Method for determining a tracking object, computer device, storage medium and program product

Abstract

The present application relates to the technical field of image processing, and discloses a method for determining a tracking object, a computer device, a storage medium and a program product, comprising: acquiring a first tracking image of a current frame, a second tracking image of a previous frame, and boundary box information corresponding to a plurality of original tracking objects included in the second tracking image; performing image processing on the first tracking image to acquire boundary box information corresponding to a plurality of present tracking objects included in the first tracking image; when it is determined that there are a plurality of candidate present tracking objects according to the boundary box information corresponding to a target original tracking object and the boundary box information corresponding to each present tracking object, determining a target present tracking object according to the boundary box information of the target original tracking object, the first tracking image, the second tracking image, a pre-constructed image processing model, and the boundary box information corresponding to each candidate present tracking object. The present application can reduce the probability of loss and error of the tracking object.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for determining a tracking object, a computer device, a storage medium, and a program product. Background Technology

[0002] In fields such as video surveillance and autonomous driving, tracking algorithms are typically used on edge devices to track people or vehicles. Edge devices refer to computing devices located near the data source or at the network edge. They can process data close to where it is generated, rather than sending all data to a remote data center for processing. However, edge devices have limited computing resources; therefore, the tracking algorithms used by edge devices are generally Kalman filtering algorithms with low computational complexity.

[0003] During object tracking, the edge device typically begins by detecting the object in the image and generating relevant information about it. This tracking information usually includes the object's category (e.g., person, vehicle, etc.) and bounding box information (e.g., the coordinates of the top-left corner of the bounding box, as well as its length and width). For each detected object, the edge device can use a Kalman filter algorithm to determine its position and identification (ID) in the next frame.

[0004] During the tracking process, tracking objects may be lost or incorrectly identified, meaning that the same tracking object may have multiple IDs. For example, tracking object A may have an ID of "001" and tracking object B may have an ID of "002". During the tracking process, if tracking object A and tracking object B are close to each other, tracking object A's ID may change to "002" and tracking object B's ID may change to "001". Summary of the Invention

[0005] In view of this, the present invention provides a method, apparatus, computer device, storage medium, and program product for determining a tracking object, so as to reduce the probability of tracking object loss or errors.

[0006] In a first aspect, the present invention provides a method for determining a tracking object, the method comprising:

[0007] Acquire the first tracking image of the current frame, the second tracking image of the previous frame corresponding to the current frame, and the bounding box information of the multiple original tracking objects included in the second tracking image;

[0008] The first tracking image is processed to obtain the bounding box information corresponding to the multiple tracking objects included in the first tracking image.

[0009] Based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each of the current tracking objects, it is determined whether there is a candidate current tracking object that matches the original target tracking object, wherein the original target tracking object is any one of the multiple original tracking objects;

[0010] When multiple candidate tracking objects are identified, the target tracking object is determined based on the bounding box information of the original target tracking object, the first tracking image, the second tracking image, the pre-constructed image processing model, and the bounding box information corresponding to each candidate tracking object.

[0011] The method for determining a tracking object provided by this invention has the following advantages:

[0012] This scheme identifies potential candidate current tracking objects by combining the bounding box information of the original target object and the bounding box information of all currently tracked objects. When multiple candidate current tracking objects exist, the scheme further processes other feature information of the tracking objects in the images by combining the image processing model, the second tracking image containing the original target object, the first tracking image containing the current tracking object, and the bounding box information. Through multi-dimensional matching, the true target current tracking object can be identified. Compared to related techniques that only use a Kalman filter to determine the target current tracking object, this scheme significantly reduces the probability of tracking errors and loss.

[0013] In one optional implementation, the bounding box information includes the coordinates of a preset target point on the bounding box, the length of the bounding box, and the width of the bounding box;

[0014] The step of determining whether there is a candidate current tracking object that matches the original target tracking object based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each current tracking object includes:

[0015] The target pixel displacement is determined based on the coordinates of the target point included in the bounding box information corresponding to the original target tracking object and the coordinates of the target point included in the bounding box information corresponding to the first current tracking object, wherein the first current tracking object is any one of the multiple current tracking objects;

[0016] When it is determined that the target pixel displacement is less than or equal to the preset pixel displacement, the first current tracking object is determined as the candidate current tracking object.

[0017] Specifically, if the position of a tracked object does not change much between two adjacent frames, then this tracked object is likely the same target. Therefore, this scheme can determine the candidate current tracked object that may be the original target by calculating the pixel displacement of the original tracked object and the current tracked object between two frames and comparing it with a preset pixel displacement.

[0018] In an optional implementation, when it is determined that there are multiple candidate tracking objects, determining the target tracking object based on the bounding box information of the original target tracking object, the first tracking image, the second tracking image, the pre-constructed image processing model, and the bounding box information corresponding to each candidate tracking object includes:

[0019] Based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the original target tracking object, a sub-image corresponding to the original target tracking object is extracted from the first tracking image;

[0020] Based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the first candidate tracking object, a sub-image corresponding to the first candidate tracking object is extracted from the second tracking image, wherein the first candidate tracking object is any one of the multiple candidate tracking objects;

[0021] The target current tracking object is determined based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each of the candidate current tracking objects, and the image processing model.

[0022] Specifically, by extracting sub-images corresponding to the original target and candidate current target from the first and second tracking images, only these small regions need to be processed instead of the entire image. This significantly reduces computational overhead and improves processing speed, making it particularly suitable for applications in resource-constrained devices, such as edge devices. Furthermore, by combining the sub-images of the original target and candidate current target with a pre-built image processing model, comprehensive analysis and matching are performed. This method leverages the advantages of multi-information fusion, thereby improving matching accuracy.

[0023] In an optional implementation, when the image processing model includes a color model, each of the multiple sub-images includes RGB information corresponding to multiple pixels, wherein the multiple sub-images include a sub-image corresponding to the original target tracking object and a sub-image corresponding to each of the candidate current tracking objects;

[0024] The step of determining the target current tracking object based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each of the candidate current tracking objects, and the image processing model includes:

[0025] Based on the color model and the RGB information corresponding to each pixel in the first sub-image, determine the hue information, saturation information and brightness information of each pixel in the first sub-image, wherein the first sub-image is any one of the multiple sub-images;

[0026] The hue information of each pixel in the first sub-image is statistically analyzed to determine the main hue range corresponding to the first sub-image;

[0027] Based on the dominant color range corresponding to the first sub-image and the color information of each pixel in the first sub-image, a target pixel is selected from the first sub-image, wherein the target pixel includes multiple pixels;

[0028] After the operation of selecting the target pixels for all sub-images is completed, the color similarity between the original target tracking object and each candidate tracking object is determined based on the number, hue, saturation and brightness information of the target pixels in the sub-image corresponding to the original target tracking object, and the number, hue and saturation information of the target pixels in the sub-image corresponding to each candidate tracking object.

[0029] When it is determined that there is only one candidate current tracking object whose color similarity to the original target tracking object is greater than a first preset threshold, the candidate current tracking object is determined as the original target tracking object.

[0030] Specifically, by converting each pixel in the sub-image into hue, saturation, and brightness information, and statistically analyzing the hue information of each pixel in the sub-image to determine the dominant hue range, the target pixel is selected from this range. By focusing on the main color features of the tracked object, the probability of false matches caused by complex backgrounds or partial occlusion is reduced. Furthermore, this scheme not only considers color information but also combines pixel quantity, hue, saturation, and brightness information for a multi-dimensional comprehensive analysis, which improves the reliability of matching.

[0031] In an optional implementation, when it is determined that the color similarity between multiple candidate tracking objects and the original target tracking object is greater than the first preset threshold, the image processing model further includes a contour model, and the method further includes:

[0032] Based on the hue information, saturation information, and brightness information of the target pixels included in the sub-image corresponding to the original target tracking object, and the contour model, the contour feature data corresponding to the original target tracking object is determined;

[0033] In addition, based on the hue information, saturation information and brightness information of the target pixels included in the sub-image corresponding to the second candidate tracking object, and the contour model, contour feature data corresponding to the second candidate tracking object is determined, wherein the second candidate tracking object is any one of a plurality of candidate tracking objects whose color similarity with the original target tracking object is greater than the first preset threshold.

[0034] Based on the contour feature data of the original target object and the contour feature data of the second candidate current target object, the contour similarity between the original target object and the second candidate current target object is determined.

[0035] When it is determined that there is only one candidate current tracking object whose contour similarity to the original target tracking object is greater than a second preset threshold, the candidate current tracking object is determined as the original target tracking object.

[0036] Specifically, when multiple candidate tracking objects have a color similarity greater than a first preset threshold with the original target tracking object, the target tracking object is further screened using a contour model, which is more accurate.

[0037] In an optional implementation, when it is determined that the contour similarity between multiple candidate current tracking objects and the original target tracking object is greater than the second preset threshold, the method further includes:

[0038] The candidate current tracking object with the highest contour similarity to the original target tracking object is selected as the target current tracking object.

[0039] Specifically, the object with the highest contour similarity is selected to ensure that the final selected target object being tracked is closest to the original target object in shape and structure.

[0040] In an optional implementation, the method further includes:

[0041] Obtain the identification information of the original target object and the identification information of the current target object;

[0042] The identification information of the original target being tracked is compared with the identification information of the current target being tracked.

[0043] When it is determined that the identification information of the original target being tracked is different from the identification information of the current target being tracked, the identification information of the current target being tracked is updated to the identification information of the original target being tracked.

[0044] Specifically, since the tracking algorithm assigns identification information to each tracked object, this solution can avoid wasting identification information resources by correcting the tracking information. Furthermore, it can also ensure the accuracy of the tracking information.

[0045] In a second aspect, the present invention provides an apparatus for determining a tracking object, the apparatus comprising:

[0046] The acquisition module is used to acquire a first tracking image of the current frame, a second tracking image of the previous frame corresponding to the current frame, and bounding box information corresponding to multiple original tracking objects included in the second tracking image; and to perform image processing on the first tracking image to acquire bounding box information corresponding to multiple current tracking objects included in the first tracking image.

[0047] The determination module is used to determine whether there is a candidate current tracking object that matches the original target tracking object, based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each of the current tracking objects. The original target tracking object is any one of the multiple original tracking objects. When it is determined that there are multiple candidate current tracking objects, the target current tracking object is determined based on the bounding box information of the original target tracking object, the first tracking image, the second tracking image, the pre-constructed image processing model, and the bounding box information corresponding to each of the candidate current tracking objects.

[0048] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method for determining a tracking object as described in the first aspect or any corresponding embodiment thereof.

[0049] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method for determining a tracking object as described in the first aspect or any corresponding embodiment thereof.

[0050] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the method for determining a tracking object as described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the specific embodiments or related technologies of the present invention, the drawings used in the description of the specific embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating a method for determining a tracking object according to an embodiment of the present invention;

[0053] Figure 2 This is a schematic diagram of the process of filtering candidate tracking objects using a color model according to an embodiment of the present invention;

[0054] Figure 3 This is a schematic diagram of the process of filtering candidate tracking objects using color models and contour models according to an embodiment of the present invention;

[0055] Figure 4 This is a schematic diagram of the process of filtering candidate tracking objects using a contour model according to an embodiment of the present invention;

[0056] Figure 5 This is a flowchart illustrating a specific example of determining a tracking object according to an embodiment of the present invention;

[0057] Figure 6 This is a structural block diagram of a device for determining a tracking object according to an embodiment of the present invention;

[0058] Figure 7 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] This invention can be implemented using computer devices, which can be edge devices. Edge devices are devices operating at the network edge, close to the source of data generation, capable of local data processing, storage, and analysis. Examples include mobile phones, computers, and in-vehicle terminals. By performing tasks at the network edge layer, edge devices reduce the need for data transmission to data centers, thereby improving data processing response speed and efficiency. When implementing tracking functionality on edge devices, the selection of artificial intelligence (AI) algorithms is particularly important due to the lower computing power of edge devices.

[0061] The tracking function is implemented based on the use of various image processing models and tracking algorithms. These image processing models can include object recognition algorithms, color models, contour models, and so on. For example, the object recognition algorithm could be the YOLOv5 (You Only Look Once version 5) algorithm, and the color model could be the Hue Saturation Value (HSV) model. The tracking algorithm could be the Kalman filtering algorithm.

[0062] This invention provides a method for determining a tracking object. By using an image processing model, the accuracy of the tracking object determined by the tracking algorithm can be confirmed, thereby avoiding the problem of incorrect tracking information.

[0063] According to an embodiment of the present invention, a method embodiment for determining a tracking object is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0064] This embodiment provides a method for determining the tracking object, which can be executed by the aforementioned edge device. Figure 1 This is a flowchart of a method for determining a tracking object according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0065] Step S101: Obtain the first tracking image of the current frame, the second tracking image of the previous frame corresponding to the current frame, and the bounding box information of the multiple original tracking objects included in the second tracking image.

[0066] The bounding box information may include the coordinates of a preset target point on the bounding box, the length and width of the bounding box, etc. For example, when the bounding box is rectangular, the preset target point may be the top-left corner of the bounding box.

[0067] Step S102: Perform image processing on the first tracking image to obtain the bounding box information corresponding to the multiple tracking objects included in the first tracking image.

[0068] Specifically, after receiving the video stream transmitted by the image acquisition device, the edge device can decode the video stream to obtain the image. For each frame of the image, the edge device can perform the following processing:

[0069] First, the identification and processing of the tracked object.

[0070] Edge devices can use the YOLOv5 algorithm to identify at least one object and its category from an image. By comparing each object's category with a pre-defined list of categories, at least one trackable object belonging to that list is determined; for example, the tracked object could be a person, a vehicle, etc. For each tracked object, the YOLOv5 algorithm can output the corresponding bounding box information.

[0071] For each tracked object, the edge device can determine the coordinates of the center point within the bounding box based on the coordinates of the preset target point and the length and width of the bounding box.

[0072] Second, predicting the tracking location.

[0073] For each tracked object, the edge device can construct a state variable corresponding to that object based on the coordinates of the center point, the preset rates of change of the horizontal and vertical coordinates, and the length and width of the bounding box. The mathematical expression for the state variable can be as follows:

[0074]

[0075] in, Here, x is the x-coordinate of the center point, y is the y-coordinate of the center point, Δx is the rate of change of the x-coordinate, Δy is the rate of change of the y-coordinate, w is the width of the bounding box, and h is the length of the bounding box.

[0076] Then, the edge device can determine the target state of the tracked object in the next frame based on the state quantity corresponding to the tracked object, the preset state transition matrix, the measurement matrix, the preset process noise, the preset measurement error, and the first and second weight values ​​automatically generated by the tracking algorithm. The specific steps are as follows:

[0077] Step 1: The edge device can determine the first state variable of the tracked object based on the state transition matrix, the state variable corresponding to the tracked object, and the preset process noise.

[0078] For example, step one can be expressed as follows:

[0079]

[0080] in, To predict state variables, F is the state transition matrix, and w is the preset process noise. For example, the mathematical expression for the state transition matrix can be as follows:

[0081]

[0082] Where Δt represents the interval between two adjacent frames.

[0083] Step two: The edge device can determine the second state quantity of the tracked object based on the measurement matrix, the state quantity corresponding to the tracked object, and the preset measurement error.

[0084] For example, step two can be expressed as follows:

[0085]

[0086] Among them, z k Let H be the second state variable, H be the measurement matrix, and v be the preset measurement error. For example, the mathematical expression for the measurement matrix can be as follows:

[0087]

[0088] Step 3: The edge device can determine the target state variable based on the first weight value, the second weight value, the first state variable, and the second state variable.

[0089] For example, step three can be expressed as follows:

[0090]

[0091] Where, x k+1 Let A be the target state variable, and B be the first weight value and B be the second weight value.

[0092] Third, the identification of the tracking targets.

[0093] Because the edge device identifies the tracked object in each frame during tracking, the displacement of the tracked object between adjacent frames is relatively small. Thus, in each frame, the edge device can determine which tracked object A in the previous frame corresponds to the tracked object A in the current frame (e.g., the tracked object with the highest similarity (greater than a preset threshold)) based on the first state value of the tracked object A constructed in the current frame and the target state values ​​of all tracked objects calculated in the previous frame. The identification information of the corresponding tracked object from the previous frame is then assigned to tracked object A. If no corresponding tracked object exists in the previous frame, tracked object A can be identified as a new tracked object in the current frame, and a new identification information can be assigned to tracked object A. In this way, the edge device can generate tracking information for the tracked object in the current frame based on the target state value and identification information of the tracked object.

[0094] Thus, during the processing of the current frame, the edge device can first acquire the first tracking image of the current frame, the second tracking image of the previous frame corresponding to the current frame, and the bounding box information and identification information corresponding to the multiple original tracking objects included in the second tracking image (obtained in the processing of the previous frame using the same processing method described above). The edge device can then perform image processing on the first tracking image according to the above process to obtain the bounding box information and identification information corresponding to the multiple original tracking objects included in the first tracking image.

[0095] In the process of determining the corresponding tracking object of tracking object A in the previous frame, if tracking object A and tracking object B are two objects moving side by side, there may be a situation where the tracking object is incorrectly identified, leading to errors in the tracking information. Therefore, to avoid the above problem, further matching can be performed in subsequent steps.

[0096] Step S103: Based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each current tracking object, determine whether there is a candidate current tracking object that matches the original target tracking object.

[0097] The target original tracking object is any one of multiple original tracking objects.

[0098] Specifically, for any tracked object identified in the previous frame, the edge device can determine whether there is a candidate tracked object in the processing of the current frame in the following specific manner:

[0099] Step 1: Determine the target pixel displacement based on the coordinates of the target points included in the bounding box information corresponding to the original target object and the target points included in the bounding box information corresponding to the first current target object.

[0100] The first currently tracked object is any one of the multiple currently tracked objects.

[0101] For example, step one can be expressed as follows:

[0102]

[0103] in, x is the target pixel displacement. f The x-coordinate of the preset target point included in the bounding box information corresponding to the first tracked object is given by the x-coordinate. i The x-coordinate of the preset target point included in the bounding box information corresponding to the original tracked object, y f The y-coordinate of the preset target point included in the bounding box information corresponding to the first tracked object. i The ordinate of the preset target point is included in the bounding box information corresponding to the original tracked object.

[0104] Step 2: When the target pixel displacement is determined to be less than or equal to the preset pixel displacement, the first current tracking object is determined as the candidate current tracking object.

[0105] For example, the preset pixel displacement can be 20 pixels.

[0106] Specifically, when the target pixel displacement is determined to be less than or equal to a preset pixel displacement, it indicates that the first currently tracked object may be the original target tracking object, and the first currently tracked object can be identified as a candidate currently tracked object. When the target pixel displacement is determined to be greater than the preset pixel displacement, it indicates that the first currently tracked object is not the original target tracking object. In this case, the first currently tracked object can be filtered out and not participate in subsequent processing. In this way, it is possible to determine whether each currently tracked object is a candidate tracking object that matches the original target tracking object.

[0107] If only one candidate tracking object is ultimately identified, then that candidate tracking object can be designated as the target tracking object. If no candidate tracking object exists, no further processing is required. If multiple candidate tracking objects are ultimately identified, then step S104 can be performed to determine which tracking object is the target tracking object.

[0108] Step S104: When it is determined that there are multiple candidate tracking objects, the target tracking object is determined based on the bounding box information of the original target tracking object, the first tracking image, the second tracking image, the pre-constructed image processing model, and the bounding box information corresponding to each candidate tracking object.

[0109] The image processing model may include a color model and / or a contour model. The color model may be an HSV model.

[0110] Specifically, the edge device can perform sub-image cropping processing on the first tracking image and the second tracking image respectively. That is, it can first crop a sub-image corresponding to the original target object from the first tracking image based on the coordinates of a preset target point, the length and width of the bounding box included in the bounding box information corresponding to the original target object; and then, it can crop a sub-image corresponding to the first candidate current tracking object from the second tracking image based on the coordinates of a preset target point, the length and width of the bounding box included in the bounding box information corresponding to the first candidate current tracking object. Here, the first candidate current tracking object is any one of a plurality of candidate current tracking objects.

[0111] In this way, the edge device can obtain multiple sub-images, including the sub-image corresponding to the original target object and the sub-image corresponding to each candidate target object. Both the first and second tracking images mentioned above can be in Red Green Blue (RGB) mode, meaning that each tracking image can include RGB information corresponding to multiple pixels. Correspondingly, each of the multiple sub-images extracted above can include RGB information corresponding to multiple pixels.

[0112] Then, the edge device can determine the target current tracking object corresponding to the target original tracking object based on the sub-image corresponding to the target original tracking object, the sub-image corresponding to each candidate current tracking object, and the image processing model.

[0113] Depending on the image processing model, edge devices can perform different processing:

[0114] Scenario 1: The image processing model only includes a color model. For example... Figure 2 As shown, the processing flow for filtering using a color model can be as follows:

[0115] Step 1: Based on the color model and the RGB information corresponding to each pixel in the first sub-image, determine the hue information, saturation information, and brightness information of each pixel in the first sub-image.

[0116] The first sub-image is any one of the multiple sub-images.

[0117] Step two: Statistically analyze the hue information of each pixel in the first sub-image to determine the main hue range corresponding to the first sub-image.

[0118] Step 3: Select the target pixel from the first sub-image based on the main color range corresponding to the first sub-image and the color information of each pixel in the first sub-image.

[0119] The target pixels include multiple pixels.

[0120] Step 4: After completing the operation of selecting target pixels for all sub-images, determine the color similarity between the original target object and each candidate target object based on the number, hue, saturation, and brightness information of target pixels in the sub-image corresponding to the original target object, as well as the number, hue, saturation, and brightness information of target pixels in the sub-image corresponding to each candidate target object.

[0121] Step 5: When it is determined that there is only one candidate current tracking object whose color similarity to the original target tracking object is greater than the first preset threshold, the candidate current tracking object is determined as the target current tracking object.

[0122] Step 6: When it is determined that there are multiple candidate current tracking objects whose color similarity with the original target tracking object is greater than the first preset threshold, select the candidate current tracking object with the largest color similarity with the original target tracking object as the target current tracking object.

[0123] Specifically, when the color model is HSV, the edge device can input the first sub-image (i.e., the RGB information of all pixels in the first sub-image) into the HSV model to obtain a sub-image converted to HSV mode. The HSV mode sub-image can include HSV information corresponding to multiple pixels, namely the hue information, saturation information, and brightness information mentioned above. Hue information can be represented by a color range. For example, [0, 180], this color range can be divided into 36 color sub-ranges, each with a size of 5. Based on the hue information, the edge device can calculate a histogram of the HSV-H channels for the first sub-image. The horizontal axis of the histogram can be the color sub-range, and the vertical axis can be the number of pixels in the corresponding color sub-range. Thus, for the first sub-image, the edge device can sort the color sub-ranges from most to least least according to the number of pixels in each color sub-range, and determine at least one color sub-range that ranks before a preset sorting value. Then, the edge device can construct a main hue range based on each color sub-range.

[0124] Based on the above processing method, the edge device can determine the dominant color range corresponding to each sub-image, as well as the HSV information corresponding to all pixels in each sub-image.

[0125] The following example, comparing the original tracked object with the first candidate tracked object, illustrates the process of calculating color similarity in detail:

[0126] First, the edge device selects target pixels within the first dominant color range of sub-image A based on the first dominant color range corresponding to sub-image A and the color information corresponding to each pixel in sub-image A. Then, based on the second dominant color range of sub-image B and the color information corresponding to each pixel in sub-image B, it selects target pixels within the second dominant color range of sub-image B. Here, sub-image A is the sub-image corresponding to the original target being tracked, and sub-image B is the sub-image corresponding to the first candidate target being tracked.

[0127] Then, the edge device can count the number of target pixels in sub-image A and the number of target pixels in sub-image B. Further, it compares which sub-image has fewer target pixels and sets the fewer pixels as the upper limit for the target pixel count. Furthermore, the edge device can determine the position of each target pixel within its respective sub-image; the position can be the row and column number of the target pixel in the sub-image. Based on the position of the target pixels within their respective sub-images, the device then determines the sorting of the target pixels within those sub-images.

[0128] Finally, the edge device can use the following mathematical expression to calculate the color similarity between the original tracked object and the first candidate tracked object:

[0129]

[0130] Where d(A,B) represents the color similarity between the original target object and the first candidate current target object. V represents brightness information, S represents saturation information, H represents hue information, k represents the ranking of the target pixel in its sub-image, and n represents the number of targets. 1k Let x be the target pixel sorted k in subimage A. 2k Let x, y be the target pixel sorted k in subimage B. 1k Let y be the target pixel sorted k in subimage A. 2k Let y and z be the target pixels sorted k in subimage B. 1k Let z be the target pixel sorted k in subimage A. 2k Let z be the target pixel ranked k in subimage B. Here, x, y, and z can be expressed as follows:

[0131]

[0132] Thus, for each candidate tracking object, its color similarity to the original target tracking object can be determined using the above method. If no candidate tracking object has a color similarity greater than the first preset threshold with the original target tracking object, no processing is required. If only one candidate tracking object has a color similarity greater than the first preset threshold with the original target tracking object, that candidate tracking object can be directly identified as the original target tracking object. If multiple candidate tracking objects have a color similarity greater than the first preset threshold with the original target tracking object, the candidate tracking object with the highest color similarity can be selected as the original target tracking object.

[0133] In some optional implementations, the edge device can first determine the overlap between the dominant color range of the sub-image corresponding to the original target and the dominant color range of the sub-image corresponding to the first candidate current target, and determine whether the overlap is greater than a preset overlap threshold. If so, color similarity can be further calculated; if not, it can be determined that there is no target current target corresponding to the original target. Since the dominant color range can reflect the color similarity between two images to a certain extent, the consistency of the dominant color range can first filter out some candidate current targets, avoiding the calculation of color similarity and saving resources.

[0134] Scenario 2: The image processing model includes a color model and a contour model.

[0135] like Figure 3 As shown, after completing step four in Case 1, if the edge device determines that the color similarity between multiple candidate current tracking objects and the original target tracking object is greater than the first preset threshold, it can perform contour matching processing on the candidate current tracking objects selected in step four of Case 1 based on the contour model. For example... Figure 4 As shown, the processing flow for filtering using a contour model can be as follows:

[0136] Step 1: Based on the hue, saturation, and brightness information of the target pixels in the sub-image corresponding to the original target object, as well as the contour model, determine the contour feature data corresponding to the original target object.

[0137] Step 2: Based on the hue information, saturation information, and brightness information of the target pixels in the sub-image corresponding to the second candidate tracking object, as well as the contour model, determine the contour feature data corresponding to the second candidate tracking object.

[0138] The second candidate current tracking object is any one of a number of candidate current tracking objects whose color similarity with the original target tracking object is greater than a first preset threshold.

[0139] Step 3: Determine the contour similarity between the original target object and the second candidate target object based on the contour feature data of the original target object and the contour feature data of the second candidate target object.

[0140] Step 4: When it is determined that there is only one candidate current tracking object whose contour similarity with the original target tracking object is greater than the second preset threshold, the candidate current tracking object is determined as the target current tracking object.

[0141] Step 5: When multiple candidate tracking objects have contour similarities with the original target tracking object that are all greater than the second preset threshold, select the candidate tracking object with the largest contour similarity with the original target tracking object as the target tracking object.

[0142] Specifically, the edge device can convert the sub-image of the original target tracking object and the sub-image of the candidate current tracking objects (candidate current tracking objects whose color similarity with the original target tracking object is greater than a first preset threshold) from HSV mode back to RGB mode. During the conversion process, the pixels in the sub-image within the main color range can be converted using the following formulas (10) to (15). The RGB information of the remaining pixels can be set to preset information, such as (0, 0, 0). In this way, a sub-image of the original target tracking object and each candidate current tracking object after color screening in RGB mode can be obtained.

[0143] The data expression corresponding to RGB information can be as follows:

[0144]

[0145] p = V × (1 - S) (13)

[0146] q=V× (1-f×S) (14)

[0147] t=V× (1-(1-f)×S) (15)

[0148] Furthermore, the edge device can convert each sub-image converted to RGB mode into a grayscale image and perform binarization to obtain the corresponding target sub-image. For each target sub-image, a contour model can be used to extract the corresponding contour feature data (specifically, it can be extracted using a multinomial fitting method). The contour feature data can include the coordinates of each pixel containing the contour, i.e., the contour feature data is a dataset. The edge device can calculate the distance between two datasets according to the Hausdorff distance formula.

[0149] The Hausdorff distance formula can be expressed as follows:

[0150] d H (A, B) = max(sup) a∈A inf b∈B ||ab||,sup b∈B inf a∈A ||ab||) (16)

[0151] Where, d H (A, B) represents the contour similarity between the original target object and the second candidate target object, where a is a point in dataset A, b is a point in dataset B, sup represents the supremum (maximum value in the dataset), and inf represents the infimum (minimum value in the dataset). ||ab|| represents the distance between points a and b, which can be expressed using Euclidean distance.

[0152] Scenario 3: The image processing model only includes a contour model.

[0153] The edge device can extract contour feature data from the sub-images of the original target tracking object and each candidate current tracking object using a contour model. Then, using the Hausdorff distance formula and the contour feature data of the original target tracking object and the first candidate current tracking object, it can calculate the contour similarity between the original target tracking object and the first candidate current tracking object. It then determines whether the contour similarity between the original target tracking object and the first candidate current tracking object is greater than a second preset threshold. When it is determined that only one candidate current tracking object has a contour similarity greater than the second preset threshold with the original target tracking object, that candidate current tracking object is identified as the target current tracking object. When it is determined that multiple candidate current tracking objects have contour similarities greater than the second preset threshold with the original target tracking object, the edge device can select the candidate current tracking object with the highest contour similarity to the original target tracking object as the target current tracking object. Alternatively, it can further calculate the color similarity between the original target tracking object and the multiple candidate current tracking objects selected by the contour model. When it is determined that only one candidate current tracking object has a color similarity greater than the first preset threshold with the original target tracking object, the edge device can identify that candidate current tracking object as the target current tracking object. When it is determined that there are multiple candidate current tracking objects whose color similarity with the original target tracking object is greater than a first preset threshold, the edge device can select the candidate current tracking object with the largest color similarity with the original target tracking object as the target current tracking object.

[0154] In some optional implementations, the edge device can compare the identification information of the original target being tracked with the identification information of the currently tracked target. When it is determined that the identification information of the original target being tracked differs from that of the currently tracked target, the identification information of the currently tracked target is updated to that of the original target being tracked.

[0155] Specifically, when the edge device determines that the identification information of the original target being tracked is the same as the identification information of the current target being tracked, it indicates that the tracking information provided by the tracking algorithm is correct and does not need to be modified. When the identification information of the original target being tracked by the edge device is different from the identification information of the current target being tracked, it indicates that the tracking information provided by the tracking algorithm is incorrect, and the identification information included in the tracking information of the current target being tracked can be modified to the identification information of the original target being tracked.

[0156] In some optional implementations, during the determination of the tracking object in step 102, the tracking object identification information may not be assigned initially. Instead, the edge device can directly add the identification information of the original tracking object to the tracking information of the current tracking object. For a current tracking object that does not match any original tracking object, the edge device can determine it as a newly identified tracking object and assign it new identification information.

[0157] In some optional implementations, for each assigned identifier, the edge device can determine whether the identifier appears in the current frame. If not, it increments the value of the counter parameter corresponding to the identifier and determines whether the value of the counter parameter is greater than a third preset threshold. If it is greater than the third preset threshold, it can be considered that the tracking object corresponding to the identifier has disappeared and will not appear again. Therefore, tracking of the tracking object corresponding to the identifier can be canceled, and the tracking information of the tracking object can be deleted, thus reclaiming the identifier. This saves storage resources of the edge device and resources in the identifier resource pool. Furthermore, it reduces the number of original tracking objects that need to be traversed subsequently. Since the edge device needs to perform a large number of matching calculations for each original tracking object, timely deletion of related tracking objects can reduce the computational load.

[0158] like Figure 5 As shown in the diagram, a specific example diagram for determining the tracking object is provided. The edge device can traverse each of the multiple original tracking objects identified in the previous frame. Each time an original tracking object is encountered, the device can determine whether a corresponding target tracking object exists among the multiple current tracking objects identified in the current frame, following the method described above. This process continues until all original tracking objects in the previous frame have been traversed. Figure 5 The specific processing of each step has been described in detail above, and will not be repeated here.

[0159] The method for determining the tracking object provided in this embodiment firstly selects multiple candidate tracking objects that may be mistracked by using the bounding box information of the original tracking object and each current tracking object after determining the tracking information. Then, it further filters out the accurate target tracking object by calculating the color similarity and contour similarity between the original tracking object and the candidate tracking objects. This reduces the probability of tracking object loss or errors. Furthermore, this method only performs the next layer of filtering using an image processing model when multiple candidate tracking objects are identified. This maximizes the conservation of computing resources and is suitable for edge devices with low computing power. Additionally, in the process of calculating color similarity, only the HSV information of target pixels within the main color range is considered, which reduces the computational load.

[0160] This embodiment also provides a device for determining a tracking object, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0161] This embodiment provides a device for determining a tracking object, such as... Figure 6 As shown, it includes:

[0162] The acquisition module 601 is used to acquire the first tracking image of the current frame, the second tracking image of the previous frame corresponding to the current frame, and the bounding box information of the multiple original tracking objects included in the second tracking image respectively; and to perform image processing on the first tracking image to acquire the bounding box information of the multiple original tracking objects included in the first tracking image respectively.

[0163] The determination module 602 is used to determine whether there is a candidate current tracking object that matches the target original tracking object based on the bounding box information corresponding to the target original tracking object and the bounding box information corresponding to each current tracking object, wherein the target original tracking object is any one of multiple original tracking objects; when it is determined that there are multiple candidate current tracking objects, the target current tracking object is determined based on the bounding box information of the target original tracking object, the first tracking image, the second tracking image, the pre-constructed image processing model, and the bounding box information corresponding to each candidate current tracking object.

[0164] In one optional implementation, the bounding box information includes the coordinates of a preset target point on the bounding box, the length of the bounding box, and the width of the bounding box;

[0165] Module 602 is specifically used for:

[0166] The target pixel displacement is determined based on the coordinates of the target point included in the bounding box information corresponding to the original target tracking object and the coordinates of the target point included in the bounding box information corresponding to the first current tracking object, wherein the first current tracking object is any one of multiple current tracking objects;

[0167] When the target pixel displacement is determined to be less than or equal to the preset pixel displacement, the first current tracking object is determined as the candidate current tracking object.

[0168] In one alternative implementation, the determining module 602 is specifically used for:

[0169] Based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the original target object, a sub-image corresponding to the original target object is extracted from the first tracking image;

[0170] Based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the first candidate tracking object, a sub-image corresponding to the first candidate tracking object is extracted from the second tracking image. The first candidate tracking object is any one of multiple candidate tracking objects.

[0171] The target current tracking object is determined based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each candidate current tracking object, and the image processing model.

[0172] In an optional implementation, when the image processing model includes a color model, each of the multiple sub-images includes red, green, and blue RGB information corresponding to multiple pixels respectively, wherein the multiple sub-images include a sub-image corresponding to the original target tracking object and a sub-image corresponding to each candidate current tracking object.

[0173] Module 602 is specifically used for:

[0174] Based on the color model and the RGB information corresponding to each pixel in the first sub-image, determine the hue information, saturation information and brightness information of each pixel in the first sub-image, wherein the first sub-image is any one of multiple sub-images;

[0175] The tonal information of each pixel in the first sub-image is statistically analyzed to determine the main color range corresponding to the first sub-image;

[0176] Based on the dominant color range corresponding to the first sub-image and the color information of each pixel in the first sub-image, a target pixel is selected from the first sub-image, wherein the target pixel includes multiple pixels;

[0177] After completing the operation of selecting target pixels for all sub-images, the color similarity between the original target tracking object and each candidate current tracking object is determined based on the number of target pixels, hue information, saturation information and brightness information in the sub-image corresponding to the original target tracking object, and the number of target pixels, hue information, saturation information and brightness information in the sub-image corresponding to each candidate current tracking object.

[0178] When it is determined that there is only one candidate current tracking object whose color similarity to the original target tracking object is greater than the first preset threshold, the candidate current tracking object is determined as the target current tracking object.

[0179] In an optional implementation, when it is determined that the color similarity between multiple candidate current tracking objects and the original target tracking object is greater than a first preset threshold, the image processing model further includes a contour model. The determination module 602 is specifically used for:

[0180] Based on the hue, saturation, and brightness information of the target pixels in the sub-image corresponding to the original target object, as well as the contour model, determine the contour feature data corresponding to the original target object.

[0181] In addition, based on the hue information, saturation information and brightness information of the target pixels included in the sub-image corresponding to the second candidate current tracking object, and the contour model, the contour feature data corresponding to the second candidate current tracking object is determined, wherein the second candidate current tracking object is any one of a number of candidate current tracking objects whose color similarity with the original target tracking object is greater than a first preset threshold.

[0182] Based on the contour feature data of the original target object and the contour feature data of the second candidate target object, determine the contour similarity between the original target object and the second candidate target object.

[0183] When it is determined that there is only one candidate current tracking object whose contour similarity with the original target tracking object is greater than the second preset threshold, the candidate current tracking object is determined as the target current tracking object.

[0184] In one alternative implementation, the determining module 602 is specifically used for:

[0185] When it is determined that there are multiple candidate current tracking objects whose contour similarity with the original target tracking object is greater than the second preset threshold, the candidate current tracking object with the largest contour similarity with the original target tracking object is selected as the original target tracking object.

[0186] In one alternative implementation, the device further includes an update module:

[0187] Obtain the identification information of the original target object and the current target object being tracked;

[0188] Compare the identification information of the original target object with the identification information of the current target object;

[0189] When it is determined that the identification information of the original target being tracked is different from the identification information of the current target being tracked, the identification information of the current target being tracked will be updated to the identification information of the original target being tracked.

[0190] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0191] In this embodiment, the device for determining the tracking object is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0192] This invention also provides a computer device having the above-described features. Figure 6 The apparatus shown is for determining the object to be tracked.

[0193] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 7 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 7 Take a processor 10 as an example.

[0194] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware integrated circuit. The aforementioned hardware integrated circuit may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0195] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0196] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0197] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0198] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0199] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0200] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0201] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for determining a tracking object, characterized in that, The method is applied to an edge device, and the method includes: Acquire the first tracking image of the current frame, the second tracking image of the previous frame corresponding to the current frame, and the bounding box information corresponding to the multiple original tracking objects included in the second tracking image, wherein the bounding box information includes the coordinates of a preset target point on the bounding box, the length and width of the bounding box; The first tracking image is processed to obtain the bounding box information corresponding to the multiple tracking objects included in the first tracking image. Based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each of the current tracking objects, it is determined whether there is a candidate current tracking object that matches the original target tracking object, wherein the original target tracking object is any one of the multiple original tracking objects; When it is determined that there are multiple candidate tracking objects, a sub-image corresponding to the original tracking object is extracted from the first tracking image based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the original tracking object. Based on the coordinates of the preset target point, the length and width of the bounding box included in the bounding box information corresponding to the first candidate tracking object, a sub-image corresponding to the first candidate tracking object is extracted from the second tracking image, wherein the first candidate tracking object is any one of the plurality of candidate tracking objects; The target current tracking object is determined based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each candidate current tracking object, and the pre-constructed image processing model, wherein the image processing model includes a color model and / or a contour model. Obtain the counting parameter corresponding to the assigned identification information. When it is determined that the value of the counting parameter is greater than a third preset threshold, delete the tracking information of the tracking object corresponding to the identification information and reclaim the identification information. The counting parameter is used to indicate the number of consecutive frames in which the tracking object corresponding to the identification information disappears.

2. The method according to claim 1, characterized in that, The step of determining whether there is a candidate current tracking object that matches the original target tracking object based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each current tracking object includes: The target pixel displacement is determined based on the coordinates of the preset target point included in the bounding box information corresponding to the original target tracking object and the coordinates of the preset target point included in the bounding box information corresponding to the first current tracking object, wherein the first current tracking object is any one of the multiple current tracking objects; When it is determined that the target pixel displacement is less than or equal to the preset pixel displacement, the first current tracking object is determined as the candidate current tracking object.

3. The method according to claim 1, characterized in that, Each of the multiple sub-images includes RGB information corresponding to multiple pixels, wherein the multiple sub-images include a sub-image corresponding to the original target tracking object and a sub-image corresponding to each of the candidate current tracking objects; The step of determining the target current tracking object based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each of the candidate current tracking objects, and the image processing model includes: Based on the color model and the RGB information corresponding to each pixel in the first sub-image, determine the hue information, saturation information and brightness information of each pixel in the first sub-image, wherein the first sub-image is any one of the multiple sub-images; The hue information of each pixel in the first sub-image is statistically analyzed to determine the main hue range corresponding to the first sub-image; Based on the dominant color range corresponding to the first sub-image and the color information of each pixel in the first sub-image, a target pixel is selected from the first sub-image, wherein the target pixel includes multiple pixels; After the operation of selecting the target pixels for all sub-images is completed, the color similarity between the original target tracking object and each candidate tracking object is determined based on the number, hue, saturation and brightness information of the target pixels in the sub-image corresponding to the original target tracking object, and the number, hue and saturation information of the target pixels in the sub-image corresponding to each candidate tracking object. When it is determined that there is only one candidate current tracking object whose color similarity to the original target tracking object is greater than a first preset threshold, the candidate current tracking object is determined as the original target tracking object.

4. The method according to claim 3, characterized in that, When it is determined that the color similarity between multiple candidate tracking objects and the original target tracking object is greater than the first preset threshold, the method further includes: Based on the hue information, saturation information, and brightness information of the target pixels included in the sub-image corresponding to the original target tracking object, and the contour model, the contour feature data corresponding to the original target tracking object is determined; In addition, based on the hue information, saturation information and brightness information of the target pixels included in the sub-image corresponding to the second candidate tracking object, and the contour model, contour feature data corresponding to the second candidate tracking object is determined, wherein the second candidate tracking object is any one of a plurality of candidate tracking objects whose color similarity with the original target tracking object is greater than the first preset threshold. Based on the contour feature data of the original target object and the contour feature data of the second candidate current target object, the contour similarity between the original target object and the second candidate current target object is determined. When it is determined that there is only one candidate current tracking object whose contour similarity to the original target tracking object is greater than a second preset threshold, the candidate current tracking object is determined as the original target tracking object.

5. The method according to claim 4, characterized in that, When it is determined that the contour similarity between multiple candidate tracking objects and the original target tracking object is greater than the second preset threshold, the method further includes: The candidate current tracking object with the highest contour similarity to the original target tracking object is selected as the target current tracking object.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Obtain the identification information of the original target object and the identification information of the current target object; The identification information of the original target being tracked is compared with the identification information of the current target being tracked. When it is determined that the identification information of the original target being tracked is different from the identification information of the current target being tracked, the identification information of the current target being tracked is updated to the identification information of the original target being tracked.

7. A device for determining a tracked object, characterized in that, The device is configured in an edge device, and the device includes: The acquisition module is used to acquire a first tracking image of the current frame, a second tracking image of the previous frame corresponding to the current frame, and bounding box information corresponding to multiple original tracking objects included in the second tracking image, wherein the bounding box information includes the coordinates of a preset target point on the bounding box, the length and width of the bounding box; and to perform image processing on the first tracking image to acquire the bounding box information corresponding to multiple original tracking objects included in the first tracking image. The determination module is configured to determine, based on the bounding box information corresponding to the original target tracking object and the bounding box information corresponding to each of the current tracking objects, whether there exists a candidate current tracking object matching the original target tracking object, wherein the original target tracking object is any one of a plurality of original tracking objects; when it is determined that there are a plurality of candidate current tracking objects, the module further configures, based on the coordinates of a preset target point, the length and width of the bounding box included in the bounding box information corresponding to the original target tracking object, to extract a sub-image corresponding to the original target tracking object from the first tracking image; and based on the coordinates of a preset target point, the length and width of the bounding box included in the bounding box information corresponding to the first candidate current tracking object, to extract a sub-image corresponding to the original target tracking object from the second tracking image. The first candidate tracking object is any one of the multiple candidate tracking objects. The target tracking object is determined based on the sub-image corresponding to the original target tracking object, the sub-image corresponding to each candidate tracking object, and a pre-constructed image processing model. The image processing model includes a color model and / or a contour model. Allocated identification information and a counting parameter corresponding to the identification information are obtained. When the value of the counting parameter is greater than a third preset threshold, the tracking information of the tracking object corresponding to the identification information is deleted, and the identification information is retrieved. The counting parameter indicates the number of consecutive frames in which the tracking object corresponding to the identification information disappears.

8. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method for determining a tracking object as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method for determining a tracked object as described in any one of claims 1 to 6.

Images