Target re-finding method and device, electronic equipment and computer readable medium
By updating the state vector model and state transfer equation, determining the target recovery range, and performing object detection, the problem of difficult re-retrieval in the prior art after being blocked for a long time is solved, and efficient target recovery and tracking is achieved.
Patent Information
- Application Number
- CN202411699749.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-26
- Publication Date
- 2025-03-11
AI Technical Summary
The existing target retrieval method cannot effectively deal with the situation where the target is blocked for a long time, which takes a long time and has high computational complexity, making it difficult to find the target again after the target is lost.
By using the historical target tracking results of the target to be retrieved in the historical frame, the state vector model and its state transition equation are updated, the target recovery range is determined, and the target detection is performed within the range of the frame to be processed to achieve re-retrieval of the target.
This method can effectively deal with the situation where the target is blocked for a long time, reduces time-consuming and computational complexity, improves the efficiency of target re-retrieval, and thus improves the overall efficiency of target tracking.
Smart Images
Figure CN119600053A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of target tracking, and in particular, to a target re-recovery method, device, electronic device and computer-readable medium. Background Art
[0002] The target tracking algorithm refers to an algorithm that uses video analysis technology to track the target in the video sequence in real time and obtain its position. It can usually process the video stream in real time and continuously track the target, which is crucial for application scenarios that require instant response. With the continuous optimization of the algorithm and the improvement of computing power, the target tracking algorithm can accurately identify and track targets in complex environments, reduce false positives and false negatives, and can also combine other sensor data to obtain more information such as the target's speed, acceleration, motion trajectory, etc. to provide support for decision-making. It has broad application prospects in many fields such as security monitoring, intelligent transportation, robot navigation, and autonomous driving, and can play an important role.
[0003] In target tracking, target re-finding (also called target re-positioning or target re-capture) is an important step. Although the target tracking algorithm has a certain degree of robustness, the tracking effect may be affected in extreme lighting conditions, complex backgrounds, or fast-moving targets. The algorithm may lose the target when the target is completely blocked, or when the target leaves the field of view, or when the confidence of the target tracking result is low. At this time, the lost target needs to be re-positioned and retrieved to facilitate continued tracking and analysis.
[0004] However, current target re-finding methods have serious drawbacks such as being unable to cope with situations where the target is obscured for a long time and being time-consuming. Summary of the invention
[0005] The present application aims to solve one of the technical problems in the related art to a certain extent. To this end, the present application provides a target retrieving method, device, electronic device and computer readable medium.
[0006] As a first aspect of the present application, a target retrieving method is provided, characterized in that the method comprises:
[0007] Using the historical target tracking results of the target to be found in the historical frames, a preset model including a state vector model and its state transfer equation is updated to determine a first state vector of the target to be found in the last historical frame before it is lost and a second state vector in the frame to be processed after it is lost; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result;
[0008] Determining a target retrieval range according to the first state vector and the second state vector;
[0009] In the target retrieval range of the frame to be processed, target detection is performed according to the second state vector to obtain a target re-retrieval result of the frame to be processed.
[0010] Optionally, the dimension of the elements in the state vector further includes speed, and determining the target retrieval range according to the first state vector and the second state vector includes:
[0011] A target retrieval range is determined according to the speed in the first state vector and the target tracking result in the second state vector.
[0012] Optionally, the target retrieval range is determined by the following formula:
[0013]
[0014]
[0015]
[0016]
[0017] search_area q ={x search_q ,y search_q ,w search_q ,h search_q} (1);
[0018] In formula (1), search_area q Indicates the target retrieval range, x search_q ,y search_q ,w search_q ,h search_q In order, x represents the starting search position of the first coordinate system, the starting search position of the second coordinate system, the search pixel length of the first coordinate system, and the search pixel length of the second coordinate system. q ,y q 、w q and h q represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the target tracking result in the second state vector in sequence, v x,p and v y,p They respectively represent the first coordinate system speed and the second coordinate system speed included in the speed in the first state vector, q and p respectively represent the sequence number of the frame to be processed and the sequence number of the last historical frame, and FPS represents the frame rate.
[0019] Optionally, performing target detection according to the second state vector in the target retrieval range of the frame to be processed to obtain a target re-retrieval result of the frame to be processed includes:
[0020] Performing target detection in the target retrieval range of the frame to be processed to obtain a target detection result;
[0021] Calculating the Euclidean distance between each of the target detection results and the target tracking result in the second state vector;
[0022] The target detection result with the shortest Euclidean distance among the target detection results is determined as the target re-finding result.
[0023] Optionally, the Euclidean distance is calculated by the following formula (2):
[0024]
[0025] In formula (2), represents the Euclidean distance between the mth target detection result and the target tracking result in the second state vector, x q ,y q 、w q and h q In sequence, x represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, and the second coordinate system pixel length included in the target tracking result in the second state vector. m ,y m 、w m and h m The first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the m-th target detection result are respectively represented in sequence.
[0026] Optionally, the method further comprises:
[0027] In the case that no target detection result is obtained by performing target detection in the target retrieval range of the frame to be processed, a frame subsequent to the frame to be processed is determined as a current frame to be processed.
[0028] Optionally, the state vector model includes the following formula (3):
[0029] X t =[x t ,y t ,w t ,h t ,v x,t ,v y,t ,a x,t ,a y,t ] T(3);
[0030] The state transfer equation includes the following formula (4):
[0031]
[0032]
[0033] In formulas (3) and (4), X t represents the state vector corresponding to time t, x t ,y t 、w t 、h t 、v x,t 、v y,t 、a x,t 、a y,t w respectively represent the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, the second coordinate system pixel length, the first coordinate system speed, the second coordinate system speed, the first coordinate system acceleration and the second coordinate system acceleration corresponding to the target at time t in sequence; t represents the state transition process noise corresponding to time t, v t represents the observation process noise corresponding to time t, w t The probability distribution and v t The probability distribution of all obeys the normal distribution, T represents the time constant and its value is 1.
[0034] As a second aspect of the present application, a target retrieving device is provided, wherein the device comprises:
[0035] An updating module, used to update a preset model including a state vector model and its state transfer equation using the historical target tracking results of the target to be found in the historical frames, and determine a first state vector in the last historical frame before the target to be found is lost and a second state vector in the frame to be processed after the target is lost; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result;
[0036] A range determination module, configured to determine a target retrieval range according to the first state vector and the second state vector;
[0037] The target detection module is used to perform target detection according to the second state vector in the target retrieval range of the frame to be processed, and obtain the target re-retrieval result of the frame to be processed.
[0038] As a third aspect of the present application, an electronic device is provided, wherein the electronic device includes:
[0039] one or more processors;
[0040] A memory having one or more computer programs stored thereon, wherein when the one or more computer programs are executed by the one or more processors, the one or more processors implement the target retrieving method according to the first aspect of the present application.
[0041] As a fourth aspect of the present application, a computer-readable medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the target retrieving method according to the first aspect of the present application is implemented.
[0042] In the target retrieving method provided in the embodiment of the present application, the historical target tracking results of the target to be retrieved in the historical frame are used to update the preset model including the state vector model and its state transfer equation, and determine the first state vector of the target to be retrieved in the last historical frame before the loss and the second state vector in the frame to be processed after the loss; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result; according to the first state vector and the second state vector, the target retrieval range is determined; in the target retrieval range of the frame to be processed, the target detection is performed according to the second state vector to obtain the target retrieving result of the frame to be processed. It is simple, convenient and easy to implement, with low time consumption and low computational complexity, which greatly reduces the time consumption and computational complexity, and can effectively deal with the situation where the target is blocked for a long time, avoiding the situation where the target can no longer be associated after it is lost, improving the efficiency of target retrieving, and thus improving the efficiency of target tracking. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] The present application is further described below in conjunction with the accompanying drawings:
[0044] Figure 1 It is a flowchart of an implementation method of the target re-finding method provided in the embodiment of the present application;
[0045] Figure 2 This is a flow chart of an implementation method of determining a target retrieval range provided in an embodiment of the present application;
[0046] Figure 3 It is a flowchart of an implementation method of obtaining a target retrieving result of a frame to be processed provided by an embodiment of the present application;
[0047] Figure 4 is a flow chart of another implementation of the target retrieving method provided in an embodiment of the present application;
[0048] Figure 5 It is a module diagram of an implementation method of a target retrieving device provided in an embodiment of the present application;
[0049] Figure 6 It is a module diagram of an implementation of an electronic device provided in an embodiment of the present application;
[0050] Figure 7 It is a schematic diagram of a computer-readable medium provided in an embodiment of the present application.
[0051] Description of Reference Numerals
[0052] 101: Processor 102: Memory
[0053] 103: I / O interface 104: bus DETAILED DESCRIPTION
[0054] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments based on the embodiments are intended to be used to explain the present application and cannot be understood as limiting the present application.
[0055] References to "one embodiment" or "an example" or "an example" in this specification mean that a particular feature, structure, or characteristic described in conjunction with the embodiment itself may be included in at least one embodiment disclosed in the present application. The appearance of the phrase "in one embodiment" in various places in the specification does not necessarily refer to the same embodiment.
[0056] In the process of target tracking, the target often fails to be tracked or is unreliable due to various factors such as posture changes, light background environment, and external objects blocking it. These situations often lead to target loss and tracking termination. In certain specific situations, target loss can have disastrous consequences. After the target is lost, it is necessary to relocate and retrieve the lost target to facilitate continued tracking and analysis. However, the current target re-retrieval method has serious disadvantages such as being unable to cope with the situation where the target is blocked for a long time and taking a long time.
[0057] For example, there is currently a method for predicting the target position through a deep network model (Transformer and CNN). Although the method has high accuracy, it cannot cope with the situation where the target is blocked for a long time. Once the target is blocked for a long time, the target cannot be associated with it later, the target will be lost directly, and the tracking will be disconnected. For another example, there is currently a method for updating the sensor identification and target identification corresponding to each fused target in the common view area and the position information in the bird's-eye view of the target section based on the target matching results, and then obtaining the continuous display results of each fused target in the target section. This method is very time-consuming in actual application, brings extremely high computational complexity to the equipment, and increases the operating load of the equipment.
[0058] In view of this, the inventor proposes to use the historical target tracking results of the target to be recovered before it is lost to update the preset model including the state vector model and its state transfer equation, construct an operation model, and plan the target recovery range, so as to quickly and accurately recover the lost target in the processing frame after the loss.
[0059] As a first aspect of the present application, a target retrieving method is provided, such as Figure 1 As shown, the method may include the following steps:
[0060] In step S110, the preset model including the state vector model and its state transfer equation is updated using the historical target tracking results of the target to be found in the historical frames, and the first state vector of the target to be found in the last historical frame before it is lost and the second state vector in the frame to be processed after it is lost are determined; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result;
[0061] In step S120, a target retrieval range is determined according to the first state vector and the second state vector;
[0062] In step S130, target detection is performed in the target retrieval range of the frame to be processed according to the second state vector to obtain a target re-retrieval result of the frame to be processed.
[0063] The target to be found refers to the target that is lost during the target tracking process. The historical frame refers to the frame in which a reliable target tracking result has been successfully obtained for the target to be found, that is, the frame before the target to be found is lost. The pending frame refers to the frame in which a reliable target tracking result has not been successfully obtained for the target to be found, including the frame in which the target tracking result has not been obtained for the target to be found and the frame in which the target tracking result obtained is unreliable, that is, the frame after the target to be found is lost.
[0064] It is understandable that the last historical frame before the loss refers to the latest frame among all frames before the target to be found is lost. The frame to be processed after the loss can be any frame of all frames after the target to be found is lost, that is, even if the target to be found is blocked for a long time, the target re-finding method provided in the embodiment of the present application can be executed to retrieve it.
[0065] There are currently multiple possible implementation methods for how to specifically perform step S110, so this embodiment of the present application will not elaborate on this.
[0066] Among them, the dimensions of each element in the target tracking result of the kth frame usually include x, y, w and h, and x, y, w and h respectively represent the first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length of the target in the kth frame. The first coordinate system position is the horizontal coordinate of the upper left corner of the target in the kth frame, and the second coordinate system position is the vertical coordinate of the upper left corner of the target in the kth frame. Correspondingly, the first coordinate system pixel length and the second coordinate system pixel length correspond to the pixel width value and the pixel height value of the target in the kth frame, respectively. Of course, the horizontal, vertical, width and height can also be reversed. In the embodiment of the present application, the dimensions of the elements in the state vector at least include the dimensions of each element in the target tracking result, that is, the state vector at least includes the first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length.
[0067] That is, the determined second state vector actually includes the predicted target tracking result of the target to be found in the frame to be processed after it is lost. The inventor proposes that, considering the influence of various factors such as extreme lighting conditions, complex backgrounds or rapid movement of targets, the search range can be appropriately expanded near the predicted target tracking result to cope with various situations. Specifically, the target retrieval range is determined based on the first state vector and the second state vector. Finally, target detection is performed in the target re-retrieval range of the frame to be processed.
[0068] In the target retrieving method provided in the embodiment of the present application, the historical target tracking results of the target to be retrieved in the historical frame are used to update the preset model including the state vector model and its state transfer equation, and determine the first state vector of the target to be retrieved in the last historical frame before the loss and the second state vector in the frame to be processed after the loss; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result; according to the first state vector and the second state vector, the target retrieval range is determined; in the target retrieval range of the frame to be processed, the target detection is performed according to the second state vector to obtain the target retrieving result of the frame to be processed. It is simple, convenient and easy to implement, with low time consumption and low computational complexity, which greatly reduces the time consumption and computational complexity, and can effectively deal with the situation where the target is blocked for a long time, avoiding the situation where the target can no longer be associated after it is lost, improving the efficiency of target retrieving, and thus improving the efficiency of target tracking.
[0069] The inventor proposes that speed can effectively characterize the movement trend of the target to be found in the historical frame before it is lost. Combined with the speed of the target to be found in the last historical frame before it is lost, the impact of the sudden change of the target state can be taken into account. Accordingly, in some embodiments, the dimension of the elements in the state vector also includes speed. The target retrieval range is determined according to the first state vector and the second state vector (i.e., step S120), such as Figure 2 As shown, the following steps may be included:
[0070] In step S121, a target retrieval range is determined according to the speed in the first state vector and the target tracking result in the second state vector.
[0071] In some embodiments, the target retrieval range is determined by the following formula:
[0072]
[0073]
[0074]
[0075]
[0076] search_area q ={x search_q ,y search_q ,w search_q ,h search_q} (1);
[0077] In formula (1), search_area q Indicates the target retrieval range, x search_q ,y search_q ,w search_q ,h search_q In order, x represents the starting search position of the first coordinate system, the starting search position of the second coordinate system, the search pixel length of the first coordinate system, and the search pixel length of the second coordinate system. q ,y q 、w q and h q represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the target tracking result in the second state vector in sequence, v x,p and v y,p They respectively represent the first coordinate system speed and the second coordinate system speed included in the speed in the first state vector, q and p respectively represent the sequence number of the frame to be processed and the sequence number of the last historical frame, and FPS represents the frame rate.
[0078] Among them, the first state vector can be expressed as X p =[x p ,y p ,w p ,h p ,v x,p ,v y,p ,a x,p ,a y,p ] T , the second state vector can be expressed as X q =[x q ,y q ,w q ,h q ,v x,q ,v y,q ,a x,q ,a y,q ] T ,v x,p and v y,p is the velocity in the first state vector, x q ,y q ,w q ,h q is the target tracking result in the second state vector.
[0079] It should be noted that when defining the target retrieval range, search_area q Need to be in the image frame, that is, x search_q and search_q The coordinates of pixels that do not exceed the edge of the image frame need to be w search_q and h search_q Need to be greater than 0 and less than the image frame width and height respectively.
[0080] In some embodiments, in the target retrieval range of the frame to be processed, target detection is performed according to the second state vector to obtain the target retrieval result of the frame to be processed (ie, step S130), such as Figure 3 As shown, the following steps may be included:
[0081] In step S131, target detection is performed in the target retrieval range of the frame to be processed to obtain a target detection result;
[0082] In step S132, the Euclidean distance between each of the target detection results and the target tracking result in the second state vector is calculated;
[0083] In step S133, the target detection result with the shortest Euclidean distance among the target detection results is determined as the target retrieval result.
[0084] It can be understood that if the target detection fails to obtain the target detection result from the target retrieval range of the frame to be processed, it may be caused by the target to be retrieved being completely unobstructed or out of the picture. It can be regarded as a target retrieval failure, and the target retrieval method provided in the embodiment of the present application is continued to be executed on the next frame to be processed.
[0085] It can be understood that, in the case where only one target detection result is obtained by performing target detection in the target retrieval range of the frame to be processed, the target detection result can be directly determined as the target retrieval result.
[0086] The mth target detection result can be expressed as [x m ,y m ,w m ,h m ], calculate its difference with the second state vector X q =[x q ,y q ,w q ,h q ,v x,q ,v y,q ,a x,q ,a y,q ] T Target tracking results[x q ,y q ,w q ,h q ] Finally, the target detection result corresponding to the shortest Euclidean distance among all Euclidean distances is determined: The target detection result corresponding to the shortest Euclidean distance [x i ,y i ,w i ,h i ] to confirm the target and retrieve the result.
[0087] In some embodiments, the Euclidean distance is calculated by the following formula (2):
[0088]
[0089] In formula (2), represents the Euclidean distance between the mth target detection result and the target tracking result in the second state vector, x q ,y q 、w q and h q In sequence, x represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, and the second coordinate system pixel length included in the target tracking result in the second state vector. m ,ym 、w m and h m The first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the m-th target detection result are respectively represented in sequence.
[0090] In some embodiments, Figure 4 As shown, the method may further include the following steps:
[0091] In step S140, when no target detection result is obtained by performing target detection in the target retrieval range of the frame to be processed, a frame subsequent to the frame to be processed is determined as a current frame to be processed.
[0092] Among them, by determining the next frame of the pending frame in which the target re-recovery fails as the current pending frame, the target re-recovery method provided in the embodiment of the present application is continued to be executed until the target to be found is found in the pending frame, thereby facilitating the continuation of target tracking and improving the efficiency of target tracking.
[0093] In some embodiments, the state vector model includes the following formula (3):
[0094] X t =[x t ,y t ,w t ,h t ,v x,t ,v y,t ,a x,t ,a y,t ] T (3);
[0095] The state transfer equation includes the following formula (4):
[0096]
[0097]
[0098] In formulas (3) and (4), X t represents the state vector corresponding to time t, x t ,y t 、w t 、h t 、v x,t 、v y,t 、a x,t 、a y,tw respectively represent the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, the second coordinate system pixel length, the first coordinate system speed, the second coordinate system speed, the first coordinate system acceleration and the second coordinate system acceleration corresponding to the target at time t in sequence; t represents the state transition process noise corresponding to time t, v t represents the observation process noise corresponding to time t, w t The probability distribution and v t The probability distribution of all obeys the normal distribution, T represents the time constant and its value is 1.
[0099] Among them, considering the influence of measurement error and other uncertain factors, noise w is introduced into the state transfer equation t and v t , w t It represents the uncertainty in the state transition process, v t It represents the uncertainty in the measurement process.
[0100] As a second aspect of the embodiment of the present application, a target retrieving device is provided, such as Figure 5 As shown, the device may include:
[0101] The updating module 110 is used to update the preset model including the state vector model and its state transfer equation using the historical target tracking results of the target to be found in the historical frames, and determine the first state vector of the target to be found in the last historical frame before it is lost and the second state vector in the frame to be processed after it is lost; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result;
[0102] A range determination module 120, configured to determine a target retrieval range according to the first state vector and the second state vector;
[0103] The target detection module 130 is used to perform target detection according to the second state vector in the target retrieval range of the frame to be processed, and obtain a target re-retrieval result of the frame to be processed.
[0104] As a third aspect of the embodiments of the present application, an electronic device is provided, wherein: Figure 6 As shown, the electronic device includes:
[0105] One or more processors 101;
[0106] The memory 102 stores one or more computer programs. When the one or more computer programs are executed by the one or more processors 101, the one or more processors 101 implement the target retrieving method provided in the first aspect of the embodiment of the present application.
[0107] The electronic device may further include one or more I / O interfaces 103 connected between the processor 101 and the memory 102 and configured to implement information interaction between the processor 101 and the memory 102 .
[0108] Among them, the processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU), etc.; the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory (FLASH); the I / O interface (read-write interface) is connected between the processor and the memory, and can realize information exchange between the processor and the memory, including but not limited to a data bus (Bus), etc.
[0109] In some embodiments, the processor 101 , the memory 102 , and the I / O interface 103 are connected to each other via a bus 104 , and further connected to other components of the computing device.
[0110] As a fourth aspect of the embodiment of the present application, Figure 7 As shown, a computer-readable medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, the target retrieving method provided in the first aspect of the embodiment of the present application is implemented.
[0111] It can be understood by a person of ordinary skill in the art that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. Accordingly, the computer program can be stored in a non-volatile computer-readable storage medium, and the computer program can implement the method of any of the above-mentioned embodiments when executed. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the embodiments of the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0112] The above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Those skilled in the art should understand that the present application includes but is not limited to the contents described in the drawings and the above specific embodiments. Any modification that does not deviate from the functional and structural principles of the present application will be included in the scope of the claims.
Claims
1. A target retrieving method, characterized in that: The method comprises: Using the historical target tracking results of the target to be found in the historical frames, a preset model including a state vector model and its state transfer equation is updated to determine a first state vector of the target to be found in the last historical frame before it is lost and a second state vector in the frame to be processed after it is lost; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result; Determining a target retrieval range according to the first state vector and the second state vector; In the target retrieval range of the frame to be processed, target detection is performed according to the second state vector to obtain a target re-retrieval result of the frame to be processed.
2. The method according to claim 1, characterized in that The dimension of the elements in the state vector also includes speed, and determining the target retrieval range according to the first state vector and the second state vector includes: A target retrieval range is determined according to the speed in the first state vector and the target tracking result in the second state vector.
3. The method according to claim 2, characterized in that The target retrieval range is determined by the following formula: search_area q ={x search_q ,y search_q ,w search_q ,h search_q } (1); In formula (1), search_area q Indicates the target retrieval range, x search_q ,y search_q ,w search_q ,h search_q In order, x represents the starting search position of the first coordinate system, the starting search position of the second coordinate system, the search pixel length of the first coordinate system, and the search pixel length of the second coordinate system. q ,y q 、w q and h q represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the target tracking result in the second state vector in sequence, v x,p and v y,p They respectively represent the first coordinate system speed and the second coordinate system speed included in the speed in the first state vector, q and [ respectively represent the sequence number of the frame to be processed and the sequence number of the last historical frame, and FPS represents the frame rate.
4. The method according to claim 1, characterized in that: The step of performing target detection according to the second state vector in the target retrieval range of the frame to be processed to obtain a target re-retrieval result of the frame to be processed includes: Performing target detection in the target retrieval range of the frame to be processed to obtain a target detection result; Calculating the Euclidean distance between each of the target detection results and the target tracking result in the second state vector; The target detection result with the shortest Euclidean distance among the target detection results is determined as the target re-finding result.
5. The method according to claim 4, characterized in that The Euclidean distance is calculated by the following formula (2): In formula (2), represents the Euclidean distance between the mth target detection result and the target tracking result in the second state vector, x q ,y q 、w q and h q In sequence, x represents the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, and the second coordinate system pixel length included in the target tracking result in the second state vector. m ,y m 、w m and h m The first coordinate system position, the second coordinate system position, the first coordinate system pixel length and the second coordinate system pixel length included in the m-th target detection result are respectively represented in sequence.
6. The method according to claim 4, characterized in that The method further comprises: In the case that no target detection result is obtained by performing target detection in the target retrieval range of the frame to be processed, a frame subsequent to the frame to be processed is determined as a current frame to be processed.
7. The method according to any one of claims 1 to 6, characterized in that The state vector model includes the following formula (3): X t =[x t ,y t ,w t ,h t ,v z,t ,v y,t ,a x,t ,a y,t ] T (3); The state transfer equation includes the following formula (4): In formulas (3) and (4), X t represents the state vector corresponding to time t, x t ,y t 、w t 、h t 、v x,t 、v y,t 、a x,t 、a y,t w respectively represent the first coordinate system position, the second coordinate system position, the first coordinate system pixel length, the second coordinate system pixel length, the first coordinate system speed, the second coordinate system speed, the first coordinate system acceleration and the second coordinate system acceleration corresponding to the target at time t in sequence; t represents the state transition process noise corresponding to time t, v t represents the observation process noise corresponding to time t, w t The probability distribution and v t The probability distribution of all obeys the normal distribution, T represents the time constant and its value is 1.
8. A target retrieving device, characterized in that: The device comprises: An updating module, used to update a preset model including a state vector model and its state transfer equation using the historical target tracking results of the target to be found in the historical frames, and determine a first state vector in the last historical frame before the target to be found is lost and a second state vector in the frame to be processed after the target is lost; wherein the dimension of the elements in the state vector at least includes the dimension of each element in the target tracking result; A range determination module, configured to determine a target retrieval range according to the first state vector and the second state vector; The target detection module is used to perform target detection according to the second state vector in the target retrieval range of the frame to be processed, and obtain the target re-retrieval result of the frame to be processed.
9. An electronic device, characterized in that: The electronic device comprises: one or more processors; A memory having one or more computer programs stored thereon, wherein when the one or more computer programs are executed by the one or more processors, the one or more processors implement the target retrieving method according to any one of claims 1-7.
10. A computer readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the target retrieving method according to any one of claims 1 to 7 is implemented.