtering condition can be first determined whether each to-be-matched target in the to-be-de-duplicated to-be-matched subset satisfies the first filtering condition, and the to-be-matched target that satisfies the first filtering condition is filtered; and then it is determined whether each to-be-matched target remaining in the to-be-de-duplicated to-be-matched subset after filtering satisfies the second filtering condition, and the to-be-matched target that satisfies the second filtering condition is filtered, and the to-be-matched target that does not satisfy the second filtering condition is retained. Alternatively, each to-be-matched target in the to-be-de-duplicated to-be-matched subset can be sequentially taken as a current to-be-matched target, it is determined whether the current to-be-matched target satisfies the first filtering condition, if yes, the current to-be-matched target is filtered, if not, it is determined whether the current to-be-matched target satisfies the second filtering condition, if yes, the current to-be-matched target is filtered, if not, the current to-be-matched target is retained. The latter case is described in detail as follows:

[0071] 1) sequentially taking each target in the to-be-matched subset to be de-duplicated as the current to-be-matched target, determining the camera corresponding to the current to-be-matched target and each reference target. Wherein, the corresponding camera can be determined based on the corresponding camera identifier.

[0072] 2) determining whether there is a connection relationship between the cameras corresponding to the current to-be-matched target and each reference target; if there is a connection relationship between the cameras corresponding to the current to-be-matched target and at least one reference target, then go to 3); otherwise, go to 5).

[0073] 3) determining whether the current to-be-matched target and at least one reference target are both in the camera overlapping region; if both are in the camera overlapping region, then go to 4); otherwise, go to 5).

[0074] 4) determining that the current to-be-matched target meets the filtering condition.

[0075] 5) determining that the current to-be-matched target does not meet the filtering condition.

[0076] In some embodiments, after 2) and before 3), it can also be determined whether the number of to-be-matched targets remaining in the to-be-matched subset to be de-duplicated is greater than a first number; if it is greater than the first number, then go to 3) to perform 3) and the steps thereafter; otherwise, end the filtering. The first number is the minimum value of the number of to-be-matched targets remaining in the to-be-matched subset to be de-duplicated, such as 1, 2 (allowing a certain tolerance) and the like.

[0077] The target tracking method provided by the present application is described in detail in the form of an example as follows: Figure 6

[0078] 1) obtain a first to-be-matched set X, including n to-be-matched targets (X1~Xn).

[0079] 2) calculate the appearance feature similarity S between X1~Xn feat , based on dividing X into m to-be-matched subsets X filter1 , the number of to-be-matched targets in X filter1 is N filter1 .

[0080] 3) de-duplicate the to-be-matched targets in X filter1 , to obtain X filter3 .

[0081] ① determine whether N filter1 =1 is met; if it is met, then go to ②; otherwise, do not de-duplicate based on the first filtering condition.

[0082] ② filter the to-be-matched targets in N filter1 that meet the first filtering condition and put them into a second to-be-matched set, X​filter1 Update X filter2 , X filter2 The number of targets to be matched in X filter2 .

[0083] 3) Determine whether N filter2 = 1 is satisfied; if satisfied, go to 4; otherwise, do not remove duplicates based on the second filtering condition.

[0084] 4) Filter the targets to be matched in X filter2 that satisfy the second filtering condition and put them into a second set of targets to be matched, and update X filter2 to X filter3 .

[0085] 4) Select a target to be matched with the highest quality score from each X filter3 and put it into the second set of targets to be matched.

[0086] 5) Match each target to be matched in the second set of targets to be matched with a number of matched targets to obtain a matching result.

[0087] 6) Determine the matching result of the target to be matched in X filter3 that belongs to the second set of targets to be matched as the matching result of all targets to be matched in X filter3 .

[0088] Figure 7 is a structural schematic diagram of an embodiment of the target tracking device. As shown in Figure 7 , the target tracking device includes a processor 21 and a memory 22 coupled to the processor 21.

[0089] The memory 22 stores program instructions for implementing the method of any of the above embodiments; the processor 21 is configured to execute the program instructions stored in the memory 22 to implement the steps of the above method embodiments. The processor 21 can also be referred to as a CPU (Central Processing Unit). The processor 21 can be an integrated circuit chip with signal processing capability. The processor 21 can also be a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0090] The target tracking device can be a camera, an electronic device (such as a mobile phone or a computer) in communication with the camera, or a server, etc.

[0091] Figure 8is a structural schematic diagram of an embodiment of the computer readable storage medium of the present application. As shown in Figure 8 The computer readable storage medium 30 of the embodiment of the present application stores program instructions 31 which, when executed, implement the method provided by the above-mentioned embodiments of the present application. The program instructions 31 can form a program file and be stored in the above-mentioned computer readable storage medium 30 in the form of a software product, so that a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor executes all or part of the steps of the method of each embodiment of the present application. The aforementioned computer readable storage medium 30 includes various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0092] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are merely schematic, for example, the division of units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0093] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit. The above is only an embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings of the present application, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A target tracking method characterized by, The method comprises: obtaining a first to-be-matched set containing a plurality of to-be-matched targets, the plurality of to-be-matched targets being respectively captured by a plurality of cameras; dividing the first to-be-matched set into a plurality of to-be-matched subsets based on appearance feature similarities between the plurality of to-be-matched targets, the appearance feature similarities between different to-be-matched targets in each to-be-matched subset being smaller than the appearance feature similarities between to-be-matched targets in the same to-be-matched subset; de-duplicating the to-be-matched targets in a to-be-de-duplicated to-be-matched subset, the to-be-de-duplicated to-be-matched subset including at least one to-be-matched subset; tracking the to-be-matched targets in the de-duplicated to-be-matched subsets; the de-duplicating the to-be-matched targets in at least one to-be-matched subset comprises: selecting one to-be-matched target from each to-be-de-duplicated to-be-matched subset to form a second to-be-matched set; the tracking the to-be-matched targets in the de-duplicated to-be-matched subsets comprises: matching each to-be-matched target in the second to-be-matched set with a plurality of matched targets to obtain a matching result of each to-be-matched target in the second to-be-matched set; for each to-be-de-duplicated to-be-matched subset, determining the matching result of the to-be-matched target belonging to the to-be-de-duplicated to-be-matched subset and included in the second to-be-matched set as the matching result of all the to-be-matched targets in the to-be-de-duplicated to-be-matched subset.

2. The method of claim 1, wherein, Before the de-duplicating the to-be-matched targets in the to-be-de-duplicated to-be-matched subset, the method further comprises: for each to-be-de-duplicated to-be-matched subset, finding out to-be-filtered to-be-matched targets from the to-be-de-duplicated to-be-matched subset based on relationships between different to-be-matched targets in the to-be-de-duplicated to-be-matched subset, the relationships between different to-be-matched targets including at least one of a connection relationship between cameras corresponding to different to-be-matched targets and a positional relationship between different to-be-matched targets; filtering the to-be-filtered to-be-matched targets from the to-be-matched subset and including the to-be-filtered to-be-matched targets in a second to-be-matched set.

3. The method of claim 2, wherein, The finding out the to-be-filtered to-be-matched targets from the to-be-de-duplicated to-be-matched subset based on the relationships between different to-be-matched targets in the to-be-de-duplicated to-be-matched subset comprises: determining whether each to-be-matched target in the to-be-de-duplicated to-be-matched subset meets a filtering condition; determining the to-be-matched target meeting the filtering condition as the to-be-filtered to-be-matched target; wherein the filtering condition includes at least one of the following: the camera corresponding to the to-be-matched target has no connection relationship with the cameras corresponding to each reference target, the camera corresponding to the to-be-matched target has a connection relationship with the cameras corresponding to at least one reference target, and the to-be-matched target and the at least one reference target are both in a camera field-of-view overlap region, the reference target being another to-be-matched target of the same to-be-de-duplicated to-be-matched subset to which the to-be-matched target belongs, and the camera field-of-view overlap region being a field-of-view overlap region of the camera corresponding to the to-be-matched target and the camera corresponding to the at least one reference target.

4. The method of claim 3, wherein, The determining whether each of the to-be-matched targets in the to-be-matched subset to be deduplicated satisfies a filtering condition comprises: sequentially taking each of the to-be-matched targets in the to-be-matched subset to be deduplicated as a current to-be-matched target, determining cameras corresponding to the current to-be-matched target and each of the reference targets; judging whether there is a connection relationship between the cameras corresponding to the current to-be-matched target and each of the reference targets; if there is a connection relationship between the cameras corresponding to the current to-be-matched target and at least one of the reference targets, judging whether the current to-be-matched target and the at least one of the reference targets are both in the camera field-of-view overlapping region; if both are in the camera field-of-view overlapping region, determining that the current to-be-matched target satisfies the filtering condition.

5. The method of claim 4, wherein, Before the judging whether the current to-be-matched target and the at least one of the reference targets are both in the camera overlapping region, the method further comprises: if there is a connection relationship between the cameras corresponding to the current to-be-matched target and at least one of the reference targets, judging whether the number of the to-be-matched targets remaining in the to-be-matched subset is greater than a first number; if greater than the first number, executing the judging whether the current to-be-matched target and the at least one of the reference targets are both in the camera overlapping region and the steps after the judging.

6. The method of claim 1, wherein, The respectively selecting one of the to-be-matched targets from each of the deduplicated to-be-matched subsets into a second to-be-matched set comprises: respectively selecting one of the to-be-matched targets with the best quality from each of the deduplicated to-be-matched subsets into the second to-be-matched set.

7. The method of claim 1, wherein, Before the deduplicating the to-be-matched targets in the to-be-matched subset to be deduplicated, the method further comprises: for each to-be-matched subset to be deduplicated, judging whether the number of the to-be-matched targets in the to-be-matched subset to be deduplicated is greater than a second number; if greater than the second number, executing the deduplicating the to-be-matched targets in the to-be-matched subset to be deduplicated.

8. A target tracking device, characterized by, The device comprises a processor and a memory connected to the processor, wherein the memory stores program instructions; the processor is configured to execute the program instructions stored in the memory to implement the method in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores program instructions, which can be executed by a processor, and when executed, implement the method in any one of claims 1-7. The computer readable storage medium stores program instructions, which can be executed by a processor, and when executed, implement the method in any one of claims 1-7.

Images