Multi-target adaptive tracking method and device, electronic equipment and storage medium

Abstract

The invention relates to a multi-target adaptive tracking method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a target image and a target tracking task, determining a plurality of tracking targets in the target image based on the target tracking task, and distributing a corresponding window for each tracking target; extracting initial positions of a plurality of tracking targets based on the linear projection of the target image at a plurality of angles; determining a switching condition of each tracking target between an independent state and a meeting state based on the initial position, tracking the plurality of tracking targets by combining the initial position and a self-adaptive tracking strategy corresponding to the switching condition, updating the window, and calculating a tracking position of each tracking target corresponding to each time of window updating, and ending the target tracking task until a preset tracking termination condition is met. Therefore, the technical problems that in the related technology, when multiple targets are processed, the time cost of data acquisition and processing is high, the updating rate is low, the tracking requirement of the high-speed moving target is difficult to meet, large data errors exist, and the tracking effect is poor are solved.

Description

Technical Field

[0001] The present application relates to the field of image data processing technology, and in particular to a multi-target adaptive tracking method, device, electronic device and storage medium. Background Art

[0002] Single-pixel target tracking technology differs fundamentally from traditional imaging tracking in its tracking mechanism. It uses a spatial light modulator (such as a digital micromirror array) to modulate the target's light field signal. A photodetector (such as a photomultiplier tube) detects the modulated signal and processes it to calculate the object's position.

[0003] In the related art, single-pixel multi-target positioning methods still have certain limitations in terms of update rate and applicability. In particular, their robustness needs to be further improved when dealing with complex and changeable multi-target motion scenes. For example, in the related art, in order to achieve multi-target tracking, it is necessary to continuously process the target's projection curve. This process relies on the frequent flipping of the DMD (Digital Micromirror Device) to generate a large number of projection patterns to obtain sufficient measurement information. However, due to the limited DMD flipping speed, this method requires a long time for data acquisition and processing when processing multiple targets, which seriously affects the real-time performance of the system, resulting in a low update rate and difficulty in meeting the tracking requirements of high-speed moving targets. In addition, the windowing method used in the related art has a certain hysteresis. Especially when the target's motion state changes dramatically, it may not be able to accurately "frame" the target, affecting its tracking accuracy. The core idea of ​​this method is to convert the target's motion changes into the Fourier domain and treat the horizontal and vertical displacements as unknowns in a nonlinear system of equations to solve. However, Fourier domain measurement data often contains errors. Even small errors can cause the calculated target displacement to deviate significantly from the actual motion. In some cases, the accumulation of errors can even render the equations unsolvable, compromising the stability and reliability of positioning.

[0004] To sum up, in the relevant technologies, when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there are large data errors, resulting in poor tracking effects, which urgently needs to be improved. Summary of the Invention

[0005] The present application provides a multi-target adaptive tracking method, device, electronic device and storage medium to solve the technical problems in related technologies that, when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there are large data errors, resulting in poor tracking effects.

[0006] The first aspect of the present application provides a multi-target adaptive tracking method, including the following steps: acquiring a target image and a target tracking task, determining multiple tracking targets in the target image based on the target tracking task, and assigning a corresponding window to each tracking target; extracting the initial positions of the multiple tracking targets based on the linear projections of the target image at multiple angles; determining the switching situation of each tracking target between an independent state and an encounter state based on the initial position, tracking the multiple tracking targets in combination with the initial position and an adaptive tracking strategy corresponding to the switching situation, updating the window, and calculating the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met to end the target tracking task.

[0007] Optionally, in one embodiment of the present application, determining the switching status of each tracking target between the independent state and the encounter state based on the initial position includes: calculating the distance between any tracking target and an adjacent tracking target based on the initial position; if the distance between at least one adjacent tracking target and the any tracking target is less than a preset distance threshold, switching the any tracking target to the encounter state; otherwise, switching the any tracking target to the independent state.

[0008] Optionally, in one embodiment of the present application, the adaptive tracking strategy corresponding to the initial position and the switching situation is used to track the multiple tracking targets, update the window, and calculate the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met, including: when any tracking target is in the independent state, tracking any tracking target to update the window based on the position change of any tracking target; combining a preset independent algorithm, the initial position and the updated window to calculate the tracking position corresponding to any tracking target when the window is updated each time, and using the tracking position corresponding to each window update as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target switches to the encounter state.

[0009] Optionally, in one embodiment of the present application, the combination of a preset independent algorithm, the initial position and the updated window to calculate the corresponding tracking position of any tracking target each time the window is updated includes: generating a mask corresponding to any tracking target based on the initial position and the window area of ​​the updated window; using error diffusion dithering to diffuse the quantization error of the pixels within the window area to adjacent pixels to obtain a grayscale mask that meets a preset gradient condition; using the grayscale mask to detect multiple light intensities within the window area, so as to use the multiple light intensities to calculate the center of mass of any tracking target, so as to obtain the corresponding tracking position based on the center of mass.

[0010] Optionally, in one embodiment of the present application, the combining of the initial position and the adaptive tracking strategy corresponding to the switching situation to track the multiple tracking targets, update the window, and calculate the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met, includes: when any of the tracking targets is in the encounter state, obtaining the any tracking target and another tracking target that encounters the any tracking target, and merging the windows of the any tracking target and the other tracking target to obtain a merged window; within the merged window, calculating the average center of mass of the any tracking target and the other tracking target, and obtaining the initial position of the average center of mass; tracking the any tracking target and the other tracking target to update the merged window based on the position change of the any tracking target and the position change of the other tracking target; combining a preset joint estimation algorithm, the initial position of the average center of mass and the updated merged window, calculating the tracking position corresponding to the average center of mass when the merged window is updated each time, and using the tracking position corresponding to each update of the merged window as the initial position of the next round of tracking, until the preset tracking termination condition is met, or the any tracking target switches to the independent state.

[0011] Optionally, in one embodiment of the present application, the combination of a preset joint estimation algorithm, the initial position of the average centroid and the updated merge window to calculate the tracking position corresponding to the average centroid each time the merge window is updated includes: detecting at least one set of measured positions of the average centroid in the merge window within a preset detection period; calculating at least one set of offsets using the initial position of the average centroid and the at least one set of measured positions, and calculating at least one set of predicted positions of the average centroid using the at least one set of offsets; optimizing a corresponding smoothing factor based on the at least one set of predicted positions and the at least one set of measured positions, so as to use the optimized smoothing factor to predict the tracking position corresponding to the average centroid each time the merge window is updated, and using the corresponding tracking position as the initial position for the next round of tracking, until the preset tracking termination condition is met, or any tracking target is switched to the independent state.

[0012] The second aspect of the present application provides a multi-target adaptive tracking device, including: an acquisition module for acquiring a target image and a target tracking task, so as to determine multiple tracking targets in the target image based on the target tracking task, and assign a corresponding window to each tracking target; an extraction module for extracting the initial positions of the multiple tracking targets based on the linear projections of the target image at multiple angles; a tracking module for determining the switching situation of each tracking target between an independent state and an encounter state based on the initial position, so as to track the multiple tracking targets in combination with the initial position and an adaptive tracking strategy corresponding to the switching situation, update the window, and calculate the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met to end the target tracking task.

[0013] Optionally, in one embodiment of the present application, the tracking module includes: a first calculation unit, configured to calculate the distance between any tracking target and an adjacent tracking target based on the initial position; and a switching unit, configured to switch any tracking target to the encounter state if the distance between at least one adjacent tracking target and the any tracking target is less than a preset distance threshold; otherwise, switch any tracking target to the independent state.

[0014] Optionally, in one embodiment of the present application, the tracking module includes: a first tracking unit, used to track any tracking target when any tracking target is in the independent state, so as to update the window based on the position change of any tracking target; a second calculation unit, used to calculate the tracking position corresponding to any tracking target each time the window is updated in combination with a preset independent algorithm, the initial position and the updated window, and use the tracking position corresponding to each time the window is updated as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target switches to the encounter state.

[0015] Optionally, in one embodiment of the present application, the second calculation unit includes: a generation subunit, used to generate a mask corresponding to any tracking target based on the initial position and the window area of ​​the updated window; a processing subunit, used to diffuse the quantization error of the pixels in the window area to adjacent pixels using error diffusion dithering to obtain a grayscale mask that meets a preset gradient condition; a first calculation subunit, used to use the grayscale mask to detect multiple light intensities in the window area, so as to calculate the center of mass of any tracking target using the multiple light intensities, so as to obtain the corresponding tracking position based on the center of mass.

[0016] Optionally, in one embodiment of the present application, the tracking module includes: a merging unit, configured to, when any tracking target is in the encounter state, obtain the any tracking target and another tracking target that encounters the any tracking target, and merge the windows of the any tracking target and the other tracking target to obtain a merged window; a third calculation unit, configured to calculate the average center of mass of the any tracking target and the other tracking target within the merged window, and obtain an initial position of the average center of mass; a second tracking unit, configured to track the any tracking target and the other tracking target to update the merged window based on the position changes of the any tracking target and the other tracking target; a fourth calculation unit, configured to calculate the tracking position corresponding to the average center of mass when the merged window is updated each time the merged window is updated, in combination with a preset joint estimation algorithm, the initial position of the average center of mass and the updated merged window, and use the tracking position corresponding to each update of the merged window as the initial position of the next round of tracking, until the preset tracking termination condition is met, or the any tracking target is switched to the independent state.

[0017] Optionally, in one embodiment of the present application, the fourth calculation unit includes: a detection subunit, used to detect at least one set of measurement positions of the average center of mass in the merged window within a preset detection period; a second calculation subunit, used to calculate at least one set of offsets using the initial position of the average center of mass and the at least one set of measurement positions, and calculate at least one set of predicted positions of the average center of mass using the at least one set of offsets; an optimization subunit, used to optimize the corresponding smoothing factor based on the at least one set of predicted positions and the at least one set of measurement positions, so as to use the optimized smoothing factor to predict the corresponding tracking position of the average center of mass when the merged window is updated each time, and use the corresponding tracking position as the initial position of the next round of tracking, until the preset tracking termination condition is met, or any tracking target is switched to the independent state.

[0018] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-target adaptive tracking method as described in the above embodiment.

[0019] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the multi-target adaptive tracking method as described in the above embodiment.

[0020] The fifth aspect of the present application provides a computer program product, including a computer program, which, when executed, is used to implement the above multi-target adaptive tracking method.

[0021] The embodiment of the present application can determine multiple tracking targets based on the target image and the target tracking task, and assign a corresponding window to each tracking target, and extract the initial positions of the multiple tracking targets based on the linear projection of the target image at multiple angles to determine the state of each tracking target, and continuously determine the switching situation of each tracking target between the independent state and the encounter state during the tracking process, so as to track multiple tracking targets in combination with the initial position and the adaptive tracking strategy corresponding to the switching situation, and update the window to calculate the position of the tracking target each time the window is updated until the tracking is terminated, thereby adaptively adjusting the tracking method in a variable motion scene to achieve stable tracking of multiple targets. Thus, the technical problem in the related art that when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there is a large data error, resulting in poor tracking effect.

[0022] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0024] Figure 1 A flowchart of a multi-target adaptive tracking method provided according to an embodiment of the present application;

[0025] Figure 2 A schematic diagram of the Radon projection principle provided according to one embodiment of the present application;

[0026] Figure 3 A schematic diagram of the principle of realizing multi-angle projection by flipping a micromirror according to one embodiment of the present application;

[0027] Figure 4 A schematic diagram of the principle of grayscale mask binarization according to one embodiment of the present application;

[0028] Figure 5 A schematic diagram of the principle of segmenting multiple targets using a window function according to an embodiment of the present application;

[0029] Figure 6 A schematic diagram of positioning interference caused by approaching a tracking target according to one embodiment of the present application;

[0030] Figure 7 A schematic diagram of close-range multi-target tracking provided according to an embodiment of the present application;

[0031] Figure 8A schematic diagram of the principle of a multi-target tracking algorithm provided according to one embodiment of the present application;

[0032] Figure 9 A schematic diagram of the principle of a multi-target adaptive tracking method provided according to one embodiment of the present application;

[0033] Figure 10 A schematic structural diagram of a multi-target adaptive tracking device provided according to an embodiment of the present application;

[0034] Figure 11 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0035] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0036] The following describes a multi-target adaptive tracking method, device, electronic device, and storage medium according to an embodiment of the present application with reference to the accompanying drawings. In view of the related art mentioned in the above background technology, when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there is a large data error, resulting in a poor tracking effect. The present application provides a multi-target adaptive tracking method, in which multiple tracking targets can be determined according to the target image and the target tracking task, and a corresponding window is assigned to each tracking target. The initial positions of the multiple tracking targets are extracted based on the linear projection of the target image at multiple angles to determine the state of each tracking target, and the switching situation of each tracking target between the independent state and the encounter state is continuously determined during the tracking process, so as to track the multiple tracking targets in combination with the initial position and the adaptive tracking strategy corresponding to the switching situation, and update the window to calculate the position of the tracking target each time the window is updated until the tracking is terminated, thereby adaptively adjusting the tracking mode in a variable motion scene to achieve stable tracking of multiple targets. This solves the technical problems in related technologies, such as high time cost of data acquisition and processing when processing multiple targets, low update rate, difficulty in meeting the tracking requirements of high-speed moving targets, large data errors, and poor tracking effects.

[0037] Specifically, Figure 1 A flowchart of a multi-target adaptive tracking method provided in an embodiment of the present application.

[0038] like Figure 1 As shown, the multi-target adaptive tracking method includes the following steps:

[0039] In step S101 , a target image and a target tracking task are acquired, so as to determine a plurality of tracking targets in the target image based on the target tracking task, and allocate a corresponding window to each tracking target.

[0040] During the actual execution process, the embodiment of the present application can obtain a target image containing a tracking target, wherein the tracking target can be determined by a target tracking task, and the target tracking task can be set accordingly according to the actual needs of the technician. After determining multiple tracking targets, the embodiment of the present application can assign a corresponding window to each tracking target, wherein the window is a local area or subset. During the tracking process, the window moves with the tracking target, and the tracking target in the image can be processed pixel by pixel according to the window corresponding to the tracking target.

[0041] In step S102 , initial positions of multiple tracking targets are extracted based on linear projections of the target image at multiple angles.

[0042] For example, the embodiment of the present application may utilize a method such as Radon projection to extract the initial position.

[0043] Among them, Radon projection is a mathematical transformation method commonly used in signal processing and image reconstruction. It converts the position information of the object from the spatial domain to the projection domain by linearly projecting the two-dimensional image at different angles, thereby simplifying the problem of multi-target recognition. Its basic principle is as follows Figure 2 For a two-dimensional image containing multiple targets, image processing methods in related technologies usually require direct analysis of complex feature information in the image. However, Radon projection analyzes the projection curves of the targets at different angles, converting complex image features into one-dimensional projection signals, thereby extracting the target's position information.

[0044] Specifically, the core idea of ​​Radon projection is to integrate the target's light field along a given angle, mapping the target's two-dimensional spatial distribution information into a set of one-dimensional curves in the projection domain. By selecting projections at multiple angles, a set of projection curves in different directions can be obtained, and these projection curves can be used to invert the target's position in the original image. For multi-target recognition, the advantage of Radon projection is that it can effectively reduce the dimensionality of the data, transforming the originally complex multi-target detection problem into a simpler curve matching problem, thereby improving computational efficiency and reducing computational complexity.

[0045] Based on the above projection scheme, the embodiment of the present application can also use DMD to perform projection modulation of the target light field at different angles. Among them, DMD is composed of many independently flippable micromirrors, each of which can flip at an angle of ±12° at high speed to control the reflection direction of the incident light. In the specific operation process, the scene light field where the target is located is first imaged on the DMD through the lens group, and then the flipping mode of the micromirrors on the DMD is controlled according to the preset projection angle, so that the micromirror array is flipped in columns to form projections at different angles. This process is as follows: Figure 3 As shown. After the DMD performs projection modulation, the reflected light carries the projection information of the target and is further transmitted to the PMT (Photomultiplier Tube) by the optical system. As a highly sensitive photodetector, the PMT can efficiently collect the modulated light signal and convert it into an electrical signal, which is then processed by a computer. By analyzing the signal output by the PMT, the projection curves of the target at different projection angles can be obtained. These curves reflect the light intensity distribution of the target in different directions. And through subsequent mathematical processing and inversion algorithms, the approximate position information of the target is finally obtained.

[0046] Based on the combination of the above technical solutions, the embodiment of the present application can obtain the initial position of each tracking target based on the linear projection of the target image at multiple angles, so as to facilitate subsequent position tracking.

[0047] In step S103, the switching situation of each tracking target between the independent state and the encounter state is determined based on the initial position, and multiple tracking targets are tracked in combination with the initial position and the adaptive tracking strategy corresponding to the switching situation. The window is updated, and the tracking position of each tracking target corresponding to each window update is calculated until the preset tracking termination condition is met to end the target tracking task.

[0048] Furthermore, since the position of the dynamic tracking target is constantly changing, during the actual tracking process, there may be situations where two or more tracking targets meet or separate. At this time, according to the different states of the tracking target, that is, the independent state and the meeting state, the embodiment of the present application can realize adaptive tracking strategy adjustment, thereby achieving accurate tracking of the tracking target.

[0049] There are many situations in which the state switching of the tracking target may occur, for example, the tracking target always maintains an independent state (i.e., does not meet other tracking targets), the tracking target meets other tracking targets in an encounter state, the tracking target meets other tracking targets and then separates from the independent state, etc.

[0050] According to the continuous state switching caused by the position change of the tracking target, the embodiment of the present application continuously updates the position of the tracking target through an adaptive switching tracking strategy, and outputs the tracking position obtained after each tracking. The updated tracking position is used as the basic position for the next round of tracking until a tracking stop signal is received, that is, the tracking termination condition is met, thereby completing the target tracking task.

[0051] Optionally, in one embodiment of the present application, determining the switching status of each tracking target between the independent state and the encounter state based on the initial position includes: calculating the distance between any tracking target and an adjacent tracking target based on the initial position; if the distance between at least one adjacent tracking target and any tracking target is less than a preset distance threshold, switching any tracking target to the encounter state; otherwise, switching any tracking target to the independent state.

[0052] In the embodiment of the present application, the distance between the tracking target and other tracking targets can be calculated to determine whether they have met. If the distance is far, it can be determined as an independent state, and if the distance is close, it can be determined as an encounter state. The state of the tracking target can be switched in real time according to the distance continuously updated during the tracking process.

[0053] The preset distance threshold may be set by those skilled in the art according to actual conditions and is not specifically limited here.

[0054] Optionally, in one embodiment of the present application, multiple tracking targets are tracked in combination with the initial position and an adaptive tracking strategy corresponding to the switching situation, the window is updated, and the tracking position of each tracking target corresponding to each window update is calculated until a preset tracking termination condition is met, including: when any tracking target is in an independent state, tracking any tracking target to update the window based on the position change of any tracking target; calculating the tracking position corresponding to any tracking target when the window is updated each time in combination with a preset independent algorithm, the initial position and the updated window, and using the tracking position corresponding to each window update as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target switches to an encounter state.

[0055] In some embodiments, for a tracking target in an independent state, the embodiments of the present application can use an independent algorithm combined with the initial position to calculate the tracking position of the tracking target for the first time when the window of the tracking target is updated, and use the tracking position of the initial tracking as the initial position for the next round of tracking, so as to update the position of the tracking target during tracking until a termination signal is received, or the tracking target is switched from an independent state to an encounter state, and the tracking position at each position update is output, or the tracking position before the state switching is output as the initial position for tracking in the encounter state.

[0056] Optionally, in one embodiment of the present application, the tracking position corresponding to any tracking target each time the window is updated is calculated in combination with a preset independent algorithm, an initial position and an updated window, including: generating a mask corresponding to any tracking target based on the initial position and the window area of ​​the updated window; using error diffusion dithering to diffuse the quantization error of the pixels in the window area to adjacent pixels to obtain a grayscale mask that meets a preset gradient condition; using the grayscale mask to detect multiple light intensities in the window area, and using the multiple light intensities to calculate the center of mass of any tracking target, and obtaining the corresponding tracking position based on the center of mass.

[0057] Here, the preset independent algorithm is explained.

[0058] After obtaining the initial position, the embodiment of the present application can use the windowed geometric moment to separate multiple objects, converting the multi-target positioning problem into the problem of multiple single-target positioning. First, the principle of using DMD to construct geometric moments for positioning is introduced. The centroid formula in the discrete scene is as follows:

[0059]

[0060] Where D(x,y) represents the light field region whose centroid is to be determined. x and y represent the coordinate values ​​on the x-axis and y-axis of the coordinate region, respectively. In a single-pixel tracking system, the DMD essentially generates a mask that reflects position information to replace the x- and y-coordinates in the equation. The mask matrix generated by the DMD is as follows:

[0061]

[0062] Among them, I and J represent the length and width of the DMD size respectively. Because DMD cannot represent specific numerical values, it is considered to control the flipping of the micromirror to form a progressive grayscale mask, and use gradient grayscale values ​​instead of gradient numerical values. However, DMD can only reflect light to generate binary patterns, and cannot directly generate grayscale gradient patterns. In order to realize grayscale modulation based on DMD, a spatial dithering modulation method is adopted, which diffuses the quantization error of the pixel to its neighboring pixels through error diffusion dithering, thereby reducing the error in the local area. This method achieves an average quantization error close to zero, and uses two grayscales of 0 and 255 to show a gradient effect from black to white, such as Figure 4 The DMD is controlled to flip according to the dithered mask, and the two rotation angles of the DMD micromirror correspond to two grayscale values.

[0063] The center of mass of the scene light field can be expressed as:

[0064]

[0065] Where ω is the modulation window area on the DMD. Using the two coding masks L1 and L2 generated by the DMD, after modulating the scene light field D(i, j) containing the target information, a photodetector is used to collect and measure it, thus achieving the summation operation in the formula, and the modulated light intensities are measured as S1 and S2 respectively. When all the micromirrors of the DMD are adjusted to the direction of the photodetector, it acts as a reflector, reflecting the scene light field D(i, j) to the detector for detection and measurement, and obtaining the light intensity S3. In the above case, the light intensity detected by the detector can be expressed as:

[0066]

[0067] When the DMD quickly flips to modulate the light field, the photodetector can immediately receive the reflected light and calculate the center of mass of the target. The calculated target center of mass coordinates (x c ,y c ) can be expressed as:

[0068]

[0069] Based on the above calculation, the embodiment of the present application can calculate the tracking position of the tracking target in the independent state according to the initial position, and use the tracking position as the initial position of the next round of tracking for cyclic calculation until a termination signal is received or the tracking target switches to the encounter state.

[0070] Optionally, in one embodiment of the present application, multiple tracking targets are tracked in combination with the initial position and an adaptive tracking strategy corresponding to the switching situation, the window is updated, and the tracking position of each tracking target corresponding to each window update is calculated until a preset tracking termination condition is met, including: when any tracking target is in an encounter state, obtaining any tracking target and another tracking target that encounters any tracking target, and merging the windows of any tracking target and the other tracking target to obtain a merged window; within the merged window, calculating the average center of mass of any tracking target and the other tracking target, and obtaining the initial position of the average center of mass; tracking any tracking target and the other tracking target to update the merged window based on the position change of any tracking target and the position change of the other tracking target; combining the preset joint estimation algorithm, the initial position of the average center of mass and the updated merged window, calculating the tracking position corresponding to the average center of mass when the merged window is updated each time, and using the tracking position corresponding to each update of the merged window as the initial position of the next round of tracking, until the preset tracking termination condition is met, or any tracking target is switched to an independent state.

[0071] In other embodiments, for a tracking target in an encounter state, the embodiments of the present application may utilize a joint estimation algorithm to implement position tracking of the tracking target.

[0072] In an embodiment of the present application, the windows of any tracking target and another tracking target can be merged, and the average center of mass under the merged window can be calculated, so as to predict the tracking position of the average center of mass based on the initial position of the average center of mass and the joint estimation algorithm until a termination signal is received, or any tracking target switches from an encounter state to an independent state, and outputs the tracking position at each position update, or outputs the tracking position before the state switching as the initial position of tracking in the independent state.

[0073] It should be noted that, in order to clearly explain the principles, the embodiment of the present application only describes the situation where two tracking targets meet. In the actual tracking process, if multiple tracking targets meet, the above technical solution can also be used to achieve joint tracking.

[0074] Optionally, in one embodiment of the present application, in combination with a preset joint estimation algorithm, the initial position of the average center of mass and the updated merge window, the tracking position corresponding to the average center of mass each time the merge window is updated is calculated, including: within a preset detection period, detecting at least one set of measured positions of the average center of mass in the merge window; using the initial position of the average center of mass and at least one set of measured positions to calculate at least one set of offsets, and using at least one set of offsets to calculate at least one set of predicted positions of the average center of mass; optimizing the corresponding smoothing factor based on at least one set of predicted positions and at least one set of measured positions, so as to use the optimized smoothing factor to predict the tracking position corresponding to the average center of mass each time the merge window is updated, and using the corresponding tracking position as the initial position for the next round of tracking, until the preset tracking termination condition is met, or any tracking target is switched to an independent state.

[0075] Here, the technical solution of tracking using a joint estimation algorithm is described.

[0076] The embodiment of the present application can achieve tracking of the tracking target in the encounter state through EWMA (Exponentially Weighted Moving Average), which is a statistical method for smooth measurement and prediction.

[0077] The present embodiment can utilize the rapid flipping frequency of the DMD and assume that the ratio of the target's velocity between two consecutive measurements within a short period of time remains constant. Therefore, the velocity at the next moment can be predicted based on the previous velocity, thereby obtaining a predicted value of the target's position at the next moment. Therefore, theoretically, the relative position of the object can be evaluated and an appropriate adaptive tracking method can be selected. The detailed theoretical derivation is as follows.

[0078] In a multi-target scenario, when the distances between the tracking targets are relatively far apart, the embodiment of the present application can consider windowing and isolating each tracking target, converting the multi-target positioning problem into a multiple single-target positioning problem. A rectangular window is selected with the initial position of the target to be positioned as the center, and the position of the DMD within the window is controlled to generate a new modulation mask. The micromirrors outside the window are flipped to a negative angle to isolate the interference of the light intensity of other targets, such as Figure 5 As shown. In this case, the embodiment of the present application can utilize the EWMA method, using the target center of mass value obtained by the geometric moment at this moment as the measured value and predicting the center of mass value at the next moment as the predicted value. By appropriately assigning weights, an estimated center of mass value at the next moment is obtained, thereby determining the position of the window function. After obtaining the actual measured center of mass value at the next moment, it is fed back to the estimated value to update the weights of the EWMA method, thereby ensuring stable tracking of objects at various speeds.

[0079] When the distance between targets is close, different objects may be close to each other, which may result in multiple objects in one frame. In this case, the center of mass information of the target cannot be directly obtained, such as Figure 5 As shown. To this end, we propose a joint tracking method based on EWMA to achieve continuous tracking of multiple targets. The following example will illustrate. Figure 7 As shown, four objects are positioned and tracked, and two of them have frame crossover interference because they are too close. At time t0, according to the trajectory prediction, frame N2 and frame N4 will cross in the next detection cycle because they are too close. At this time, it is impossible to directly obtain the center of mass of the two objects at the next cycle node. The embodiment of the present application can combine the two small frames into a large frame, that is, obtain a merged window, and then calculate the average center of mass of the two tracked targets.

[0080] Assume that the tracking targets in the merge window are g1 and g2 respectively, and the flowchart of the adaptive EWMA method is as follows: Figure 8 As shown:

[0081] Step S1: Measure and predict. First, set the scale parameter based on the displacement value of the previous detection cycle, and use the displacement of the past time period to represent the displacement of the future time period. N2 and N4 are the same frame, and the center of mass is measured at the middle time and the end time of the detection cycle respectively, and the average center of mass of g1 and g2 at these two moments is obtained. The difference between the average center of mass at time t0 can be obtained. The average center of mass displacement for the time period (t0, t0+Δt) is given by two equations for the average center of mass displacement, using the total reflected light intensity of g1 and g2 as weights. The two unknowns are the set scale parameters. Solving the equations yields the displacement Δx of a single object. mea At the same time, the motion state of the object is judged based on the first three detection cycles, thereby predicting the displacement Δx pre.

[0082] Step S2: Use the adaptive EWMA method to obtain a displacement closer to the true value. By selecting an appropriate smoothing factor, the measured value and the predicted value are weighted. The smoothing factor is then improved based on the difference between the measured value and the predicted value and used for the calculation at the next moment. Finally, the horizontal displacement of g1 and g2 in the corresponding time period is obtained and output as Δx g1 (t0+Δt) and Δx g2 (t0+Δt). At the same time, the obtained and Used for prediction and measurement of the next moment to achieve continuous high-precision tracking.

[0083] In summary, the working principle of the multi-target adaptive tracking method of the embodiment of the present application can be as follows: Figure 9 As shown, the embodiment of the present application is divided into several stages:

[0084] Initialization phase

[0085] At the beginning of tracking, the embodiment of the present application can first use Radon transform to obtain the initial position information of each target to provide basic data for subsequent tracking.

[0086] Target independent tracking (one window one object) stage

[0087] This embodiment of the present application updates the window function corresponding to each tracked target: a separate window is assigned to each tracked target, ensuring that each target is located in its own window during initial positioning. Determining whether the targets have met: If no targets have met, the position of each tracked target is calculated using an independent algorithm, and independent tracking continues. If the targets have met, the joint estimation phase begins.

[0088] Joint estimation (multiple objects in one window) stage when targets meet

[0089] Merge Window: The embodiment of the present application can merge the corresponding window functions of the encountered targets into a large window for overall processing. Calculate the target position using a joint algorithm: Within the merge window, the position information of multiple tracking targets is obtained using a joint calculation method.

[0090] Target separation determination stage

[0091] The embodiment of the present application can determine whether the tracking target is separated: if the tracking target is separated, it returns to the independent estimation stage and continues to track the tracking object according to the one-window-one-object principle. If the tracking target is still not separated, it maintains the one-window-multiple-object principle and continues the joint estimation until the tracking target is separated.

[0092] Ending stage

[0093] When the tracking target no longer needs to be tracked, the process terminates.

[0094] In summary, the embodiments of the present application can use an adaptive multi-target tracking method based on EWMA using a single-pixel detector, significantly reducing the number of masks used and improving the update rate. For N objects, only 3N masks need to be loaded into the DMD to obtain the center of mass of each object. Compared with positioning methods in related arts, the embodiments of the present application are faster and more efficient.

[0095] At the same time, the embodiments of the present application ensure high-precision tracking. The output value is used to continuously update the weight parameters to form positive feedback, which can adaptively adjust the tracking method, thereby achieving stable tracking in constantly changing motion scenes. In numerical simulations, the tracking accuracy can reach 0.7 pixels for the complex and changeable motion state of the target. Even if the targets are close to each other and cause interference, the tracking accuracy can still reach 1.1 pixels.

[0096] In the experiment, the embodiment of the present application tracked three moving targets. The results showed that the embodiment of the present application achieved stable tracking with a normalized root mean square error (NRMSE) of 0.00785. For scenes involving multiple targets with complex and dynamic motion states, the method achieved high-speed detection while maintaining high-precision tracking performance. The formula for calculating the root mean square error can be shown as follows:

[0097]

[0098] Among them, (x t ,y t ) represents the real coordinates of the object’s movement, (x e ,x e ) represents the estimated coordinates. M is the total number of locations tracked for the target.

[0099] According to the multi-target adaptive tracking method proposed in the embodiment of the present application, multiple tracking targets can be determined based on the target image and the target tracking task, and a corresponding window can be assigned to each tracking target. The initial positions of the multiple tracking targets are extracted based on the linear projection of the target image at multiple angles to determine the state of each tracking target, and the switching situation of each tracking target between the independent state and the encounter state is continuously determined during the tracking process, so as to track multiple tracking targets in combination with the initial position and the adaptive tracking strategy corresponding to the switching situation, and update the window to calculate the position of the tracking target each time the window is updated until the tracking is terminated, thereby adaptively adjusting the tracking method in a variable motion scene to achieve stable tracking of multiple targets. Thus, the technical problem in the related art that when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there is a large data error, resulting in poor tracking effect.

[0100] Next, a multi-target adaptive tracking device proposed according to an embodiment of the present application will be described with reference to the accompanying drawings.

[0101] Figure 10 4 is a block diagram of a multi-target adaptive tracking device according to an embodiment of the present application.

[0102] like Figure 10 As shown, the multi-target adaptive tracking device 10 includes: an acquisition module 100 , an extraction module 200 and a tracking module 300 .

[0103] Specifically, the acquisition module 100 is used to acquire a target image and a target tracking task, to determine multiple tracking targets in the target image based on the target tracking task, and to allocate a corresponding window to each tracking target.

[0104] The extraction module 200 is configured to extract initial positions of multiple tracking targets based on linear projections of the target image at multiple angles.

[0105] The tracking module 300 is used to determine the switching status of each tracking target between the independent state and the encounter state based on the initial position, track multiple tracking targets by combining the initial position and the adaptive tracking strategy corresponding to the switching status, update the window, and calculate the tracking position of each tracking target corresponding to each window update until the preset tracking termination condition is met, thereby ending the target tracking task.

[0106] Optionally, in one embodiment of the present application, the tracking module 300 includes: a first calculation unit and a switching unit.

[0107] The first calculation unit is configured to calculate the distance between any tracking target and an adjacent tracking target based on the initial position.

[0108] The switching unit is configured to switch any tracking target to an encounter state if the distance between at least one adjacent tracking target and any tracking target is less than a preset distance threshold; otherwise, switch any tracking target to an independent state.

[0109] Optionally, in one embodiment of the present application, the tracking module 300 includes: a first tracking unit and a second calculation unit.

[0110] The first tracking unit is configured to track any tracking target when any tracking target is in an independent state, so as to update the window based on the position change of any tracking target.

[0111] The second calculation unit is used to calculate the tracking position corresponding to any tracking target each time the window is updated by combining the preset independent algorithm, the initial position and the updated window, and use the tracking position corresponding to each time the window is updated as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target switches to the encounter state.

[0112] Optionally, in one embodiment of the present application, the second computing unit includes: a generating subunit, a processing subunit and a first computing subunit.

[0113] The generating subunit is used to generate a mask corresponding to any tracking target based on the initial position and the window area of ​​the updated window.

[0114] The processing subunit is used to diffuse the quantization errors of the pixels in the window area to the adjacent pixels by using error diffusion dithering, so as to obtain a grayscale mask that meets the preset gradient condition.

[0115] The first calculation subunit is configured to detect multiple light intensities within the window area using a grayscale mask, calculate the centroid of any tracking target using the multiple light intensities, and obtain a corresponding tracking position based on the centroid.

[0116] Optionally, in one embodiment of the present application, the tracking module 300 includes: a merging unit, a third computing unit, a second tracking unit, and a fourth computing unit.

[0117] The merging unit is configured to, when any tracking target is in an encounter state, obtain any tracking target and another tracking target that encounters the any tracking target, and merge the windows of the any tracking target and the other tracking target to obtain a merged window.

[0118] The third calculation unit is used to calculate the average center of mass of any tracked target and another tracked target within the merging window, and obtain an initial position of the average center of mass.

[0119] The second tracking unit is configured to track any one tracking target and another tracking target, so as to update the merging window based on a position change of the any one tracking target and a position change of the other tracking target.

[0120] The fourth calculation unit is used to combine the preset joint estimation algorithm, the initial position of the average centroid and the updated merge window to calculate the tracking position corresponding to the average centroid each time the merge window is updated, and use the tracking position corresponding to each time the merge window is updated as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target is switched to an independent state.

[0121] Optionally, in one embodiment of the present application, the fourth computing unit includes: a detection subunit, a second computing subunit and an optimization subunit.

[0122] The detection subunit is configured to detect at least one set of measurement positions of the average centroid within the merging window within a preset detection period.

[0123] The second calculation subunit is configured to calculate at least one set of offsets using the initial position of the average centroid and at least one set of measured positions, and calculate at least one set of predicted positions of the average centroid using the at least one set of offsets.

[0124] The optimization subunit is used to optimize the corresponding smoothing factor based on at least one set of predicted positions and at least one set of measured positions, so as to use the optimized smoothing factor to predict the tracking position corresponding to the average center of mass when the merging window is updated each time, and use the corresponding tracking position as the initial position of the next round of tracking until the preset tracking termination condition is met, or any tracking target is switched to an independent state.

[0125] It should be noted that the above explanation of the embodiment of the multi-target adaptive tracking method is also applicable to the multi-target adaptive tracking device of this embodiment, and will not be repeated here.

[0126] According to the multi-target adaptive tracking device proposed in the embodiment of the present application, multiple tracking targets can be determined based on the target image and the target tracking task, and a corresponding window can be assigned to each tracking target. The initial positions of the multiple tracking targets are extracted based on the linear projection of the target image at multiple angles to determine the state of each tracking target, and the switching situation of each tracking target between the independent state and the encounter state is continuously determined during the tracking process, so as to track multiple tracking targets in combination with the initial position and the adaptive tracking strategy corresponding to the switching situation, and update the window to calculate the position of the tracking target each time the window is updated until the tracking is terminated, thereby adaptively adjusting the tracking method in a variable motion scene to achieve stable tracking of multiple targets. Thus, the technical problem in the related art that when processing multiple targets, the time cost of data acquisition and processing is high, the update rate is low, it is difficult to meet the tracking requirements of high-speed moving targets, and there is a large data error, resulting in poor tracking effect.

[0127] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0128] A memory 1101 , a processor 1102 , and a computer program stored in the memory 1101 and executable on the processor 1102 .

[0129] When the processor 1102 executes the program, the multi-target adaptive tracking method provided in the above embodiment is implemented.

[0130] Furthermore, the electronic device further includes:

[0131] The communication interface 1103 is used for communication between the memory 1101 and the processor 1102 .

[0132] The memory 1101 is used to store computer programs that can be run on the processor 1102 .

[0133] The memory 1101 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0134] If the memory 1101, processor 1102, and communication interface 1103 are implemented independently, the communication interface 1103, memory 1101, and processor 1102 can be connected to each other via a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 11 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0135] Optionally, in a specific implementation, if the memory 1101, the processor 1102 and the communication interface 1103 are integrated on a chip, the memory 1101, the processor 1102 and the communication interface 1103 can communicate with each other through an internal interface.

[0136] The processor 1102 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0137] This embodiment also provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the multi-target adaptive tracking method described above is implemented.

[0138] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the multi-target adaptive tracking method provided by an embodiment of the present invention.

[0139] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0140] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of this application, "N" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0141] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing a custom logical function or process step, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application pertain.

[0142] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or N wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program can be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing it in other suitable ways as necessary, and then storing it in a computer memory.

[0143] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiment, the N steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0144] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0145] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0146] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A multi-target adaptive tracking method, characterized in that: The following steps are involved: Acquire a target image and a target tracking task, determine a plurality of tracking targets in the target image based on the target tracking task, and assign a corresponding window to each tracking target; Extracting initial positions of the plurality of tracking targets based on linear projections of the target image at a plurality of angles; Based on the initial position, a switching situation of each tracking target between an independent state and an encounter state is determined, and the multiple tracking targets are tracked in combination with the initial position and an adaptive tracking strategy corresponding to the switching situation. The window is updated, and the tracking position of each tracking target corresponding to each window update is calculated until a preset tracking termination condition is met, thereby ending the target tracking task.

2. The method according to claim 1, characterized in that The determining, based on the initial position, a switching condition of each tracking target between an independent state and an encounter state includes: Calculating the distance between any tracking target and an adjacent tracking target based on the initial position; If the distance between at least one adjacent tracking target and any tracking target is less than a preset distance threshold, any tracking target is switched to the encounter state; otherwise, any tracking target is switched to the independent state.

3. The method according to claim 2, characterized in that Tracking the multiple tracking targets by combining the initial position and an adaptive tracking strategy corresponding to the switching situation, updating the window, and calculating the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met, includes: When any of the tracking targets is in the independent state, tracking the any of the tracking targets to update the window based on a position change of the any of the tracking targets; The tracking position corresponding to each tracking target at each window update is calculated by combining a preset independent algorithm, the initial position, and the updated window, and the tracking position corresponding to each window update is used as the initial position for the next round of tracking until the preset tracking termination condition is met, or the any tracking target switches to the encounter state.

4. The method according to claim 3, characterized in that The calculating the tracking position corresponding to any tracking target each time the window is updated by combining the preset independent algorithm, the initial position and the updated window includes: Generate a mask corresponding to any tracking target based on the initial position and the window area of ​​the updated window; Using error diffusion dithering to diffuse the quantization errors of the pixels in the window area to adjacent pixels, so as to obtain a grayscale mask that meets a preset gradient condition; The grayscale mask is used to detect multiple light intensities within the window area, so as to calculate the center of mass of any tracking target using the multiple light intensities, and to obtain the corresponding tracking position based on the center of mass.

5. The method according to claim 2, characterized in that Tracking the multiple tracking targets by combining the initial position and an adaptive tracking strategy corresponding to the switching situation, updating the window, and calculating the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met, includes: When any of the tracking targets is in the encounter state, acquiring the any of the tracking targets and another tracking target that encounters the any of the tracking targets, and merging windows of the any of the tracking targets and the another tracking target to obtain a merged window; Calculating, within the merge window, an average center of mass of the any one tracked target and the other tracked target, and obtaining an initial position of the average center of mass; Tracking the any one tracking target and the another tracking target to update the merge window based on a position change of the any one tracking target and a position change of the another tracking target; In combination with the preset joint estimation algorithm, the initial position of the average centroid and the updated merge window, the tracking position corresponding to the average centroid each time the merge window is updated is calculated, and the tracking position corresponding to each time the merge window is updated is used as the initial position of the next round of tracking until the preset tracking termination condition is met, or any tracking target is switched to the independent state.

6. The method according to claim 5, characterized in that The step of combining a preset joint estimation algorithm, the initial position of the average centroid, and the updated merge window to calculate a tracking position corresponding to the average centroid each time the merge window is updated includes: detecting at least one set of measurement positions of the average centroid within the merging window within a preset detection period; calculating at least one set of offsets using the initial position of the average centroid and the at least one set of measured positions, and calculating at least one set of predicted positions of the average centroid using the at least one set of offsets; Based on the at least one set of predicted positions and the at least one set of measured positions, a corresponding smoothing factor is optimized to use the optimized smoothing factor to predict the tracking position corresponding to the average centroid when the merging window is updated each time, and the corresponding tracking position is used as the initial position for the next round of tracking until the preset tracking termination condition is met, or any tracking target is switched to the independent state.

7. A multi-target adaptive tracking device, characterized in that: include: an acquisition module, configured to acquire a target image and a target tracking task, determine a plurality of tracking targets in the target image based on the target tracking task, and assign a corresponding window to each tracking target; an extraction module, configured to extract initial positions of the plurality of tracking targets based on linear projections of the target image at a plurality of angles; A tracking module is configured to determine, based on the initial position, a switching condition of each tracking target between an independent state and an encounter state, track the multiple tracking targets in combination with the initial position and an adaptive tracking strategy corresponding to the switching condition, update the window, and calculate the tracking position of each tracking target corresponding to each window update until a preset tracking termination condition is met, thereby terminating the target tracking task.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the multi-target adaptive tracking method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the multi-target adaptive tracking method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed, it is used to implement the multi-target adaptive tracking method according to any one of claims 1 to 6.

Images