Method and device for generating track fragment and storage medium
By generating virtual track segments, using the area segmentation map and the optimal path table to connect discontinuous track segments, the problem of discontinuous tracking in multi-target tracking across cameras is solved, and continuous target tracks in time and space are achieved, reducing monitoring and the difficulty of the analyst.
Patent Information
- Application Number
- CN202310682973.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-09
- Publication Date
- 2024-12-10
AI Technical Summary
Cross-camera multi-objective tracking technology is difficult to form a complete and continuous motion trajectory of the target, causing surveillance and analysts to face discontinuous trajectories to increase work difficulty.
By generating virtual track segments, connecting discontinuous track segments, using the area segmentation map and the optimal path table, continuous tracks in time and space are determined, and virtual track segments are generated to form a complete target track.
A continuous target trajectory in time and space is achieved, reducing the difficulty of monitoring and analysts' work, and the predicted virtual track segment is closer to the real trajectory of the target.
Smart Images
Figure CN119107340A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure generally relates to target tracking, and more particularly, to a method and apparatus for generating trajectory segments of a target, and a storage medium storing a corresponding program. Background Art
[0002] Multi-Target Multi-Camera Tracking technology is currently a very important research topic in the field of surveillance video. It refers to finding the temporal and spatial positions of multiple tracking targets in videos shot by multiple cameras, and associating the temporal and spatial information of the same tracking target with the tracking target to obtain the trajectory of the tracking target.
[0003] The obtained trajectory of the tracked target is usually composed of multiple trajectory segments. Each trajectory segment is likely to be captured by different cameras and is likely to be discontinuous in time and space, so it is often difficult to form a complete and continuous motion trajectory of the tracked target. For monitoring personnel or analysts, such discontinuous (fragmented) trajectories increase the difficulty of work. Summary of the invention
[0004] In response to the above technical problems, the present disclosure proposes a technology for generating virtual trajectory segments, which are used to connect discontinuous trajectory segments to form a trajectory that is continuous in time and space, thereby facilitating monitoring personnel and analysts to intuitively identify the action trajectory of the tracked target.
[0005] According to one aspect of the present disclosure, a method for generating a trajectory segment of a target for target tracking is provided, comprising: sorting a plurality of trajectory segments of the target captured by a plurality of cameras in order from early to late based on the start time of the trajectory segments; determining whether to generate a virtual trajectory segment based on the time distance and the space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space; in the case where it is determined that the virtual trajectory segment is to be generated, generating the virtual trajectory segment based on a pre-generated region segmentation map and an optimal path table, wherein the region segmentation map indicates a segmentation of a plane layout map of a scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices, wherein the optimal path table indicates an optimal path between any two vertices in the region segmentation map.
[0006] According to another aspect of the present disclosure, a device for generating a trajectory segment of a target for target tracking is provided, comprising: a memory storing a computer program; and a processor, the processor being configured to perform the following operations by executing the computer program: sorting a plurality of trajectory segments of the target captured by a plurality of cameras in an order from early to late based on the start time of the trajectory segments; determining whether to generate a virtual trajectory segment based on the time distance and the space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space; if it is determined that the virtual trajectory segment is to be generated, generating the virtual trajectory segment based on a pre-generated region segmentation map and an optimal path table, wherein the region segmentation map indicates a segmentation of a plane layout map of a scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices, wherein the optimal path table indicates an optimal path between any two vertices in the region segmentation map.
[0007] According to another aspect of the present disclosure, a non-transitory computer-readable medium storing a program is provided. When the program is executed by a computer, the computer executes the above-mentioned method for generating trajectory segments of a target. BRIEF DESCRIPTION OF THE DRAWINGS
[0008] Figure 1 The plan layout diagram corresponding to the scene in the video is schematically shown.
[0009] Figure 2 A Voronoi diagram is schematically shown.
[0010] Figure 3 The vertices remaining after processing the Voronoi graph are schematically shown.
[0011] Figure 4 The final region segmentation map is schematically shown.
[0012] Figure 5 A flow chart of a method for generating trajectory segments of a target according to the present disclosure is shown.
[0013] Figure 6 Shows the implementation Figure 5 The method function module.
[0014] Figure 7 The merging of trajectory segments is schematically shown.
[0015] Figure 8 The generation of a virtual trajectory segment is schematically illustrated.
[0016] Fig. 9 A virtual trajectory segment is schematically shown.
[0017] Fig.10 An exemplary configuration block diagram of computer hardware for implementing the present disclosure is shown. DETAILED DESCRIPTION
[0018] Figure 1 The plan layout diagram corresponding to a scene is schematically shown, in which a plurality of cameras are arranged, and target tracking is performed using videos captured by these cameras. The plan layout diagram may be prepared in advance. For example, Figure 1 A floor plan corresponding to a supermarket is shown, wherein black parts represent obstacles such as shelves, pillars, and walls, and white parts represent areas where people can pass.
[0019] The following will be combined Figures 1 to 4 The method of generating a region segmentation map for the scene is described in detail.
[0020] First, obstacles in the plan layout are identified, and multiple points are extracted on the contour of each obstacle, such as Figure 1 As shown. The extracted points include the vertices of the contour and the points on the edges of the contour. In addition, the number of points extracted on each edge of the contour may depend on the length of the edge. As an example, the number of points M and the length of the edge L may satisfy the following relationship:
[0021] Then, a Voronoi diagram may be generated based on the extracted points. Since the generation of the Voronoi diagram is a known technique, a detailed description thereof will be omitted in the present disclosure. Figure 2 The generated Voronoi diagram is schematically shown.
[0022] Then, the Voronoi diagram is processed to remove the vertices located inside the obstacles. Figure 3 The remaining vertices are shown schematically. Figure 4 The region segmentation graph finally obtained after processing is schematically shown. The region segmentation graph indicates the segmentation of the plane layout graph of the scene, and includes a plurality of vertices and connection lines between connected vertices.
[0023] Next, we will describe a method for constructing an adjacency list based on a region segmentation graph.
[0024] Since the region partition graph includes vertices and connecting lines between connected vertices, the following adjacency list can be constructed based on the region partition graph: each element in the adjacency list is stored in the format of [N1, N2, w], where N1 and N2 represent two vertices connected to each other, and w represents the weight applied to the connecting line between vertices N1 and N2.
[0025] As an example, the weight w may depend on the distance d between two locations in the floor plan graph corresponding to the vertices N1 and N2.
[0026] In addition, the weight w may further depend on the crowd density corresponding to the connecting line between vertices N1 and N2.
[0027] As a method for calculating the density of pedestrian flow, we can map each point on the trajectory to a regional segmentation map of the supermarket (e.g., Figure 4 Specifically, the distance between the trajectory point and the connecting line can be calculated according to the method of calculating the distance from a point to a straight line. In this way, for each connecting line, it is determined how many trajectory points are mapped to the connecting line, so as to calculate the thermal value for the connecting line. The calculated thermal value is divided by the length of the connecting line, and the result obtained can be used as the crowd density corresponding to the connecting line.
[0028] Next, the crowd density of all connecting lines in the region segmentation map can be mapped to the interval [m, n], where the maximum crowd density is mapped to the lower limit m of the interval, and the minimum crowd density is mapped to the upper limit n. As an example, linear mapping can be used, and m = 0.6, n = 1. It should be noted that the present disclosure is not limited to this, and other mapping methods can also be used, as long as the conditions that a smaller crowd density is mapped to a larger value and a larger crowd density is mapped to a smaller value can be met.
[0029] Considering both the distance d between the actual physical locations and the mapping value f of the crowd density, the weight of the connection line can be calculated as w=d×f. It can be seen that the smaller the distance d between the actual physical locations corresponding to vertices N1 and N2, the smaller the weight w. The greater the crowd density of the connection line, the smaller its mapping value f, and the correspondingly smaller the weight w.
[0030] In addition, according to design requirements, those skilled in the art may determine the weight of the connection line based on other factors, and the technical solution of the present disclosure is not limited to the two factors of actual physical distance and crowd density.
[0031] After constructing the adjacency list, the adjacency list can be used to determine the optimal path among the multiple paths from one vertex to another vertex in the region segmentation graph, so that the optimal path table can be generated. In the present disclosure, the optimal path refers to the following path: the sum of the weights of the various connecting lines constituting the path is the smallest. "The sum of weights is the smallest" means that the optimal path corresponds to a smaller actual physical distance and a larger crowd density.
[0032] As an example, the Dijkstra algorithm can be used to calculate the optimal path. Since the Dijkstra algorithm is a known technology, its detailed description is omitted in the present disclosure. In addition, those skilled in the art can also use other known methods to calculate the optimal path, and the present disclosure does not limit this.
[0033] The following reference Figure 5 and Figure 6 To describe the technical solution of generating trajectory segments of a target according to the contents of the present disclosure. Figure 5 A flow chart showing a method of generating trajectory segments for a target. Figure 6 Shows the implementation Figure 5 The method function module.
[0034] Execute when multiple trajectory segments belonging to a target (e.g., a person) have been extracted from videos captured by multiple cameras Figure 5 Methods. Figure 5 As shown, in step S510, based on the start time of each track segment, multiple track segments are sorted in order from early to late. This step is performed by Figure 6 The sorting module 610 in is executed.
[0035] In addition, since multiple trajectory segments of a target may be captured by different cameras, the sorted adjacent trajectory segments may contain parts corresponding to the same time period. Figure 7 As schematically shown, track segments A and track segments B captured by different cameras contain a portion corresponding to the same time period (due to different installation locations of the cameras, the video content of this portion in track segments A and B is different). To facilitate subsequent processing, it is preferred that the temporally repeated portion is deleted from one of track segments A and B, and the remaining portion is merged with the other track segment. Figure 7 As shown in (a), the temporally repeated part is deleted from the track segment B, and the remaining part is merged with the track segment A. Figure 7 As shown in (b), all of trajectory segment B overlaps with trajectory segment A in time, so trajectory segment B can be deleted and only trajectory segment A is retained.
[0036] Return to reference Figure 5 In step S520, for any two adjacent track segments in the sorted multiple track segments, preferably for any two adjacent track segments in the multiple track segments obtained after merging, it is determined whether to generate a virtual track segment for connecting them according to the time distance and spatial distance between them. This step is performed by Figure 6 The determination module 620 in is executed.
[0037] Specifically, assuming that the first trajectory segment and the second trajectory segment are two adjacent trajectory segments, the difference between the end time of the first trajectory segment and the start time of the second trajectory segment is calculated as the time distance between the first trajectory segment and the second trajectory segment. In addition, the distance between the first position of the target in the floor plan at the end of the first trajectory segment and the second position of the target in the floor plan at the start of the second trajectory segment is calculated as the spatial distance between the first trajectory segment and the second trajectory segment.
[0038] Then, when the time distance is greater than the first threshold and the spatial distance is greater than the second threshold, it is determined that a virtual trajectory segment for connecting the first trajectory segment and the second trajectory segment needs to be generated. Conversely, when the time distance is less than or equal to the first threshold and / or the spatial distance is less than or equal to the second threshold, it is determined that a virtual trajectory segment does not need to be generated. It is easy to understand that those skilled in the art can appropriately set the first threshold and the second threshold according to actual application requirements.
[0039] In the case of determining to generate a virtual track segment, a virtual track segment for connecting the first track segment and the second track segment is generated based on the region segmentation map and the optimal path table described above, as shown in step S530. This step is performed by Figure 6 The virtual trajectory segment generation module 630 in is executed.
[0040] The following will be combined Figure 8 The generation of virtual trajectory segments is described in detail. Figure 8 As shown, the first trajectory segment and the second trajectory segment are represented by curved lines with arrows, respectively. In addition, the symbol ai represents a vertex in the region segmentation graph, and the symbol AL represents a connecting line between vertices.
[0041] The first position p1 and the forward direction of the target at the end of the first trajectory segment are determined, and the first vertex b closest to the first position p1 is selected along the forward direction. In addition, the second position p2 and the source direction of the target at the beginning of the second trajectory segment are determined, and the second vertex h closest to the second position p2 is selected along the source direction.
[0042] Determine the optimal path between the first vertex b and the second vertex h based on the optimal path table. Figure 8 As shown, the determined optimal path is composed of connecting line I and connecting line J.
[0043] Then, a virtual track segment is generated by sequentially connecting the first position p1, the first vertex b, the optimal path (connecting lines I and J), the second vertex h, and the second position p2. Preferably, the virtual track segment may be further smoothed to make it look more natural. Figure 8As shown, the generated virtual trajectory segment is connected between the discontinuous first trajectory segment and the second trajectory segment, and the three trajectory segments form a complete and continuous trajectory that is easy to identify.
[0044] Fig. 9 A virtual trajectory segment generated according to the method of the present disclosure is schematically shown. Fig. 9 (a) shows a first trajectory segment and a second trajectory segment later in time than the first trajectory segment based on a plan layout diagram, wherein a circle represents the beginning of the trajectory segment and a square represents the end of the trajectory segment. Fig. 9 (b) shows a virtual trajectory segment generated using the method described above, as indicated by the dotted line.
[0045] The technology of the present disclosure has been described above in conjunction with specific embodiments. By generating virtual trajectory segments to connect discontinuous trajectory segments, a complete and continuous trajectory in time and space can be formed, which makes it easier for monitoring personnel and analysts to intuitively identify the action trajectory of the tracked target. In addition, the generated virtual trajectory segments mainly include optimal paths, and the optimal path corresponds to a smaller physical distance and a larger crowd density. Therefore, the virtual trajectory segments predicted by the method of the present disclosure will be closer to the real trajectory of the target, because people usually choose paths with short distances and / or large traffic, rather than paths with long distances or few people.
[0046] In addition, it should be noted that, although there is a possibility that the predicted virtual trajectory segment is different from the actual trajectory of the target, when the actual trajectory of the target is unknown, the virtual trajectory segment predicted according to the method of the present disclosure can still provide a reference for monitoring and analysis personnel, thereby reducing the difficulty of work.
[0047] The methods described in the above embodiments can be implemented by software, hardware, or a combination of software and hardware. The programs included in the software can be pre-stored in a storage medium provided inside or outside the device. As an example, during execution, these programs are written to a random access memory (RAM) and executed by a processor (e.g., CPU), thereby implementing the various methods and processes described in this article.
[0048] Fig.10 The block diagram shows an example configuration of computer hardware that executes the method of the present disclosure according to a program. The apparatus for generating trajectory segments of a target according to the present disclosure can be implemented based on the computer hardware.
[0049] like Fig.10As shown, in a computer 1000 , a central processing unit (CPU) 1001 , a read only memory (ROM) 1002 , and a random access memory (RAM) 1003 are connected to one another via a bus 1004 .
[0050] The input / output interface 1005 is further connected to the bus 1004. The input / output interface 1005 is connected with the following components: an input device 1006 formed with a keyboard, a mouse, a microphone, etc.; an output device 1007 formed with a display, a speaker, etc.; a storage device 1008 formed with a hard disk, a nonvolatile memory, etc.; a communication device 1009 formed with a network interface card (such as a local area network (LAN) card, a modem, etc.); and a drive 1010 for driving a removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory.
[0051] In the computer having the above-described structure, the CPU 1001 loads the program stored in the storage device 1008 into the RAM 1003 via the input / output interface 1005 and the bus 1004 and executes the program so as to perform the method described above.
[0052] The program to be executed by the computer (CPU 1001) may be recorded on a removable medium 1011 as a package medium, which is formed in the form of, for example, a magnetic disk (including a floppy disk), an optical disk (including a compact disk-read only memory (CD-ROM)), a digital versatile disk (DVD), etc.), a magneto-optical disk, or a semiconductor memory. In addition, the program to be executed by the computer (CPU 1001) may also be provided via a wired or wireless transmission medium such as a local area network, the Internet, or digital satellite broadcasting.
[0053] When the removable medium 1011 is mounted in the drive 1010, the program can be installed in the storage device 1008 via the input / output interface 1005. In addition, the program can be received by the communication device 1009 via a wired or wireless transmission medium and installed in the storage device 1008. Alternatively, the program can be pre-installed in the ROM 1002 or the storage device 1008.
[0054] The program executed by the computer may be a program that performs processing according to the sequence described in the present disclosure, or may be a program that performs processing in parallel or performs processing when necessary such as when called.
[0055] The modules or devices described herein are only in a logical sense and do not strictly correspond to physical devices or entities. For example, the functions of each module described herein may be implemented by multiple physical entities, or the functions of multiple modules described herein may be implemented by a single physical entity. In addition, the features, components, elements, steps, etc. described in one embodiment are not limited to the embodiment, but may also be applied to other embodiments, such as replacing specific features, components, elements, steps, etc. in other embodiments, or combining them.
[0056] The scope of the present disclosure is not limited to the specific embodiments described herein. It should be understood by those skilled in the art that, depending on design requirements and other factors, various modifications or changes may be made to the embodiments herein without departing from the principles and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims and their equivalents.
[0057] In addition, the present disclosure also includes the following embodiments.
[0058] (1) A method for generating a trajectory segment of a target for target tracking, comprising:
[0059] Based on the start time of the trajectory segments, sorting the multiple trajectory segments of the target captured by the multiple camera devices in order from early to late;
[0060] determining whether to generate a virtual trajectory segment based on a time distance and a space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space;
[0061] In the case where it is determined that the virtual trajectory segment is to be generated, the virtual trajectory segment is generated based on a pre-generated region segmentation map and an optimal path table,
[0062] The region segmentation graph indicates the segmentation of the plane layout graph of the scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices.
[0063] The optimal path table indicates the optimal path between any two vertices in the region segmentation graph.
[0064] (2) The method according to (1), further comprising:
[0065] After the sorting, if it is detected that any two adjacent trajectory segments among the plurality of trajectory segments contain parts corresponding to the same time period, deleting the parts from one of the adjacent trajectory segments and merging the remaining parts with another one of the adjacent trajectory segments;
[0066] For the multiple trajectory segments obtained after merging, the temporal distance and spatial distance between adjacent trajectory segments are calculated.
[0067] (3) The method according to (2), wherein the adjacent trajectory segments include a first trajectory segment and a second trajectory segment, and the method further includes:
[0068] Calculating the difference between the end time of the first trajectory segment and the start time of the second trajectory segment as the time distance;
[0069] The distance between a first position of the target in the plan layout diagram at the end of the first trajectory segment and a second position of the target in the plan layout diagram at the start of the second trajectory segment is calculated as the spatial distance.
[0070] (4) The method according to (3), further comprising:
[0071] determining one or more obstacles in the floor plan;
[0072] Extract multiple points on the contour of each obstacle;
[0073] Generate a Voronoi diagram based on the extracted points;
[0074] The vertices located within the outline of the obstacle are removed from the Voronoi map to generate the region segmentation map.
[0075] (5) The method according to (4), wherein the extracted points include vertices of the contour and points on the edges of the contour, and wherein the number of points extracted for each edge of the contour depends on the length of the edge.
[0076] (6) The method according to (4) further includes constructing an adjacency list based on the region segmentation graph, wherein each element in the adjacency list indicates two vertices connected to each other in the region segmentation graph and a weight applied to a connecting line between the two vertices,
[0077] The weight is determined based on one or more of the following:
[0078] The distance between two locations in the plan layout diagram corresponding to the two vertices; and
[0079] The crowd density corresponding to the connecting line between the two vertices.
[0080] (7) The method according to (6), further comprising:
[0081] Based on the adjacency list, determining an optimal path among a plurality of paths from one vertex to another vertex of any two vertices in the region segmentation graph to generate the optimal path table,
[0082] The optimal path includes one or more connecting lines, and the sum of the weights of the one or more connecting lines is the smallest.
[0083] (8) The method according to (7), further comprising:
[0084] For the end of the first trajectory segment, determining the first position and the advancing direction of the target on the plane layout map, and selecting a first vertex closest to the first position along the advancing direction on the region segmentation map;
[0085] For the start of the second trajectory segment, determine the second position and source direction of the target on the plane layout graph, and select a second vertex closest to the second position along the source direction on the region segmentation graph;
[0086] Determine an optimal path between the first vertex and the second vertex in the region segmentation graph based on the optimal path table;
[0087] The virtual trajectory segment is generated by sequentially connecting the first position, the first vertex, the optimal path, the second vertex, and the second position.
[0088] (9) A device for generating a trajectory segment of a target for target tracking, comprising:
[0089] a memory storing a computer program; and
[0090] A processor, the processor being configured to perform the following operations by executing the computer program:
[0091] Based on the start time of the trajectory segments, sorting the multiple trajectory segments of the target captured by the multiple camera devices in order from early to late;
[0092] determining whether to generate a virtual trajectory segment based on a time distance and a space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space;
[0093] In the case where it is determined that the virtual trajectory segment is to be generated, the virtual trajectory segment is generated based on a pre-generated region segmentation map and an optimal path table,
[0094] The region segmentation graph indicates the segmentation of the plane layout graph of the scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices.
[0095] The optimal path table indicates the optimal path between any two vertices in the region segmentation graph.
[0096] (10) According to the apparatus described in (9), the operation further includes:
[0097] After the sorting, if it is detected that any two adjacent trajectory segments among the plurality of trajectory segments contain parts corresponding to the same time period, deleting the parts from one of the adjacent trajectory segments and merging the remaining parts with another one of the adjacent trajectory segments;
[0098] For the multiple trajectory segments obtained after merging, the temporal distance and spatial distance between adjacent trajectory segments are calculated.
[0099] (11) The apparatus according to (10), wherein the adjacent trajectory segments include a first trajectory segment and a second trajectory segment, and the operation further includes:
[0100] Calculating the difference between the end time of the first trajectory segment and the start time of the second trajectory segment as the time distance;
[0101] The distance between a first position of the target in the plan layout diagram at the end of the first trajectory segment and a second position of the target in the plan layout diagram at the start of the second trajectory segment is calculated as the spatial distance.
[0102] (12) According to the apparatus described in (11), the operation further includes:
[0103] determining one or more obstacles in the floor plan;
[0104] Extract multiple points on the contour of each obstacle;
[0105] Generate a Voronoi diagram based on the extracted points;
[0106] The vertices located within the outline of the obstacle are removed from the Voronoi map to generate the region segmentation map.
[0107] (13) The apparatus according to (12), wherein the extracted points include vertices of the contour and points on the edges of the contour, and wherein the number of points extracted for each edge of the contour depends on the length of the edge.
[0108] (14) According to the apparatus described in (12), the operation also includes constructing an adjacency list based on the region segmentation graph, wherein each element in the adjacency list indicates two vertices connected to each other in the region segmentation graph and a weight applied to a connecting line between the two vertices,
[0109] The weight is determined based on one or more of the following:
[0110] The distance between two locations in the plan layout diagram corresponding to the two vertices; and
[0111] The crowd density corresponding to the connecting line between the two vertices.
[0112] (15) According to the apparatus described in (14), the operation further includes:
[0113] Based on the adjacency list, determining an optimal path among a plurality of paths from one vertex to another vertex of any two vertices in the region segmentation graph to generate the optimal path table,
[0114] The optimal path includes one or more connecting lines, and the sum of the weights of the one or more connecting lines is the smallest.
[0115] (16) According to the apparatus described in (15), the operation further includes:
[0116] For the end of the first trajectory segment, determining the first position and the advancing direction of the target on the plane layout map, and selecting a first vertex closest to the first position along the advancing direction on the region segmentation map;
[0117] For the start of the second trajectory segment, determine the second position and source direction of the target on the plane layout graph, and select a second vertex closest to the second position along the source direction on the region segmentation graph;
[0118] Determine an optimal path between the first vertex and the second vertex in the region segmentation graph based on the optimal path table;
[0119] The virtual trajectory segment is generated by sequentially connecting the first position, the first vertex, the optimal path, the second vertex, and the second position.
[0120] (17) A non-transitory computer-readable medium storing a program, wherein when the program is executed by a computer, the computer performs the method for generating trajectory segments of a target according to any one of (1) to (8).
Claims
1. A method for generating a trajectory segment of a target for target tracking, comprising: Based on the start time of the trajectory segments, a plurality of trajectory segments of the target captured by a plurality of camera devices are sorted in order from early to late; determining whether to generate a virtual trajectory segment based on a time distance and a space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space; In the case where it is determined that the virtual trajectory segment is to be generated, the virtual trajectory segment is generated based on a pre-generated region segmentation map and an optimal path table, The region segmentation graph indicates the segmentation of the plane layout graph of the scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices. The optimal path table indicates the optimal path between any two vertices in the region segmentation graph.
2. The method according to claim 1, further comprising: After the sorting, if it is detected that any two adjacent trajectory segments among the plurality of trajectory segments contain parts corresponding to the same time period, deleting the parts from one of the adjacent trajectory segments and merging the remaining parts with another one of the adjacent trajectory segments; For the multiple trajectory segments obtained after merging, the temporal distance and spatial distance between adjacent trajectory segments are calculated.
3. The method according to claim 2, wherein: The adjacent trajectory segments include a first trajectory segment and a second trajectory segment, and the method further includes: Calculating the difference between the end time of the first trajectory segment and the start time of the second trajectory segment as the time distance; The distance between a first position of the target in the plan layout diagram at the end of the first trajectory segment and a second position of the target in the plan layout diagram at the start of the second trajectory segment is calculated as the spatial distance.
4. The method according to claim 3, further comprising: determining one or more obstacles in the floor plan; Extract multiple points on the contour of each obstacle; Generate a Voronoi diagram based on the extracted points; The vertices located within the outline of the obstacle are removed from the Voronoi map to generate the region segmentation map.
5. The method according to claim 4, wherein: The extracted points include vertices of the contour and points on the edges of the contour, and wherein the number of points extracted for each edge of the contour depends on the length of the edge.
6. The method according to claim 4 further comprises constructing an adjacency table based on the region segmentation graph, wherein: Each element in the adjacency list indicates two vertices connected to each other in the region segmentation graph and a weight applied to a connecting line between the two vertices. The weight is determined based on one or more of the following: The distance between two locations in the plan layout diagram corresponding to the two vertices; and The crowd density corresponding to the connecting line between the two vertices.
7. The method according to claim 6, further comprising: Based on the adjacency list, determining an optimal path among a plurality of paths from one vertex to another vertex of any two vertices in the region segmentation graph to generate the optimal path table, The optimal path includes one or more connecting lines, and the sum of the weights of the one or more connecting lines is the smallest.
8. The method according to claim 7, further comprising: For the end of the first trajectory segment, determining the first position and the advancing direction of the target on the plane layout map, and selecting a first vertex closest to the first position along the advancing direction on the region segmentation map; For the start of the second trajectory segment, determine the second position and source direction of the target on the plane layout graph, and select a second vertex closest to the second position along the source direction on the region segmentation graph; Determine an optimal path between the first vertex and the second vertex in the region segmentation graph based on the optimal path table; The virtual trajectory segment is generated by sequentially connecting the first position, the first vertex, the optimal path, the second vertex, and the second position.
9. A device for generating a trajectory segment of a target for target tracking, comprising: a memory storing a computer program; as well as A processor, the processor being configured to perform the following operations by executing the computer program: Based on the start time of the trajectory segments, a plurality of trajectory segments of the target captured by a plurality of camera devices are sorted in order from early to late; determining whether to generate a virtual trajectory segment based on a time distance and a space distance between any two adjacent trajectory segments among the plurality of trajectory segments, wherein the virtual trajectory segment is used to connect the adjacent trajectory segments to form a trajectory of the target that is continuous in time and space; In the case where it is determined that the virtual trajectory segment is to be generated, the virtual trajectory segment is generated based on a pre-generated region segmentation map and an optimal path table, The region segmentation graph indicates the segmentation of the plane layout graph of the scene where the target is located, and includes a plurality of vertices and connecting lines between connected vertices among the plurality of vertices. The optimal path table indicates the optimal path between any two vertices in the region segmentation graph. 10 . A non-transitory computer-readable medium storing a program, wherein when the program is executed by a computer, the computer is caused to perform the method for generating trajectory segments of a target according to any one of claims 1 to 8 .