TWI798815B - Target re-identification method, device, and computer readable storage medium - Google Patents

Abstract

This invention provides a target re-identification method, apparatus, and computer readable storage medium, and the method includes obtaining a multiple of current frames collected by a multiple of cameras installed in the monitoring area; performing a target detection based on the current frames to confirm the image of a target icon captured by each camera; performing a quantitative detection based on the image of the target icon captured by each camera to obtain the number of global targets; performing a target re-recognition according to the image of the target icon and a target recognition library, wherein the target recognition library includes at least one target ID and characteristic data; confirming at least one image of an unidentified target icon according to the target re-identification result to create a new ID and label at least one unidentified target icon when the quantity of detected global targets satisfies the preset increase condition; updating the target recognition library according to the new ID and the characteristic data of the at least one image of the unidentified target icon. This method can improve the accuracy and stability of the target re-identification.

Description

Target re-identification method, device and computer-readable storage medium

The invention belongs to the field of identification, and in particular relates to a target re-identification method, device and computer-readable storage medium.

This section is intended to provide a background or context for implementations of the invention that are set forth in the claims. The descriptions herein are not admitted to be prior art by inclusion in this section.

At present, with the popularization of video surveillance technology and the ever-increasing security requirements, target re-identification applied in the field of video surveillance has gradually become one of the hot spots in the field of computer vision research.

It is very important to achieve cross-camera target re-identification in monitoring places with high security requirements such as data centers and shopping malls. In the target re-identification process, when a new target enters the monitoring area, it is necessary to assign a new ID to the new target for subsequent identification. The industry usually uses the calculation of the characteristics between the bull's-eye image and the feature data in the target recognition library. The method of similarity is used to judge whether to create and assign a new ID. In some scenarios, the accuracy of the above judgment will be greatly affected by problems such as target occlusion and shooting angle, which may cause inaccurate target re-identification.

Aiming at the above-mentioned problems existing in the prior art, a target re-identification method, device and computer-readable storage medium are proposed, and the above-mentioned problems can be solved by using the method, device and computer-readable storage medium.

The present invention provides the following solutions.

In the first aspect, a target re-identification method is provided, including: acquiring a plurality of current frames collected by a plurality of cameras arranged in a monitoring area; performing target detection according to the plurality of current frames, and determining a bull's-eye image captured by each camera; Quantity detection is carried out according to the bull's-eye chart image captured by each camera to obtain the number of targets in the whole area; target re-identification is performed according to the bull's-eye chart image and the target recognition library, and the target recognition library includes at least one target's identity and characteristic data; when the global target is detected When the number of targets meets the preset increase conditions, at least one unrecognized bull's-eye image is determined according to the result of target re-identification, and a new identity is created to mark at least one unrecognized bull's-eye image; according to the new identity and at least one unidentified The feature data of the bull's-eye image is updated to the object recognition library.

In a possible implementation manner, performing target detection according to a plurality of current frames further includes: inputting a plurality of current frames into a trained target detection model to extract a bull's-eye image captured by each camera; wherein, the target detection The model is a human detection model based on the YOLOv4-tiny network.

In a possible implementation manner, the method further includes: training the YOLOv4-tiny network according to the real collected images in the monitoring area to obtain a target detection model.

In a possible implementation manner, the bull's-eye chart image is a partial image containing target features in the current frame, and the quantity detection is performed according to the bull's-eye chart images captured by each camera, and further includes: Convert the position of the bull's-eye chart image to obtain the global position corresponding to the bull's-eye chart image captured by each camera; determine the global position coincidence degree of the bull's-eye chart image captured by different cameras, and compare the global position coincidence degree of the different cameras according to the global position coincidence degree The bull's-eye chart images are screened to detect the number of bull-eye chart images retained after filtering.

In a possible implementation manner, the method further includes: when the result of the number detection is less than the number of previous global targets, judging whether there are The target that leaves the monitoring area from the predetermined area; if there is no target, the number of previous global targets will still be retained as the number of global targets determined this time; if there are targets, the result of the number detection will be used as the number of global targets determined this time; Wherein, the number of previous global objects is obtained by performing object detection and number detection on previous frames of multiple current frames.

In a possible implementation manner, performing position conversion on the captured bull's-eye chart image according to the viewing position of each camera further includes: projecting the bottom center point of the bull's-eye chart image in the current frame according to the viewing position of each camera Transform to determine the ground coordinates of each bull's-eye chart image.

In a possible implementation, the method further includes: inputting a plurality of current frames into the trained target quantity detection model to perform target detection and quantity detection to obtain the global target quantity; wherein, the target quantity detection model is based on YOLOv4-tiny A pedestrian count detection model created by the network.

In a possible implementation manner, performing target re-identification according to the bull's-eye chart image and the target recognition library further includes: calculating the similarity between the bull's-eye chart image and the characteristic data in the target recognition library, and according to the calculated similarity, Performing target re-identification on the bull's-eye chart image; when the result of target re-identification indicates that the first bull's-eye chart image matches the first target in the target recognition library, mark the first bull's-eye chart image according to the identity of the first target.

In a possible implementation manner, the method further includes: if the current frame is not the first frame, and the number of global objects corresponding to the current frame increases compared with the number of global objects corresponding to the previous frame, then the number of global objects conforms to the preset increase Condition; if the current frame is the first frame, the number of preset global targets meets the preset increase condition.

In a possible implementation manner, updating the target recognition library according to the new identity and the feature data of at least one unrecognized bull's-eye image also includes: judging whether the at least one unrecognized bull-eye image satisfies a preset image quality condition; The new identity mark and the unrecognized bull's-eye image meeting the preset image quality conditions are correspondingly stored in the target recognition library.

In a possible implementation manner, after performing target re-identification according to the bull's-eye chart image and the target recognition library, the method further includes: re-identifying the first target in the target recognition library according to the first bull's-eye chart image or the feature value of the first bull-eye chart image The feature data is updated dynamically.

In a possible implementation manner, the method further includes replacing and updating the target recognition library, specifically including: performing an update on the target recognition library according to the comparison result of the source time corresponding to the characteristic data of each target in the target recognition library and the current time. Replace and update; and/or, according to the comparison result of the global position corresponding to the feature data of each target in the target recognition library and the current global position of each target, replace and update the target recognition library; and/or, according to the target recognition The feature similarity between multiple feature data of each target in the library is used to replace and update the target recognition library.

In a possible implementation manner, the method further includes: after the quantity of the feature data of any one target exceeds a preset threshold, starting a replacement update.

In a second aspect, a target re-identification device is provided, including: an acquisition module, configured to acquire a plurality of current frames captured by a plurality of cameras arranged in a monitoring area; a target detection module, configured to perform target detection based on a plurality of current frames. Detection, to determine the bull's-eye chart image captured by each camera; the quantity detection module is used to perform quantity detection based on the bull's-eye chart image captured by each camera, and obtain the number of targets in the whole area; the target re-identification module is used to use the bull's-eye chart Image and target recognition library for target re-identification, the target recognition library includes at least one target's identity and characteristic data; the identity identification module is used for when the number of targets in the whole domain is detected to meet the preset increase conditions, according to the result of target re-identification Determine at least one unrecognized bull's-eye image, and create a new identity to mark at least one of the unrecognized bull's-eye image; target recognition library update module, used for feature data according to the new identity and at least one unrecognized bull's-eye image Update object recognition library.

In a possible implementation, the target detection module is also used to: input multiple current frames into the trained target detection model to extract the bull's-eye image captured by each camera; wherein, the target detection model is based on A human detection model created by the YOLOv4-tiny network.

In a possible implementation manner, the target detection module is also used to: train the YOLOv4-tiny network according to the real collected images in the monitoring area to obtain a target detection model.

In a possible implementation manner, the bull's-eye image is a partial image containing target features in the current frame, and the quantity detection module is also used to: perform position conversion on the captured bull's-eye image according to the viewing position of each camera to obtain The global position corresponding to the bull's-eye chart image captured by each camera; determine the global position coincidence degree of the bull's-eye chart image captured by different cameras, and filter the bull's-eye chart images captured by different cameras according to the global position coincidence degree, and detect and filter The number of bullseye chart images to keep.

In a possible implementation manner, the number detection module is also used for: when the result of the number detection is less than the number of previous global objects, then according to the multiple current frames collected by multiple cameras and the last frame of the multiple current frames , to determine whether there is a target that leaves the monitoring area from the predetermined area; if there is no target, the number of previous global targets is still retained as the number of global targets determined this time; if there is a target, the result of the number detection is used as the determined The number of global objects; wherein, the number of previous global objects is obtained by performing object detection and number detection on previous frames of multiple current frames.

In a possible implementation, the number detection module is also used for: performing projective transformation on the bottom central point of the bull's-eye image in the current frame according to the viewing position of each camera, so as to determine the ground coordinates of each bull's-eye image.

In a possible implementation, the device is also used to: input multiple current frames into the trained target quantity detection model to perform target detection and quantity detection to obtain the global target quantity; wherein, the target quantity detection model is based on YOLOv4- Pedestrian count detection model created by tiny network.

In a possible implementation, the target re-identification module is also used to: calculate the similarity between the bull's-eye chart image and the feature data in the target recognition library, and perform target re-targeting on the bull's-eye chart image according to the calculated similarity. Recognition: when the result of target re-identification indicates that the first bull's-eye chart image matches the first target in the target recognition library, mark the first bull's-eye chart image according to the identity of the first target.

In a possible implementation, the identification module is also used to: if the current frame is not the first frame, and the number of global objects corresponding to the current frame increases compared with the number of global objects corresponding to the previous frame, then the number of global objects Meet the preset increase conditions; if the current frame is the first frame, the preset number of global targets meets the preset increase conditions.

In a possible implementation, the target recognition library update module is further used to: determine whether at least one unrecognized bull's-eye image satisfies a preset image quality condition; The image of the recognized bull's-eye chart is correspondingly stored in the target recognition library.

In a possible implementation manner, the target recognition library update module is further configured to: dynamically update the characteristic data of the first target in the target recognition library according to the first bull's-eye image or the feature value of the first bull's-eye image.

In a possible implementation, the target recognition library update module is also used to: replace and update the target recognition library according to the comparison result of the source time corresponding to the characteristic data of each target in the target recognition library and the current time; and /or, according to the comparison result of the global position corresponding to the feature data of each target in the target recognition library and the current global position of each target, the target recognition library is replaced and updated; and/or, according to each target in the target recognition library The feature similarity between multiple feature data of a target is used to replace and update the target recognition library.

In a possible implementation manner, the target recognition library update module is further configured to: start a replacement update after the quantity of feature data of any one target exceeds a preset threshold.

In terms of third parties, a target re-identification device is provided, including: one or more multi-core processors; memory for storing one or more programs; when the one or more programs are processed by the one or more multi-cores When the processor is executed, the one or more multi-core processors are made to implement: the method in the first aspect.

In a fourth aspect, a computer-readable storage medium is provided, the computer-readable storage medium stores a program, and when the program is executed by a multi-core processor, the multi-core processor executes the method according to the first aspect.

The above at least one technical solution adopted in the embodiment of the present application can achieve the following beneficial effects: In this embodiment, the number of global targets in the monitoring area is detected, and the creation and distribution of new identities are controlled by the detected number of global targets, It can well guarantee the accuracy of the distribution of identity marks, and ensure the accuracy and stability of target re-identification.

It should be understood that the above description is only an overview of the technical solution of the present invention, so as to understand the technical means of the present invention more clearly, so as to be implemented according to the contents of the description. In order to make the above and other objects, features and advantages of the present invention more comprehensible, specific embodiments of the present invention are illustrated below.

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

In the description of the embodiments of the present application, it should be understood that terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, acts, components, parts or combinations thereof disclosed in the specification, and do not It is intended to exclude the possibility of the existence of one or more other features, figures, steps, acts, parts, parts or combinations thereof.

Unless otherwise specified, "/" means or, for example, A/B can mean A or B; "and/or" in this article is just a description of the relationship between associated objects, which means that there can be three relationships, for example, A and/or B may mean: A exists alone, A and B exist simultaneously, and B exists alone.

The terms "first", "second", etc. are used for descriptive purposes only, and should not be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, a feature defined as "first", "second", etc. may expressly or implicitly include one or more of that feature. In the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

All codes in this application are exemplary, and those skilled in the art will think of various modifications without departing from the idea of this application according to factors such as the programming language used, specific needs and personal habits.

The target real-time tracking method is characterized in that, comprising: obtaining a plurality of current frames collected by a plurality of cameras arranged in the monitoring area; performing target detection according to the plurality of current frames, and determining a bull's-eye image captured by each camera; Perform quantity detection according to the bull's-eye image captured by each camera to obtain the number of targets in the whole area; perform target re-identification according to the bull's-eye image and a target recognition library, wherein the target recognition library includes at least one target identity and Feature data; when it is detected that the number of targets in the whole domain meets the preset increase condition, at least one unrecognized bull's-eye image is determined according to the result of the target re-identification, and a new identity is created for the at least one unrecognized bull's-eye image Marking; updating the target recognition library according to the new identity and the feature data of the at least one unrecognized bull's-eye image.

In addition, it should be noted that, in the case of no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and examples.

Fig. 1 is a schematic flow diagram of a real-time target tracking method according to an embodiment of the present application, which is used to track a specified target in a preset scene. In this process, from the perspective of equipment, the execution subject may be one or more electronic devices ; From a program point of view, the execution subject can be a program carried on these electronic devices accordingly.

As shown in Figure 1, the method 100 includes:

Step 101, obtaining multiple current frames captured by multiple cameras set in the monitoring area;

Specifically, the monitoring area refers to the sum of the viewing areas of multiple cameras, the multiple cameras include at least two cameras, and the viewing areas of the multiple cameras are adjacent to each other or at least partially overlapped, so that the target to be tracked can be in Move in the monitoring area and appear in the viewing area of any one or more cameras. Wherein, the current frames of the multiple cameras are respectively extracted from the surveillance videos of the multiple cameras, wherein the current frames of each camera have the same acquisition time. Optionally, the tracking target in the present disclosure is preferably a pedestrian. Those skilled in the art can understand that the above tracking target can also be other movable objects, such as animals, vehicles, etc., which is not specifically limited in the present disclosure.

，攝像頭202採集的監控視頻可解析為圖像幀序列

，其中上述解析可以即時線上進行或離線進行。基於此，可以按時序從上述多個圖像幀序列中依次提取兩個攝像頭的當前幀

和

以進行本公開所示出的即時目標跟蹤，其中，下標n的取值可以是

。 For example, in complex monitoring scenarios, such as in corridors, large shopping malls, computer rooms and other places, a large number of cameras are usually used to monitor various areas and obtain multiple monitoring videos. FIG. 2 shows a schematic monitoring scene, in which a camera 201 and a camera 202 are set, and FIG. 3 shows a viewfinder picture of the above-mentioned camera 201 and camera 202 . Wherein, the surveillance video collected by the camera 201 can be parsed into a sequence of image frames

, the surveillance video collected by the camera 202 can be parsed into a sequence of image frames

, wherein the above analysis can be performed online or offline in real time. Based on this, the current frames of the two cameras can be sequentially extracted from the above multiple image frame sequences in sequence

and

To carry out the real-time target tracking shown in this disclosure, wherein, the value of the subscript n can be

.

As shown in Figure 1, the method 100 may include:

Step 102, perform target detection according to multiple current frames, and determine the bull's-eye image captured by each camera;

和

，然後，在任意基於深度學習的目標檢測模型中輸入預處理後的當前幀

和

進行檢測，輸出針對每個攝像頭的一系列行人圖像（靶心圖表像的一種示例）。目標檢測模型比如可以是YOLO(統一即時目標檢測，You Only Look Once)模型等，本公開對此不作具體限制。如圖5所示，示出了對多個當前幀

和

進行檢測得到的多個目標（行人）檢測框，可以理解，根據目標（行人）檢測框從當前幀中可以截取到目標（行人）圖像，其中攝像頭201捕獲到的目標（行人）圖像包括

，攝像頭202捕獲到的目標（行人）包括圖像

。可以將截取出的目標（行人）圖像進行歸一化處理，以便於後續的跟蹤展示。 Specifically, the bull's-eye chart image may be a partial image containing target features in the current frame. For example, as shown in Figure 4, the current frames of the camera 201 and the camera 202 are shown

and

, and then input the preprocessed current frame into any deep learning-based target detection model

and

Detection is performed, outputting a sequence of pedestrian images for each camera (an example of a bull's-eye chart image). The target detection model may be, for example, a YOLO (You Only Look Once) model, etc., which is not specifically limited in the present disclosure. As shown in Figure 5, it shows that for multiple current frames

and

The multiple target (pedestrian) detection frames obtained by detection can be understood that the target (pedestrian) image can be intercepted from the current frame according to the target (pedestrian) detection frame, wherein the target (pedestrian) image captured by the camera 201 includes

, the target (pedestrian) captured by the camera 202 includes the image

. The intercepted target (pedestrian) image can be normalized for subsequent tracking display.

Further, in a possible implementation manner, in order to detect the bull's-eye image more accurately, step 102 may also include: inputting multiple current frames into the trained target detection model to extract the bull's-eye captured by each camera Diagram image; Among them, the target detection model is a human detection model created based on the YOLOv4-tiny network.

Specifically, the instant target detection algorithm YOLOV4-TINY based on deep learning can be improved to obtain YOLOV4-TINY-P (YOLOv4-tiny-People), and trained to generate a human body detection model, which can be used to target the overall characteristics of pedestrians. Recognition, and is not affected by facial occlusions such as wearing a mask. In addition, without a professional face camera, the above-mentioned target detection can be directly completed by using multiple ordinary surveillance cameras.

Optionally, other target detection algorithms may also be used, such as the faster-rcnn target detection algorithm, yolov4 target detection algorithm, etc., which are not specifically limited in this application.

Optionally, for other target detection scenarios such as vehicle detection and animal detection, other target detection models may be used correspondingly, which is not specifically limited in the present application.

Further, in some implementations, in order to make the target detection model maintain high accuracy for specific monitoring scenarios, the following steps can also be performed to obtain the above-mentioned target detection model: according to the real collected images in the monitoring area, pair YOLOv4- The tiny network is trained to obtain a target detection model.

For example, when applied to a computer room scene, targeted training can be performed on pedestrians in actual scenes such as a computer room, and target positive and negative samples are added based on the actual scene. For example, chairs, backpacks, servers, etc. are negative samples, and pedestrians are positive samples. , so as to avoid the misidentification of objects such as backpacks and chair sundries in the distance as pedestrians of different forms due to light reasons. The training data can be jointly trained with the actual computer room scene data, PASCAL VOC2007 and VOC2012 and other target detection data sets to further improve the model detection ability.

As shown in Figure 1, the method 100 also includes:

Step 103 , performing quantity detection according to the bull's-eye image captured by each camera, to obtain the quantity of targets in the whole area.

Any possible target statistical method can be used to perform the above quantitative detection, which is not specifically limited in the present application.

，攝像頭202捕獲到的一個目標（行人）圖像，包括

，將局部目標數量累加，由於不同攝像頭之間存在取景範圍的交叉，因此必然存在由不同攝像頭拍攝到了同一目標的不同角度的靶心圖表像的情況，可以通過位置分析判斷攝像頭201捕獲到的

與攝像頭202捕獲到的

重合，從局部目標數量累加結果中刪減重合的數量，由此可以得到全域目標數量為3個。 For example, the number of local targets in each camera can be detected separately, and then the number of multiple local targets can be accumulated to analyze the overlapped bull's-eye images captured by different cameras, and correspondingly deleted. Referring to FIG. 5, three target (pedestrian) images are captured by the camera 201, including

, a target (pedestrian) image captured by the camera 202, including

, add up the number of local targets, because there is an intersection of viewing ranges between different cameras, there must be a situation where different cameras capture the bull’s-eye images of the same target at different angles, and the location analysis can be used to judge the captured by the camera 201

Captured with camera 202

Overlap, the number of overlaps is deleted from the accumulation result of the number of local targets, so that the number of global targets can be obtained as 3.

In some implementations, in order to accurately obtain the number of global targets in the monitoring area, step 103 may also include the following steps: performing position conversion on the captured bull's-eye chart image according to the viewing position of each camera to obtain the target number captured by each camera. The global position corresponding to the bull's-eye chart image; determine the global position coincidence degree of the bull's-eye chart image captured by different cameras, and screen the bull's-eye chart image according to the global position coincidence degree; determine the monitoring area according to the number of bull's-eye chart images retained after screening The number of global targets for .

It can be understood that the bull's-eye image is a partial image containing target features in the current frame, and the global position and Monitors the number of global targets in the area.

，攝像頭202捕獲到的靶心圖表像

，根據攝像頭201的取景位置對捕獲到的靶心圖表像

進行位置轉換，根據攝像頭202的取景位置對捕獲到的靶心圖表像

進行位置轉換，得到圖6示出的每個靶心圖表像的全域位置，可以看出，攝像頭201捕獲到的靶心圖表像

和攝像頭202捕獲到的靶心圖表像

的全域位置重合度很高，假設其超過預設的重合度閾值，即可認為靶心圖表像

和

實際為同一目標，可以僅保留一個，進而可以判斷監控區域中的全域目標數量為3個。 Referring to Fig. 5, the bull's-eye chart image is captured by the camera 201

, the bulls-eye image captured by the camera 202

, according to the viewfinder position of the camera 201, the captured bull's-eye chart image

Carry out position conversion, according to the viewfinder position of camera 202 to the captured bull's-eye chart image

Perform position conversion to obtain the global position of each bull's-eye chart image shown in FIG. 6 . It can be seen that the bull's-eye chart image captured by the camera 201

and the bull's-eye chart image captured by the camera 202

The coincidence degree of the global position of is very high. Assuming that it exceeds the preset coincidence degree threshold, the bull's-eye image can be considered as

and

It is actually the same target, only one can be reserved, and then it can be judged that the number of global targets in the monitoring area is 3.

In some implementations, further, since background occlusion targets may appear in the monitoring area, the number of detected global targets may decrease compared with the actual number. Based on this, the following steps may also be performed: when the number is detected When the result of is less than the number of previous global targets, then according to the multiple current frames collected by multiple cameras and the previous frame of multiple current frames, it is judged whether there is a target that leaves the monitoring area from the predetermined area; if there is no target from the If the target in the predetermined area leaves the monitoring area, the number of previous global targets will still be retained as the number of global targets determined this time; if there are targets that leave the monitoring area from the predetermined area, the result of the number detection will be used as the number of global targets determined this time ;

Wherein, the number of previous global targets is obtained based on target detection and number detection on multiple previous frames. Specifically, by replacing the multiple current frames in steps 101 to 103 with the previous frame of the multiple current frames, the same solution can be used to obtain the number of previous global objects, which will not be repeated in this application.

For example, it is assumed that when real-time target tracking is performed on the last frame of multiple current frames collected by multiple cameras, the number of global targets detected in the monitoring area is 5, that is, 5 target objects are included in total. When performing real-time target tracking on multiple current frames captured by multiple cameras, the result of the quantity detection indicates that there are only 4 target objects in the monitoring area. Compared with the previous frame, the number of objects has decreased, and it is necessary to consider whether there is a temporary Target occlusion. Specifically, the exit area such as the monitoring area can be divided into predetermined areas, and it is judged whether there is a target. According to the last frame of a plurality of current frames, it can be determined that the target is located in the exit area, and according to a plurality of current frames, it can be determined that the target is located in the exit area. The exit area disappears. If there are such targets, it can be considered that the target has actually left the monitoring area, and the result of the above number detection can be used as the number of targets in the whole domain. On the contrary, if there is no such target object, it can be considered that there is a situation such as target occlusion, and the previous number of global targets is still reserved as the number of global targets determined this time.

In some implementations, performing position conversion on the captured bull's-eye chart image according to the viewfinder position of each camera further includes: performing projective transformation on the bottom center point of the bull's-eye chart image in the current frame according to the viewfinder position of each camera, The ground coordinates of each bull's-eye chart image are thereby determined. In this way, the target groups to be recognized captured within the viewing range of each camera can be combined into a unified coordinate system.

For example, the position of the bottom center point of each bull's-eye chart image captured by each camera in Figure 5 can be obtained, and the bottom center point of each bull's-eye chart image can be converted to obtain the actual ground position of the target to be identified in the monitoring scene, Fig. 6 shows the ground coordinates corresponding to each bull's-eye chart image obtained by projective transformation. Specifically, it can be seen that the ground aisle under the perspective of each camera is an approximately trapezoidal area, so for the bull's-eye image captured by each camera, firstly, the bottom center of each bull's-eye image can be obtained through trapezoid-rectangular transformation The coordinates of the point in the standard rectangular area, and then rotate the standard rectangular area according to the actual layout of the monitoring scene, and calculate the rotated coordinates of the bottom center point of each bull's-eye chart image through the calculation of the rotation matrix, and finally according to the actual layout of the monitoring scene. After rotation, the coordinates are translated and scaled to obtain the final ground coordinate position.

Further, in some embodiments, the target number detection model can be pre-trained to be used for real-time detection of the global target number in the monitoring area. When performing the target real-time tracking method, multiple current frames are input into the trained target number Check the model to perform object detection and count detection, and directly get the number of objects in the whole domain.

For example, an improved people counting algorithm YOLOv4-TINY-PC (YOLOv4-tiny-People Counting) can be proposed by improving the real-time target detection algorithm YOLOV4-TINY based on deep learning, in which the YOLOV4-TINY algorithm does not have multi-camera capabilities Capable of counting people, YOLOv4-TINY-PC can get real-time information on the number of people in the monitoring area to count the flow of people. Specifically, the target number detection model can obtain the bull's-eye chart image recognized by each camera through the pedestrian detection algorithm (YOLOv4-TINY-P), and perform position conversion on the bull's-eye chart image to obtain the global position coordinates in the overall monitoring area. Divide the areas of each camera in the computer room, divide the cameras in the computer room into main cameras and auxiliary cameras, and filter the number detection results of each camera so that there is no overlap with each other, and finally get the number of targets in all cameras in the current frame, which is the whole domain target quantity.

In this embodiment, the target number detection model is a pedestrian number detection model created based on the YOLOv4-tiny network. Optionally, a pedestrian number detection model can also be created based on other networks such as faster-rcnn and yolov4. Optionally, target number detection models such as vehicle number detection models and animal number detection models can also be created for other application scenarios.

As shown in Figure 1, the method also includes:

Step 104, perform target re-identification according to the bull's-eye image and the target recognition library.

Wherein, the target identification database includes at least one target identification and characteristic data. For example, the object recognition library may include {object 1: signature data 1, . . . , signature data N}; {object 2: signature data 1, . . . , signature data N}, and so on.

Further, in a possible implementation manner, after step 104, it may also include: calculating the similarity between the bull's-eye chart image and the feature data in the target recognition library, and according to the calculated similarity, the bull's-eye chart image Carry out target re-identification; when the result of target re-identification indicates that the first bull's-eye chart image matches the first target in the target recognition library, mark the identity of the first target on the first bull's-eye chart image.

匹配到目標識別庫中的目標2並進行標識。行人圖像

同樣匹配目標識別庫中的目標1並進行標識。 For example, referring to FIG. 5, the similarity between the shown pedestrian image b and the feature data of each target contained in the target recognition library is calculated, assuming that the similarity between the pedestrian image b and the feature data of target 1 is the highest, And if the similarity exceeds the preset matching threshold, it can be considered that the result of target re-identification indicates that the pedestrian image b matches the target 1 in the target recognition library, and the pedestrian image b can be further identified as target 1. Based on a similar approach, pedestrian images

Match to target 2 in the target recognition library and identify it. Pedestrian image

Also match and identify target 1 in the target recognition library.

As shown in Figure 1, the method also includes:

Step 105. When it is detected that the number of global targets meets the preset increase conditions, determine at least one unrecognized bull's-eye image according to the result of target re-identification, and create a new identity (hereinafter referred to as new ID) for at least one unrecognized bull's-eye image. to mark.

It can be understood that when a new target enters the monitoring area, a new ID needs to be assigned to the new target. The industry usually uses the method of calculating the feature similarity between the bull's-eye image and the feature data in the target recognition library to determine whether to create And assign a new ID, and in some scenarios, the accuracy of the above judgment will be greatly affected due to problems such as target occlusion and shooting angle. For example, when a target already in the monitoring area has poor shooting quality of the bull’s-eye image and the corresponding feature data in the target recognition library cannot be matched, it is easy to mistake it for a new target. Compared with this embodiment, only when the detected number of global targets meets the preset growth condition, for example, when the number of global targets increases compared with the previous number of targets obtained by detecting the number of targets based on the previous frame of multiple current frames, Only then will new IDs be generated, and the creation and distribution of new IDs can be controlled through the number of global targets, which can well ensure the accurate growth of the number of IDs and ensure stability.

匹配到目標2並進行標識。行人圖像

同樣匹配到目標1並進行標識。此時，行人圖像

並未匹配到目標識別庫中的任何一個目標，也即行人圖像

為上述目標重識別過程中確定的未識別靶心圖表像，進一步地，可以根據創建新ID（比如目標3）並對該行人圖像

進行標記。由此可以實現為新的目標分配一個新ID。 Based on this, when the number of detected global targets meets the preset growth condition, at least one unrecognized bull's-eye image is further determined according to the result of target re-identification. For example, referring to Fig. 5, it is assumed that the object recognition library may include {target 1: feature data 1, ..., feature data N}; Image b is matched to target 1 and identified. Pedestrian image

Match to target 2 and identify it. Pedestrian image

Also match to target 1 and identify it. At this time, the pedestrian image

Did not match any target in the target recognition library, that is, the pedestrian image

For the unrecognized bull's-eye image determined in the above target re-identification process, further, a new ID (such as target 3) can be created and the pedestrian image

to mark. This makes it possible to assign a new ID to a new object.

It is worth noting that since the target recognition database is constantly being eliminated and updated, the above-mentioned new targets refer to targets that do not have matching identification and feature information stored in the current target recognition database. In other words, if a pedestrian has entered the monitoring area before and left, it may still be used as a new target when entering the monitoring area next time, and it is necessary to assign a newly created identity to the new target and store the characteristic data accordingly.

In a possible implementation, further, step 105 may also include the step of detecting whether the number of global objects meets the preset increase condition, specifically including: if the current frame is not the first frame, and the number of global objects corresponding to the current frame is compared with When the number of previous global objects corresponding to the previous frame of the plurality of current frames increases, the number of global objects meets a preset increase condition. If the current frame is the first frame, the number of preset global targets meets the preset increase conditions. Specifically, the number of previous global objects has been described above, and will not be repeated here.

As shown in Figure 1, the method also includes:

Step 106, update the target recognition library according to the new identity and the feature data of at least one unrecognized bull's-eye image.

In one embodiment, in order to improve the recognition accuracy of the new ID, step 106 may specifically include: judging whether at least one unrecognized bull's-eye image satisfies the preset image quality condition; Conditional unrecognized bull's-eye chart images are correspondingly stored in the target recognition library.

It can be understood that in the target recognition database, there are few feature data corresponding to the new ID, in order to ensure the accuracy of subsequent target re-identification involving the new ID, stricter quality control is required for the first feature data corresponding to the new ID. For example, at least one unrecognized bullseye image corresponding to a new ID comes from a different camera. Some unrecognized bullseye images may have image quality problems such as small original image size, blurred collection, and environmental occlusion. Whether the unrecognized bull's-eye image corresponding to the ID satisfies the preset image quality conditions, so as to comprehensively judge whether it meets the first characteristic data of the new ID. In this way, incomplete shooting, occlusion, etc. can be filtered out to improve the accuracy of new ID recognition.

In one embodiment, further, after the above step 103, in order to ensure the immediacy of the target recognition library and avoid redundancy, the above method may further include: The feature data of the first target in the target recognition library is dynamically updated. In this way, the feature data with high real-time characteristics can be used for feature matching, which is conducive to improving the recognition accuracy.

It can be understood that when the eigenvalues of the bull's-eye image are used instead of the bull's-eye image for updating, the eigenvalues can be directly used in subsequent calculations, avoiding repeated calculations, greatly reducing the calculation time, and ensuring immediate results.

In one embodiment, in order to avoid feature redundancy in the target recognition library, the method further includes replacing and updating the target recognition library, specifically including the following three replacement and updating scenarios: (1) According to the characteristics of each target in the target recognition library The source time corresponding to the data is compared with the current time, and the target recognition library is replaced and updated. For example, all feature data acquired before the specified time length of the current time may be deleted. It is also possible to delete all characteristic data acquired before another specified time period for one or more targets whose characteristic data amount exceeds a threshold value. In this way, the immediacy of the object recognition library can be guaranteed, which is beneficial to subsequent object re-identification. (2) According to the comparison result of the global position corresponding to the characteristic data of each target in the target recognition database and the current global position of each target, the target recognition database is replaced and updated. For example, it can be understood that the source of the feature data is the previously obtained bull's-eye chart image, so the feature data can correspond to a certain global position according to the bull's-eye chart image from which it originated, and for one or more targets, the distance from the target's current global position can be more than A certain range of feature data is deleted. (3) According to the feature similarity between multiple feature data of each target in the target recognition library, replace and update the target recognition library. For example, for each target in the target recognition library, two or more feature data whose feature similarity is higher than a preset value are deleted, so as to reduce feature duplication in the target recognition library.

In an implementation manner, the method further includes: after the quantity of the feature data of any one target exceeds a preset threshold, starting a replacement update. For example, if the preset threshold is set to 100, in the target recognition library, after the number of feature data of each target exceeds 100, the replacement and update described in the above embodiment will be started to effectively avoid redundancy while ensuring sufficient feature data .

Regarding the method flowchart of the embodiment of the present application, some operations are described as different steps performed in a certain order. Such flowcharts are illustrative and not restrictive. Certain steps described herein can be grouped together and performed in a single operation, can be divided into multiple sub-steps, and can be performed in an order different than that shown herein . Can be implemented in any way by any circuit structure and/or tangible mechanism (for example, by software running on a computer device, hardware (for example, logic functions implemented by a processor or a chip), etc., and/or any combination thereof) The individual steps are shown in the flowchart.

Based on the same technical concept, an embodiment of the present invention further provides an object re-identification device, configured to implement the object re-identification method provided in any of the above-mentioned embodiments. Fig. 7 is a schematic structural diagram of an object re-identification device provided by an embodiment of the present invention.

As shown in Figure 7, the target re-identification device 700 includes:

Obtaining module 701, used to obtain a plurality of current frames collected by a plurality of cameras arranged in the monitoring area;

The target detection module 702 is used for performing target detection according to a plurality of current frames, and determining the bull's-eye chart image captured by each camera;

Quantity detection module 703, is used for carrying out quantity detection according to the bull's-eye chart image captured by each camera, obtains the quantity of global target;

The target re-identification module 704 is used to perform target re-identification according to the bull's-eye image and the target recognition library, and the target recognition library includes at least one target identification and characteristic data;

The identity identification module 705 is configured to determine at least one unrecognized bull's-eye image according to the result of target re-identification when it is detected that the number of targets in the whole domain meets the preset increase condition, and create a new identity to carry out at least one unidentified bull's-eye image. mark;

The target recognition database updating module 706 is configured to update the target recognition database according to the new identity and the feature data of at least one unrecognized bull's-eye image.

In a possible implementation, the target detection module is also used to: input multiple current frames into the trained target detection model to extract the bull's-eye image captured by each camera; wherein, the target detection model is based on A human detection model created by the YOLOv4-tiny network.

In a possible implementation manner, the target detection module is also used to: train the YOLOv4-tiny network according to the real collected images in the monitoring area to obtain a target detection model.

In a possible implementation manner, the bull's-eye image is a partial image containing target features in the current frame, and the quantity detection module is also used to: perform position conversion on the captured bull's-eye image according to the viewing position of each camera to obtain The global position corresponding to the bull's-eye chart image captured by each camera; determine the global position coincidence degree of the bull's-eye chart image captured by different cameras, and filter the bull's-eye chart images captured by different cameras according to the global position coincidence degree, and detect and filter The number of bullseye chart images to keep.

In a possible implementation manner, the number detection module is also used for: when the result of the number detection is less than the number of previous global objects, then according to the multiple current frames collected by multiple cameras and the last frame of the multiple current frames , to determine whether there is a target that leaves the monitoring area from the predetermined area; if there is no target, the number of previous global targets is still retained as the number of global targets determined this time; if there is a target, the result of the number detection is used as the determined The number of global objects; wherein, the number of previous global objects is obtained by performing object detection and number detection on previous frames of multiple current frames.

In a possible implementation, the number detection module is also used for: performing projective transformation on the bottom central point of the bull's-eye image in the current frame according to the viewing position of each camera, so as to determine the ground coordinates of each bull's-eye image.

In a possible implementation, the device is also used to: input multiple current frames into the trained target quantity detection model to perform target detection and quantity detection to obtain the global target quantity; wherein, the target quantity detection model is based on YOLOv4- Pedestrian count detection model created by tiny network.

In a possible implementation, the target re-identification module is also used to: calculate the similarity between the bull's-eye chart image and the feature data in the target recognition library, and perform target re-targeting on the bull's-eye chart image according to the calculated similarity. Recognition: when the result of target re-identification indicates that the first bull's-eye chart image matches the first target in the target recognition library, mark the first bull's-eye chart image according to the identity of the first target.

In a possible implementation, the identification module is also used to: if the current frame is not the first frame, and the number of global objects corresponding to the current frame increases compared with the number of global objects corresponding to the previous frame, then the number of global objects Meet the preset increase conditions; if the current frame is the first frame, the preset number of global targets meets the preset increase conditions.

In a possible implementation, the target recognition library update module is further used to: determine whether at least one unrecognized bull's-eye image satisfies a preset image quality condition; The image of the recognized bull's-eye chart is correspondingly stored in the target recognition library.

In a possible implementation manner, the target recognition library update module is further configured to: dynamically update the characteristic data of the first target in the target recognition library according to the first bull's-eye image or the feature value of the first bull's-eye image.

In a possible implementation, the target recognition library update module is also used to: replace and update the target recognition library according to the comparison result of the source time corresponding to the characteristic data of each target in the target recognition library and the current time; and /or, according to the comparison result of the global position corresponding to the feature data of each target in the target recognition library and the current global position of each target, the target recognition library is replaced and updated; and/or, according to each target in the target recognition library The feature similarity between multiple feature data of a target is used to replace and update the target recognition library.

In a possible implementation manner, the target recognition library update module is further configured to: start a replacement update after the quantity of feature data of any one target exceeds a preset threshold.

It should be noted that the object re-identification device in the embodiment of the present application can realize the various processes of the foregoing object re-identification method embodiment, and achieve the same effect and function, which will not be repeated here.

Fig. 8 is an object re-identification device according to an embodiment of the present application, which is used to execute the object re-identification method shown in Fig. 1 , the device includes: at least one processor; and a memory communicatively connected to at least one processor ; Wherein, the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor, so that at least one processor can execute the methods described in the above-mentioned embodiments.

According to some embodiments of the present application, a non-volatile computer storage medium of a target re-identification method is provided, on which computer-executable instructions are stored, and the computer-executable instructions are configured to be executed when run by a processor: the above-mentioned embodiment the method described.

Each embodiment in the present application is described in a progressive manner, the same and similar parts of each embodiment can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, as for the embodiments of the apparatus, equipment and computer-readable storage medium, since they are basically similar to the method embodiments, the descriptions thereof are simplified, and relevant parts can be referred to part of the description of the method embodiments.

The device, device, and computer-readable storage medium provided in the embodiments of the present application correspond to the method one by one. Therefore, the device, device, and computer-readable storage medium also have beneficial technical effects similar to their corresponding methods. The beneficial technical effect of the method has been described in detail, therefore, the beneficial technical effect of the device, equipment and computer-readable storage medium will not be repeated here.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, systems or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) having computer-usable program code embodied therein .

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each process and/or block in the flowchart and/or block diagram, and a combination of processes and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing equipment to produce a machine so that the instructions executed by the processor of the computer or other programmable data processing equipment Produce means for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, such that the instructions stored in the computer readable memory produce an article of manufacture including the instruction means , the instruction device implements the functions specified in one or more procedures of the flow chart and/or one or more blocks of the block diagram.

These computer program instructions may also be loaded into a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce computer-implemented The instructions executed above provide steps for implementing the functions specified in the procedure or procedures of the flowchart and/or the block or blocks of the block diagram.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

Memory may include non-permanent memory in computer readable media, random access memory (RAM) and/or nonvolatile memory in the form of read only memory (ROM) or flash memory ( flash RAM). The memory is an example of a computer readable medium.

Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology for storage of information. Information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for computers include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM) , read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital multifunction Compact disc (DVD) or other optical storage, magnetic cassette, magnetic tape magnetic disk storage or other magnetic storage device or any other non-transmission medium used to store information that can be accessed by a computing device. In addition, while operations of the methods of the present invention are depicted in the figures in a particular order, there is no requirement or implication that these operations must be performed in that particular order, or that all illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and/or one step may be decomposed into multiple steps for execution.

Although the spirit and principles of the invention have been described with reference to a number of specific embodiments, it should be understood that the invention is not limited to the specific embodiments disclosed, nor does division of aspects imply that features in these aspects cannot be combined to achieve optimal performance. Benefit, this division is only for the convenience of expression. The present invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

100: method 101: Obtain multiple current frames collected by multiple cameras set in the monitoring area 102: Perform target detection based on multiple current frames, and determine the bull's-eye image captured by each camera 103: Quantity detection is performed according to the bull's-eye image captured by each camera, and the number of global targets is obtained 104: Target re-identification based on bull's-eye image and target recognition library 105: When it is detected that the number of global targets meets the preset increase conditions, determine at least one unrecognized bull's-eye image according to the result of target re-identification, and create a new identity (hereinafter referred to as the new ID) for at least one unrecognized bull's-eye image. mark 106: Update the target recognition library according to the new identity and at least one characteristic data of the unrecognized bull's-eye chart image 201, 202: camera 700: Target re-identification device 701: Get module 702: Target detection module 703: Quantity detection module 704: Target re-identification module 705: Identity module 706: Object recognition library update module

The advantages and benefits described herein, as well as other advantages and benefits, will be apparent to those of ordinary skill in the art upon reading the following detailed description of the exemplary embodiments. The drawings are only for the purpose of illustrating exemplary embodiments and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to denote the same parts. In the attached picture:

[Fig. 1] is a schematic flow chart of a target re-identification method according to an embodiment of the present invention; [Fig. 2] is a ground schematic diagram of a monitoring area according to an embodiment of the present invention; [FIG. 3] is a schematic diagram of a viewfinder screen of multiple cameras according to an embodiment of the present invention; [Fig. 4] is a schematic diagram of a current frame of multiple cameras according to an embodiment of the present invention; [Fig. 5] is a schematic diagram of a bull's-eye image captured by multiple cameras according to an embodiment of the present invention; [Fig. 6] is a schematic diagram of the global position of the bull's-eye image captured by multiple cameras according to an embodiment of the present invention; [Fig. 7] is a schematic structural diagram of a target re-identification device according to an embodiment of the present invention; [ Fig. 8 ] is a schematic structural diagram of an object re-identification device according to another embodiment of the present invention.

In the drawings, the same or corresponding reference numerals denote the same or corresponding parts.

100: method

101: Obtain multiple current frames collected by multiple cameras set in the monitoring area

102: Perform target detection based on multiple current frames, and determine the bull's-eye image captured by each camera

103: Quantity detection is performed according to the bull's-eye image captured by each camera, and the number of global targets is obtained

104: Target re-identification based on bull's-eye image and target recognition library

105: When it is detected that the number of global targets meets the preset increase conditions, determine at least one unrecognized bull's-eye image according to the result of target re-identification, and create a new identity (hereinafter referred to as the new ID) to perform at least one unrecognized bull's-eye image. mark

106: Update the target recognition library according to the new identity and at least one characteristic data of the unrecognized bull's-eye chart image

Claims (12)

A target re-identification device, including: an acquisition module for acquiring a plurality of current frames collected by a plurality of cameras arranged in a monitoring area; a target detection module for performing target detection according to the plurality of current frames , to determine the bull’s-eye chart image captured by each camera; the quantity detection module is used to perform quantity detection according to the bull’s-eye chart image captured by each camera, to obtain the number of targets in the whole area, when the result of the quantity detection is less than the First, when the number of targets in the whole area is present, then according to the plurality of current frames collected by the plurality of cameras and the last frame of the plurality of current frames, it is judged whether there is a target leaving the monitoring area from the predetermined area; if not For the target, the number of the previous global target is still retained as the number of the global target determined this time; if the target exists, the result of the number detection is used as the number of the global target determined this time; Wherein, the number of previous global targets is obtained according to the target detection and the number detection of the last frame of the plurality of current frames; the target re-identification module is used to identify the target according to the bull's-eye image and the target The library performs target re-identification, the target identification library includes at least one target identity and characteristic data; the identity identification module is used to re-identify according to the target when it detects that the number of targets in the entire domain meets the preset increase condition As a result of determining at least one unrecognized bull's-eye chart image, creating a new identity mark to mark the at least one unrecognized bull's-eye chart image; the target recognition library update module is used for according to the new identity mark and the at least one The feature data of the unrecognized bull's-eye image is updated to the target recognition library. The target re-identification device according to claim 1, wherein the target detection module is also used for: The multiple current frames are input into the trained target detection model to extract the bull's-eye image captured by each camera; wherein the target detection model is a human body detection model created based on the YOLOv4-tiny network. The target re-identification device according to claim 1, wherein the target detection module is also used to: train the YOLOv4-tiny network according to the real collected images in the monitoring area to obtain the Object detection model. The target re-identification device according to claim 2, wherein the bullseye image is a partial image containing target features in the current frame, and the quantity detection module is also used for: according to the viewing position of each camera Perform position conversion on the captured bull's-eye image to obtain the global position corresponding to the bull's-eye image captured by each camera; determine the global position coincidence of the bull's-eye images captured by different cameras, The bull's-eye chart images captured by different cameras are screened according to the coincidence degree of the global position, and the number of the bull-eye chart images retained after screening is detected. The target re-identification device according to claim 1, wherein the number detection module is further configured to: perform projection transformation on the bottom central point of the bull's-eye chart image in the current frame according to the viewing position of each camera , so as to determine the ground coordinates of each of the bull's-eye chart images. The target re-identification device according to any one of claims 1-5, wherein the device is further configured to: input the multiple current frames into a trained target quantity detection model to perform the target detection and the quantity detection to obtain the quantity of the global target; Wherein, the target number detection model is a pedestrian number detection model created based on the YOLOv4-tiny network. The object re-identification device according to any one of claims 1-5, wherein the object re-identification module is further used to: calculate the distance between the bull's-eye image and the feature data in the object recognition library similarity, and perform target re-identification on the bull's-eye chart image according to the calculated similarity; when the result of the target re-identification indicates that the first bull's-eye chart image matches the first target in the target recognition library, The first bull's-eye image is marked according to the identity of the first target. The target re-identification device according to any one of claim items 1-5, wherein the identity identification module is also used for: if the current frame is not the first frame, and the current frame corresponds to the number of targets in the whole domain When the number of global objects corresponding to the previous frame increases, the number of global objects meets the preset increase condition; if the current frame is the first frame, the preset number of global objects meets the preset Add conditions. The object re-identification device according to any one of claims 1-5, wherein the object recognition library update module is further used to: determine whether the at least one unrecognized bull's-eye image satisfies a preset image quality condition ; Correspondingly storing the new identity mark and the unrecognized bull's-eye image meeting the preset image quality condition into the target recognition library. The target re-identification device according to any one of claim items 1-5, wherein the target recognition library update module is also used for: The feature data of the first target in the target recognition library is dynamically updated according to the first bull's-eye chart image or the feature value of the first bull's-eye chart image. The target re-identification device according to any one of claims 1-5, wherein the target recognition library update module is further configured to: according to the feature data corresponding to each target in the target recognition library The target recognition library is replaced and updated according to the comparison result of the source time and the current time; and/or, according to the global position corresponding to the characteristic data of each target in the target recognition library and the location of each target According to the comparison result of the current global position, replace and update the target recognition library; and/or, according to the feature similarity between multiple feature data of each target in the target recognition library, update the target recognition library Make a replacement update. The object re-identification device according to any one of claims 1-5, wherein the object recognition library update module is further configured to: after the quantity of the feature data of any one of the objects exceeds a preset threshold, Start said replacement update.

Images