Target infrared imaging characteristic-based anti-interference tracking method and system

Abstract

The invention provides an anti-interference tracking method and system based on target infrared imaging characteristics. The method comprises the following steps: acquiring a continuous infrared image sequence of a target object at different time points; determining a potential motion track of the target object in the infrared image sequence by using space-time correlation analysis; a dynamic weight map is generated based on the potential motion trail, and the dynamic weight map adaptively adjusts the importance of each pixel according to the change of background noise and target features; combining the dynamic weight map with the infrared image sequence to reinforce the target area and suppress the non-target area to obtain a reinforced infrared image sequence; and a multi-hypothesis tracking strategy is adopted, real-time positioning is performed on the target object in the enhanced infrared image sequence, and continuous tracking of the target object is kept by updating the dynamic weight map. According to the technical scheme provided by the invention, the tracking precision can be improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of image processing technology, and in particular to an anti-interference tracking method and system based on target infrared imaging features. Background Art

[0002] With the development of technology, infrared imaging technology has been widely used in many fields, such as security monitoring, military reconnaissance, and unmanned driving, because of its ability to provide good visual information under low light conditions. These application scenarios have put forward higher requirements for the accuracy and stability of target tracking, especially in the presence of multiple interference factors (such as complex backgrounds, obstructions, weather changes, etc.).

[0003] Many solutions currently on the market rely on traditional image processing algorithms or simple machine learning models to achieve target detection and tracking. These methods usually use static feature matching or template-based methods to locate the target, and combine prediction models such as Kalman filters to estimate the target's future position. In addition, some systems try to use deep learning frameworks to directly extract key frames from the original video stream to optimize the tracking process.

[0004] Traditional methods often have difficulty coping with dynamic changes in the background. Once they encounter complex environmental conditions (such as strong direct sunlight, rain or snow), they are prone to misjudgment or loss of the target. Secondly, when the image contains a lot of noise, many existing algorithms perform poorly and cannot effectively distinguish the target signal from the background noise, resulting in reduced tracking accuracy. Summary of the invention

[0005] The embodiments of the present application provide an anti-interference tracking method and system based on target infrared imaging features, so as to solve the problem of poor tracking accuracy in the prior art.

[0006] In a first aspect, an embodiment of the present application provides an anti-interference tracking method based on target infrared imaging features, comprising:

[0007] Acquire a sequence of continuous infrared images of the target object at different time points;

[0008] Determining the potential motion trajectory of the target object in the infrared image sequence by using spatiotemporal correlation analysis;

[0009] Generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features;

[0010] combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence;

[0011] A multi-hypothesis tracking strategy is adopted to locate the target object in real time in the enhanced infrared image sequence, and the target object is continuously tracked by updating the dynamic weight map.

[0012] Optionally, the determining the potential motion trajectory of the target object in the infrared image sequence by using spatiotemporal correlation analysis includes:

[0013] Performing inter-frame difference calculation on the continuous infrared image sequence to obtain change information between adjacent frames;

[0014] constructing an optical flow field based on the change information, wherein the optical flow field represents the movement of pixels in the infrared image sequence from a previous frame to a next frame;

[0015] Removing noise and outliers in the optical flow field by filtering technology to obtain an optical flow estimate;

[0016] Based on the optical flow estimation, a dynamic clustering algorithm is applied to identify motion regions associated with the target object;

[0017] A correspondence between the motion region and the continuous frames is established to form a potential motion trajectory of the target object.

[0018] Optionally, generating a dynamic weight map based on the potential motion trajectory includes:

[0019] Extracting a region associated with the target object from the potential motion trajectory as a region of interest;

[0020] Performing feature analysis on the region of interest to identify key feature points representing the target object;

[0021] Calculating the importance coefficient of each pixel point to the target object according to the position of the key feature point and the change of the key feature point over time;

[0022] In combination with the background noise level in the current frame, a noise suppression factor is assigned to each pixel to reflect the degree to which the pixel is affected by the background noise;

[0023] The importance coefficient is combined with the noise suppression factor to generate a dynamic weight map.

[0024] Optionally, combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area to obtain an enhanced infrared image sequence includes:

[0025] For each pixel point in each frame of the infrared image sequence, a weight value in a dynamic weight map corresponding to the position of the pixel point is applied to enhance or weaken the intensity value of the pixel point to form an enhanced infrared image sequence;

[0026] Among them, the process of enhancing or weakening the intensity value of the pixel point includes: for the pixel points in the target area, using a first preset weight value to enhance the intensity value of the pixel point to improve the visibility of the target area; for the pixel points in the non-target area, using a second preset weight value to weaken the intensity value of the pixel point to reduce the visibility of the target area.

[0027] Optionally, the adopting of a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence and maintaining continuous tracking of the target object by updating the dynamic weight map includes:

[0028] For each frame of enhanced infrared image sequence, multiple hypotheses about the position and state of the target object are constructed;

[0029] Predicting the position of the target object in each hypothesized next frame using a prediction model;

[0030] In the next frame of enhanced infrared image sequence, the likelihood score corresponding to each hypothesis is calculated according to the predicted position;

[0031] The hypothesis with the highest likelihood score is selected as the current best estimate, and a preset number of suboptimal hypotheses are retained to cope with uncertainty;

[0032] According to the result of the current optimal estimation, the weight distribution in the dynamic weight map is updated so that the weight value is more conducive to highlighting the target object and suppressing the background noise, so as to maintain continuous tracking of the target object.

[0033] Optionally, calculating the importance coefficient of each pixel point to the target object includes:

[0034] The importance coefficient of each pixel point to the target object is calculated by the following calculation formula:

[0035]

[0036] Among them, I(x,y,t) represents the importance coefficient of the pixel at position (x,y) at time t, N is the number of key feature points, ω i (t) is the adaptive weight factor of the i-th key feature point at time t, x i (t),y i (t) are the horizontal and vertical coordinates of the i-th key feature point at time t, σ i (t) is the spatial scale parameter associated with the i-th key feature point, It is a two-dimensional Gaussian function used to measure the current position (x, y) and the key feature point (x i(t),y i The influence of the distance between (t)) on the importance coefficient, is the gradient information of key feature point i at time t, It is a nonlinear function based on the gradient information of key feature points Direction parameter θ i , local texture characteristics ψ i (t) and attention weight A t to adjust the importance factor.

[0037] Optionally, updating the weight distribution in the dynamic weight map according to the result of the current optimal estimation includes:

[0038] The updated weight distribution in the dynamic weight graph is calculated by the following calculation formula:

[0039] W′(x,y,t+1)

[0040] =W(x,y,t)+α(t)·(β(t)·I(x,y,t)-W(x,y,t))+γ(t)

[0041] ·(N(x,y,t)-μ N (t))+λ·ΔW(x,y,t)+

[0042]

[0043] Among them, W'(x,y,t+1) is the weight value of the updated dynamic weight map at the position (x,y) at time t+1, W(x,y,t) is the weight value at the current time t, α(t) is the learning rate that changes with time, β(t) is the target importance gain coefficient that changes with time, I(x,y,t) is the importance coefficient of the position as defined above, γ(t) is the noise suppression gain coefficient that changes with time, N(x,y,t) is a function that reflects the background noise level at the position (x,y) at time t, and μ N (t) is the mean value of the background noise level of the entire image at time t, λ is the weight coefficient of the smoothing term, ΔW(x,y,t) is the Laplace operator result of the weight map at the (x,y) position, η is the weight coefficient of the additional adjustment term, It is a function based on convolutional neural network (CNN) and attention mechanism, which receives the current weight map W(x, y, t), context information C t and the hidden state H t , output a correction term for further optimizing the weight graph.

[0044] In a second aspect, an embodiment of the present application provides an anti-interference tracking system based on target infrared imaging features, comprising:

[0045] An acquisition module is used to acquire a sequence of continuous infrared images of a target object at different time points;

[0046] An analysis and determination module, used for determining a potential motion trajectory of a target object in the infrared image sequence by using spatiotemporal correlation analysis;

[0047] A generating and adjusting module is used to generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features;

[0048] A combining module, used for combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence;

[0049] The real-time positioning and tracking module is used to adopt a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and to keep tracking the target object by updating the dynamic weight map.

[0050] In a third aspect, an embodiment of the present application provides a computing device, comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an anti-interference tracking method and system based on target infrared imaging characteristics as described in any one of the first aspects above.

[0051] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, an anti-interference tracking method and system based on target infrared imaging features as described in any one of the first aspects above is implemented.

[0052] In the embodiment of the present application, a continuous infrared image sequence of the target object at different time points is obtained; the potential motion trajectory of the target object in the infrared image sequence is determined by using spatiotemporal correlation analysis; a dynamic weight map is generated based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to the changes in background noise and target characteristics; the dynamic weight map is combined with the infrared image sequence to enhance the target area and suppress the non-target area to obtain an enhanced infrared image sequence; a multi-hypothesis tracking strategy is used to locate the target object in real time in the enhanced infrared image sequence, and the dynamic weight map is updated to maintain continuous tracking of the target object. The technical solution provided by the present application can improve tracking accuracy.

[0053] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0055] Figure 1 A flowchart of an anti-interference tracking method based on target infrared imaging features provided in an embodiment of the present application;

[0056] Figure 2 A schematic diagram of the structure of an anti-interference tracking system based on target infrared imaging features provided in an embodiment of the present application;

[0057] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0059] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0061] Figure 1 A flowchart of an anti-interference tracking method based on target infrared imaging features is provided for an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0062] 101. Acquire a sequence of continuous infrared images of the target object at different time points;

[0063] In this step, the goal is to collect a series of continuous infrared images, which are taken at different time points of the target area. In this way, the changes of the target over time can be captured, providing basic data for subsequent analysis.

[0064] In the embodiment of the present application, a high-resolution infrared camera is used to continuously shoot the monitored area, and the frame rate is set to ensure that fast-moving objects can be clearly recorded. For example, for a nighttime vehicle tracking task, a rate of 30 frames per second can be set to obtain all infrared images of the vehicle from entering the field of view to leaving the field of view.

[0065] 102. Determine the potential motion trajectory of the target object in the infrared image sequence by using spatiotemporal correlation analysis;

[0066] In this step, the possible path of the target is identified by performing correlation analysis in the time and space domain on the acquired series of infrared images. This method takes into account the relationship between adjacent frames and helps to improve the prediction accuracy.

[0067] In the embodiment of the present application, the optical flow method is used to calculate the pixel displacement between two frames, and then the moving direction and speed of the object in the entire sequence are inferred. Assuming that a pedestrian is being tracked, the walking route can be estimated by comparing the position changes of the pedestrian in the previous and next frames, and a preliminary trajectory model can be constructed accordingly.

[0068] This application takes into account that in the existing infrared image sequence analysis technology, there are often the following problems in determining the motion trajectory of the target object: first, inaccurate target detection caused by environmental interference (such as temperature changes, background clutter, etc.); second, it is difficult to distinguish the real moving target when the contrast between the target and the background is insufficient; third, the image is blurred due to camera shaking or the rapid movement of the target itself. These problems may lead to deviations in the tracking and identification of the target object, thereby affecting subsequent applications, such as security monitoring, obstacle detection in unmanned vehicles, etc. In order to solve the above-mentioned technical difficulties, this application proposes a method based on spatiotemporal correlation analysis to more accurately determine the potential motion trajectory of the target object in the infrared image sequence. This method can effectively filter out noise interference and improve the ability to identify real moving targets by comprehensively considering the continuity in the time dimension and the changes in spatial position.

[0069] The options are as follows:

[0070] Optionally, "the use of spatiotemporal correlation analysis to determine the potential motion trajectory of the target object in the infrared image sequence" in 102 includes: performing inter-frame difference calculation on the continuous infrared image sequence to obtain change information between adjacent frames; constructing an optical flow field based on the change information, wherein the optical flow field represents the movement of pixels in the infrared image sequence from the previous frame to the next frame; removing noise and outliers in the optical flow field through filtering technology to obtain an optical flow estimate; based on the optical flow estimate, applying a dynamic clustering algorithm to identify the motion area related to the target object; and establishing a correspondence between the motion area and continuous frames to form a potential motion trajectory of the target object.

[0071] Inter-frame difference calculation: refers to comparing the differences in pixel values ​​between two consecutive frames to detect changes in image content.

[0072] Optical flow field: is a data structure used to describe the displacement of all points in an image sequence from one frame to the next.

[0073] Filtering technology: used to reduce the noise components in the signal and make the effective information clearer and more discernible.

[0074] Dynamic clustering algorithm: A method that automatically forms categories based on data characteristics, which in this case means grouping pixel regions with similar motion characteristics together.

[0075] Potential motion trajectory: The path that the target may take, predicted based on the above processing results.

[0076] Inter-frame difference calculation refers to obtaining a differential image by subtracting adjacent frames. This step can highlight the changes in the scene.

[0077] Constructing an optical flow field is to use an optical flow algorithm (such as the Lucas-Kanade algorithm) to estimate the velocity vector of each pixel, that is, how they move from the current frame to the next frame.

[0078] Filtering denoising refers to the use of median filtering or other appropriate filters to remove erroneous matching points generated during the optical flow estimation process.

[0079] Dynamic clustering uses clustering algorithms such as K-means to aggregate pixels with similar velocity vectors to form several clusters representing the motion modes of different objects.

[0080] Establishing correspondences and forming trajectories means tracking the evolution of these clusters over time, thereby depicting the complete movement routes of each target object.

[0081] In the embodiment of the present application, it is assumed that there is a street video clip shot at night, which contains a scene of pedestrians walking. First, the frame difference method is used to obtain the difference image between each pair of adjacent frames, and then the optical flow algorithm is applied to estimate the moving direction and distance of each pixel. Then, some isolated abnormal points are eliminated by median filtering. Then, the K-means algorithm is used to classify the remaining valid optical flow vectors to find out the clusters belonging to pedestrians. Finally, the position coordinates of the center points of these clusters are linked in chronological order to obtain the exact path of pedestrians crossing the street.

[0082] By implementing this solution, we can not only overcome the problems of high false detection rate and weak anti-interference ability in traditional methods, but also significantly improve the accuracy and stability of target tracking. Especially for complex and changeable practical application scenarios, this technical means based on spatiotemporal correlation analysis has shown stronger adaptability and robustness.

[0083] 103. Generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features;

[0084] In this step, a map is created that reflects the importance of each area based on the known or predicted target motion path. This map automatically adjusts the weight value of each pixel according to the background noise level and the characteristics of the target itself, thereby highlighting the key parts.

[0085] In the embodiment of the present application, a model is trained using a machine learning algorithm to learn to distinguish between real targets and interference factors. When a new frame is detected, the model will evaluate the probability of each pixel belonging to the target and assign corresponding weights accordingly. For example, when searching for wild animals in a forest environment, the system may give higher weights to areas that show animal heat signature characteristics.

[0086] The present application takes into account that in the prior art, target tracking based on potential motion trajectories usually faces the following problems: first, interference from background noise, which may cause the target object to be misjudged or lost; second, changes in target features (such as partial occlusion, posture changes, etc.) make it difficult for the target to be continuously and stably identified. These problems limit the performance and reliability of the target tracking system. In order to solve these problems, the present application proposes a method for generating a dynamic weight map, which can adaptively adjust the importance of each pixel in the image according to changes in background noise and target features, thereby improving the accuracy and robustness of target detection and tracking.

[0087] The options are as follows:

[0088] Optionally, the "generating a dynamic weight map based on the potential motion trajectory" in 103 includes: extracting an area associated with the target object from the potential motion trajectory as a region of interest; performing feature analysis on the region of interest to identify key feature points representing the target object; calculating the importance coefficient of each pixel to the target object based on the position of the key feature points and the change of the key feature points over time; assigning a noise suppression factor to each pixel to reflect the degree to which the pixel is affected by the background noise in combination with the background noise level in the current frame; and combining the importance coefficient with the noise suppression factor to generate a dynamic weight map.

[0089] Region of interest: refers to the part of the image that is considered to contain important information (such as a moving object).

[0090] Key feature points: refers to specific locations or structures that can be used to uniquely identify the target object.

[0091] Importance coefficient: A value that measures the contribution of a pixel to determining the target position.

[0092] Noise suppression factor: A parameter that indicates the degree to which a pixel is affected by background noise.

[0093] Dynamic Weight Map: An image where each pixel value represents how important that pixel is to the current task (e.g. object tracking).

[0094] Extracting the region of interest means selecting those regions that are most likely to belong to the target object from the potential motion trajectory obtained in the previous step as the focus of further processing.

[0095] Feature analysis and key point identification use feature extraction algorithms such as SI FT and SURF to find stable and distinguishable feature points in the region of interest.

[0096] Calculating the importance coefficient means giving each pixel a score based on the position of the key feature points and their changes over time. This score reflects the importance of the pixel to the tracking target.

[0097] The noise suppression factor is assigned by evaluating the noise level of each area in the current frame and assigning a suppression factor to each pixel to reduce the impact of noise on the pixel.

[0098] Generating a dynamic weight map is to combine the importance coefficient of each pixel and the noise suppression factor to create a new image in which different gray levels or colors represent the importance of different pixels.

[0099] In the embodiment of the present application, it is assumed that pedestrians are being monitored in a video stream. First, the potential motion trajectory of the pedestrian is obtained using the aforementioned technology, and the specific area where the pedestrian is located is determined as the area of ​​interest. Next, the SIFT algorithm is applied to identify some significant feature points on the pedestrian, such as the head, shoulders, etc. Then, based on the position of these feature points and the way they move with the pedestrian, the importance coefficient of each pixel is calculated-pixels close to the feature points will receive higher weights. At the same time, considering the higher noise level that may exist in night shooting, the noise distribution in the entire scene will also be estimated, and a noise suppression factor will be given to each pixel accordingly. Finally, the importance coefficient is multiplied by the noise suppression factor, and the result constitutes a dynamic weight map. In this weight map, the pixel values ​​of the pedestrian's body are higher, while the pixel values ​​of the background and other irrelevant areas are lower.

[0100] By adopting this method, the system can more accurately distinguish between targets and non-target areas in complex backgrounds, especially in the presence of a lot of noise, and can still maintain good tracking results. In addition, by dynamically focusing on the target feature points, the focus can be effectively maintained even when the target is occluded or the posture changes, greatly improving the continuity and accuracy of the tracking process. This improvement not only enhances the system's anti-interference ability, but also improves the overall application scope and practicality.

[0101] 104. Combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence;

[0102] In this step, the dynamic weight map obtained in the previous step is applied to the original infrared image sequence. Through weighted processing, the target area becomes more obvious while reducing the influence of non-target areas.

[0103] In an embodiment of the present application, an image fusion technology is developed, which can intelligently enhance the target contour lines and reduce the background brightness according to the information of the weight map. This is particularly useful in military reconnaissance scenarios, and can help operators more easily find enemy units hidden in complex terrain.

[0104] This application takes into account that in existing infrared imaging technology, especially when tracking targets at night or in low light conditions, the following problems are often encountered: first, the contrast between the background and the foreground (i.e., the target) is insufficient, making it difficult to clearly identify the target; second, thermal noise and interference sources in the environment may degrade the image quality, further affecting the detection of the target. In order to solve these problems, this application proposes a method of combining a dynamic weight map with an infrared image sequence to improve the visibility and recognition accuracy of the target by enhancing the target area and suppressing the non-target area.

[0105] The options are as follows:

[0106] Optionally, "the step of combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area to obtain an enhanced infrared image sequence" in 104 includes: for each pixel in each frame of the infrared image sequence, applying the weight value in the dynamic weight map corresponding to the position of the pixel to enhance or weaken the intensity value of the pixel to form an enhanced infrared image sequence; wherein the process of enhancing or weakening the intensity value of the pixel includes: for the pixel in the target area, using a first preset weight value to enhance the intensity value of the pixel to improve the visibility of the target area; for the pixel in the non-target area, using a second preset weight value to weaken the intensity value of the pixel to reduce the visibility of the target area.

[0107] Infrared image sequence: A series of infrared photos taken continuously to capture the heat distribution in a scene.

[0108] Dynamic weight map: A map where each pixel value indicates how important that location is for the current task (e.g., highlighting an object).

[0109] The first preset weight value: a value greater than 1, used to amplify the intensity of the pixels in the target area.

[0110] The second preset weight value: a value less than 1 but greater than 0, used to reduce the intensity of pixels in the non-target area.

[0111] Combining the dynamic weight map with the infrared image means finding a corresponding weight value for each pixel in each frame of the infrared image according to its corresponding position on the dynamic weight map.

[0112] The pixel intensity is adjusted by multiplying the original intensity value of the pixels belonging to the target area by a higher first preset weight value, thereby increasing the brightness of these pixels.

[0113] For pixels in the non-target area, a lower second preset weight value is applied to reduce their brightness.

[0114] The enhanced infrared image sequence is generated after the above processing, and the entire image sequence will become more focused on the target object, while the background information becomes less conspicuous.

[0115] In an embodiment of the present application, it is assumed that in a security monitoring scenario, a person needs to be accurately tracked from a series of infrared images taken at night. First, a dynamic weight map for this person has been obtained according to the aforementioned technology. Next, consider a specific frame of infrared image I, and define a first preset weight value W1=1.5 and a second preset weight value W2=0.7. For any pixel P in this frame of image, if it is located in the previously determined area of ​​interest (that is, it is considered to be part of the target), the new intensity value P'=P*W1; otherwise, if P is in a non-area of ​​interest, P'=P*W2. After such processing, the new image obtained not only makes the target more obvious, but also effectively reduces the influence of the surrounding environment.

[0116] Assume the original pixel intensity is I(x,y), and the corresponding dynamic weight map is W(x,y). If a pixel belongs to the target area, then:

[0117] I'(x,y)=I(x,y)×1.5

[0118] On the contrary, if the pixel does not belong to the target area, then:

[0119] I'(x,y)=I(x,y)×0.7

[0120] Here I'(x,y) represents the pixel intensity value after processing.

[0121] This solution can significantly improve the visibility of target objects in complex backgrounds, especially in the presence of strong background interference, and can more effectively highlight the object of interest. In addition, by adjusting the relative intensity of pixels in different areas, the overall visual effect of the image can be optimized, which is convenient for subsequent manual analysis or automated processing. This method not only enhances the system's anti-noise ability, but also improves the performance of the infrared imaging system in actual application scenarios.

[0122] 105. Using a multi-hypothesis tracking strategy, real-time positioning of the target object in the enhanced infrared image sequence is performed, and continuous tracking of the target object is maintained by updating the dynamic weight map.

[0123] In this step, the final stage adopts a multi-hypothesis tracking method to locate the target in the enhanced image sequence in real time and continuously update the dynamic weight map to ensure that effective tracking can be maintained even in the event of sudden changes in the target behavior.

[0124] In an embodiment of the present application, a particle filter-based tracking framework is designed, which contains multiple hypothesis branches running in parallel, each of which represents a possible target state. As new information arrives, the system re-evaluates the likelihood of each hypothesis and selects the most likely one as the current best estimate. If the tracked object suddenly changes direction or speed, new hypothesis branches are added to adapt to this change to ensure that the tracking process is not interrupted. For example, when monitoring the movement of abnormal packages on a baggage conveyor belt during airport security checks, the system needs to be able to respond quickly to any unexpected changes in behavior patterns.

[0125] This application takes into account the following problems in existing target tracking systems, especially when dealing with targets in dynamic environments: first, the target may be partially obscured or move rapidly, resulting in tracking interruption; second, background noise and interference may affect the accurate prediction of the target position. In order to solve these problems, this application proposes a method using a multi-hypothesis tracking strategy, which improves the accuracy of real-time positioning by constructing multiple hypotheses about the target position and state and combining them with an enhanced infrared image sequence. In addition, by continuously updating the dynamic weight map, it can better adapt to changes in the target, thereby maintaining continuous tracking of the target object.

[0126] The options are as follows:

[0127] Optionally, "the use of a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and updating the dynamic weight map to maintain continuous tracking of the target object" in 105 includes: for each frame of the enhanced infrared image sequence, constructing multiple hypotheses about the position and state of the target object; using a prediction model to predict the position of the target object in the next frame of each hypothesis; in the next frame of the enhanced infrared image sequence, calculating the likelihood score corresponding to each hypothesis based on the predicted position; selecting the hypothesis with the highest likelihood score as the current optimal estimate, and retaining a preset number of suboptimal hypotheses to deal with uncertainty; and updating the weight distribution in the dynamic weight map based on the result of the current optimal estimate, so that the weight value is more conducive to highlighting the target object and suppressing background noise, so as to maintain continuous tracking of the target object.

[0128] Multi-hypothesis tracking strategy: A tracking method based on multiple prediction models that can simultaneously consider multiple possible target trajectories to cope with uncertainty.

[0129] Prediction model: A mathematical model used to estimate the location and state of a target at a certain moment in the future.

[0130] Likelihood score: A numerical measure of how well a hypothesis matches actual observed data.

[0131] Best estimate: Among all the hypotheses, the one with the highest likelihood score is considered to be the estimate that is closest to the actual situation.

[0132] Suboptimal assumptions: In addition to the optimal estimate, some assumptions that are more likely to be correct are retained to deal with uncertainty and unexpected situations.

[0133] Constructing multiple hypotheses means generating several different hypotheses about the current position and state of the target for each frame of enhanced infrared image.

[0134] Predicting the next frame position means using an appropriate prediction model (such as a Kalman filter) to predict the possible position of the target in the next frame under each hypothesis.

[0135] Calculating the likelihood score means assigning a likelihood score to each hypothesis based on comparing the predicted position with the actual observed data in the next frame.

[0136] The optimal and suboptimal hypotheses are selected to select the hypothesis with the highest score as the current best estimate, and retain a certain number of suboptimal hypotheses to deal with possible occlusions or other uncertain factors.

[0137] Updating the dynamic weight map is to adjust the dynamic weight map according to the latest best estimation result so that the weight value is more conducive to highlighting the target and suppressing background noise.

[0138] In the embodiment of the present application, it is assumed that a video stream is being monitored in which a car is driving. After the previous steps, a series of enhanced infrared images have been obtained. Now, these images are input into the multi-hypothesis tracking system. First, five different hypotheses H1 to H5 are constructed for the current frame, each of which represents the possible position and speed of the car. Then, the Kalman filter is used to predict the position of the car in the next frame under each hypothesis. When a new frame of image arrives, the likelihood score of each hypothesis is calculated by comparing the predicted position with the actual detected position of the car. Hypothesis H1 obtained the highest score of 0.9, while the scores of other hypotheses were 0.7, 0.5, 0.3 and 0.1 respectively. Therefore, H1 was selected as the current best estimate, and H2 and H3 were retained as backup hypotheses. Finally, based on the result of H1, the dynamic weight map was updated to further enhance the pixel intensity of the area where the car is located and reduce the impact of the surrounding environment.

[0139] Let H i represents the ith hypothesis, L(H i ) represents the likelihood score of the hypothesis. For each hypothesis, its likelihood score can be calculated by the following formula:

[0140]

[0141] Here, x pred,i It is assumed that H i The predicted position under x obs is the actual observed target position, T is the transpose of the matrix or vector, and P is the prediction error covariance matrix. This formula shows that the smaller the deviation between the predicted value and the observed value, the higher the likelihood score.

[0142] By introducing a multi-hypothesis tracking strategy, we can not only effectively handle complex situations such as target occlusion and rapid movement, but also mitigate the impact of environmental changes to a certain extent. In addition, as the dynamic weight map is continuously updated, the system can focus more on the target area and reduce the interference of background noise, thereby significantly improving the stability and accuracy of target tracking. This method not only enhances the robustness of the system, but also improves its performance in various practical application scenarios.

[0143] The formula in this application is used to calculate the importance coefficient of each pixel to the target object, which is an important step in many computer vision tasks such as object detection, segmentation, tracking, etc. By evaluating the importance of each pixel, the system can more effectively focus on the key areas of the target object and ignore background noise, thereby improving the accuracy of the task.

[0144] The options are as follows:

[0145] Optionally, calculating the importance coefficient of each pixel point to the target object includes:

[0146] The importance coefficient of each pixel point to the target object is calculated by the following calculation formula:

[0147]

[0148] Among them, I(x,y,t) represents the importance coefficient of the pixel at position (x,y) at time t, N is the number of key feature points, ω i (t) is the adaptive weight factor of the i-th key feature point at time t, x i (t),y i (t) are the horizontal and vertical coordinates of the i-th key feature point at time t, σ i (t) is the spatial scale parameter associated with the i-th key feature point, It is a two-dimensional Gaussian function used to measure the current position (x, y) and the key feature point (x i (t),y i The influence of the distance between (t)) on the importance coefficient, is the gradient information of key feature point i at time t, It is a nonlinear function based on the gradient information of key feature points Direction parameter θ i , local texture characteristics ψ i (t) and attention weight A t to adjust the importance factor.

[0149] The formula comprehensively considers multiple factors to evaluate the importance of pixels, including the distance between the pixel and the key feature point, the adaptive weight of the key feature point, the gradient information of the key feature point, the direction parameter, the local texture characteristics and the attention weight. This design enables the importance coefficient to fully reflect the characteristics of the target object, helping the system to locate and identify the target more accurately.

[0150] Two-dimensional Gaussian function

[0151] Function: measure the current position (x, y) and the key feature point (x i (t),y i The influence of the distance between (y)) on the importance coefficient.

[0152] Principle: The closer the pixel is to the key feature point, the higher its importance coefficient; the farther the distance is, the lower the importance coefficient.

[0153] Adaptive weight factor ω i (t):

[0154] Function: Adjust the contribution of each key feature point to the importance coefficient.

[0155] Principle: Different key feature points have different importance to the target object. The influence of each key feature point can be flexibly adjusted through the adaptive weight factor.

[0156] Gradient information

[0157] Function: Use the gradient information of key feature points to enhance the calculation of importance coefficients.

[0158] Principle: Gradient information reflects the edges and contours of the image, and these areas are usually more critical to the recognition of the target object.

[0159] Nonlinear functions

[0160] Function: Comprehensively consider gradient information, direction parameters, local texture characteristics and attention weights to adjust the importance coefficient.

[0161] Principle: Nonlinear functions can capture more complex feature relationships and improve the accuracy of importance coefficients.

[0162] The parameters are obtained as follows:

[0163] N is the number of key feature points, which is usually obtained through feature detection algorithms (such as SIFT, Harris corner detection, etc.).

[0164] ω i (t) is an adaptive weight factor that can be set by an optimization algorithm (such as gradient descent) or empirically.

[0165] x i (t),y i (t) is the horizontal and vertical coordinates of the i-th key feature point at time t, obtained through the feature detection algorithm.

[0166] σ i (t) is the spatial scale parameter, which is usually set according to the scale information of the key feature points.

[0167] It is the gradient information of key feature point i at time t, which is obtained by image gradient calculation.

[0168] θ i is a directional parameter, usually obtained through a feature detection algorithm.

[0169] ψ i (t) is the local texture feature, which can be obtained by local texture analysis (such as LBP).

[0170] A t is the attention weight, which can be obtained through an attention mechanism model (such as self-attention mechanism).

[0171] In the embodiment of the present application, it is assumed that in a target detection task, the importance coefficient of a certain pixel point (x, y) at time t needs to be calculated. The specific parameters are as follows:

[0172] N = 3: There are 3 key feature points. ω1(t) = 0.6, ω2(t) = 0.3, ω3(t) = 0.1: Adaptive weight factors.

[0173] x1(t)=10,y1(t)=20; x2(t)=30,y2(t)=40; x3(t)=50,y3(t)=60: coordinates of key feature points.

[0174] σ1(t)=5,σ2(t)=10,σ3(t)=15: spatial scale parameters.

[0175] Gradient information.

[0176] θ1=π / 4, θ2=π / 2, θ3=3π / 4: direction parameters.

[0177] ψ1(t)=0.7,ψ2(t)=0.6,ψ3(t)=0.5: local texture characteristics.

[0178] A t =0.9: attention weight.

[0179] (x,y)=(25,35): The coordinates of the pixel currently being calculated.

[0180] Assume that the nonlinear function F is defined as:

[0181]

[0182] Substitute into the formula to calculate I(x,y,t):

[0183]

[0184] Calculate each term separately:

[0185] For i=1:

[0186]

[0187] ω1(t)·0.000123·0.36≈0.000027

[0188] For i=2:

[0189]

[0190] ω2(t)·0.7788·0=0

[0191] For i=3:

[0192]

[0193] ω3(t)·0.0625·(-0.0945)≈-0.00176

[0194] Finally, add all the terms together:

[0195] I(25,35,t)=0.000027+0+(-0.00176)≈-0.001733

[0196] The calculation results show that the importance coefficient of the pixel at position (25,35) is about -0.001733. A negative value may indicate that the position is less important to the target object in the current situation, perhaps because the position is far away from the key feature point, or the gradient information, direction parameters, local texture characteristics and other factors of the key feature point have little influence. This can help the system give priority to other more important areas in subsequent processing and improve the efficiency and accuracy of target detection.

[0197] The formula in this application is derived from an adaptive weight update mechanism in a dynamic system, which is usually applied to scenarios where weights need to be adjusted in real time to optimize performance, such as target detection, tracking, image processing, etc. This mechanism allows the system to continuously adjust its weight distribution according to changes in the environment and newly acquired data, thereby improving the accuracy and robustness of the system.

[0198] The options are as follows:

[0199] Optionally, updating the weight distribution in the dynamic weight map according to the result of the current optimal estimation includes:

[0200] The updated weight distribution in the dynamic weight graph is calculated by the following calculation formula:

[0201] W′(x,y,t+1)

[0202] =W(x,y,t)+α(t)·(β(t)·I(x,y,t)-W(x,y,t))+γ(t)

[0203] ·(N(x,y,t)-μ N (t))+λ·ΔW(x,y,t)+

[0204]

[0205] Among them, W'(x,y,t+1) is the weight value of the updated dynamic weight map at the position (x,y) at time t+1, W(x,y,t) is the weight value at the current time t, α(t) is the learning rate that changes with time, β(t) is the target importance gain coefficient that changes with time, I(x,y,t) is the importance coefficient of the position as defined above, γ(t) is the noise suppression gain coefficient that changes with time, N(x,y,t) is a function that reflects the background noise level at the position (x,y) at time t, and μ N (t) is the mean value of the background noise level of the entire image at time t, λ is the weight coefficient of the smoothing term, ΔW(x,y,t) is the Laplace operator result of the weight map at the (x,y) position, η is the weight coefficient of the additional adjustment term, It is a function based on convolutional neural network (CNN) and attention mechanism, which receives the current weight map W(x, y, t), context information C t and the hidden state H t , output a correction term for further optimizing the weight graph.

[0206] The formula aims to build a dynamic model that can adaptively adjust the weight distribution according to environmental changes. By combining multiple factors (such as target importance, background noise, smoothness, etc.), the model can allocate resources more intelligently, ensuring that key areas receive sufficient attention while reducing noise interference and maintaining the smoothness of weight distribution.

[0207] Learning rate (α(t)): A parameter that controls the speed at which weight values ​​are updated, usually varying over time to adapt to the needs of different stages.

[0208] Smoothness term (λ · ΔW(x,y,t)): A weight map smoothness term calculated using the Laplacian operator to maintain the spatial consistency of the weight map.

[0209] Additional adjustments The function output based on the convolutional neural network and attention mechanism is used to further optimize the weight map.

[0210] Object importance gain β(t)·I(x,y,t): emphasizes the importance of object regions and ensures that these regions have higher weights in the weight map.

[0211] Noise suppression gain γ(t)·(N(x,y,t)-μ N (t)): Reduce the impact of background noise and improve the signal-to-noise ratio.

[0212] The parameters are obtained as follows:

[0213] W(x,y,t) and W'(x,y,t+1) are obtained through the initialization of the model and the calculation results of the previous time step.

[0214] Parameters such as α(t), β(t), γ(t), λ, and η are usually determined experimentally or automatically adjusted using some optimization algorithm (such as gradient descent).

[0215] I(x,y,t) is manually labeled according to the characteristics and position of the target or automatically identified by an algorithm.

[0216] N(x,y,t) and μ N (t) Obtained through statistical analysis of the background area.

[0217] ΔW(x,y,t) is calculated by applying the Laplacian operator to the current weight map.

[0218] Calculated using a pre-trained convolutional neural network and attention mechanism model.

[0219] In the embodiment of the present application, it is assumed that a target detection system is being developed, and the system needs to track a moving target in a video stream. In order to simplify the calculation, all parameters are set to fixed values, but in actual applications, these parameters should be dynamically adjusted.

[0220] Set the initial conditions:

[0221] W(x,y,t)0.5,I(x,y,t)=1,N(x,y,t)=0.2,μ N (t)=0.1,ΔW(x,y,t)=0.05,C t =[some context information],H t =[some hidden state].

[0222] Fixed parameters: α(t)=0.1, β(t)=0.8, γ(t)=0.5, λ=0.01, η=0.1.

[0223] Assumptions (This usually needs to be calculated through CNN and attention mechanism).

[0224] Substitute into the formula to calculate W'(x,y,t+1):

[0225] W'(x,y,t+1)=0.5+0.1·(0.8·1-0.5)+0.5·(0.2-0.1)+0.01·0.05+0.1·0.05

[0226] W'(x,y,t+1)=0.5+0.03+0.005+0.0005+0.005

[0227] W'(x,y,t+1)≈0.5405

[0228] The calculation results show that the weight at position (x, y) is updated from 0.5 to approximately 0.5405. This means that based on the current observation data and context information, the system believes that this position has a slightly increased importance for the target detection task. This may be due to the target appearing near this position, or the background noise at this position is low, which increases the system's trust in this area. In this way, the system can better adapt to environmental changes and improve the accuracy and reliability of target detection.

[0229] Figure 2 A schematic diagram of the structure of an anti-interference tracking system based on target infrared imaging features is provided for an embodiment of the present application. Figure 2 As shown, the device comprises:

[0230] An acquisition module 21 is used to acquire a sequence of continuous infrared images of a target object at different time points;

[0231] An analysis and determination module 22, configured to determine a potential motion trajectory of a target object in the infrared image sequence by using a spatiotemporal correlation analysis;

[0232] A generating and adjusting module 23, configured to generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features;

[0233] A combining module 24 is used to combine the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence;

[0234] The real-time positioning and tracking module 25 is used to adopt a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and to keep tracking the target object by updating the dynamic weight map.

[0235] Figure 2 The anti-interference tracking system based on target infrared imaging characteristics can be executed Figure 1 The implementation principle and technical effect of the anti-interference tracking method based on target infrared imaging features described in the illustrated embodiment will not be described in detail. The specific manner in which each module and unit performs operations in the anti-interference tracking system based on target infrared imaging features in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0236] In one possible design, Figure 2 The anti-interference tracking system based on target infrared imaging characteristics of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0237] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0238] The processing component 32 is used to: obtain a continuous infrared image sequence of the target object at different time points; determine the potential motion trajectory of the target object in the infrared image sequence using spatiotemporal correlation analysis; generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target characteristics; combine the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area to obtain an enhanced infrared image sequence; adopt a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and keep continuous tracking of the target object by updating the dynamic weight map.

[0239] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0240] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0241] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0242] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0243] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0244] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0245] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment is an anti-interference tracking method based on infrared imaging features of a target.

[0246] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0247] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0248] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0249] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An anti-interference tracking method based on target infrared imaging features, characterized in that: include: Acquire a sequence of continuous infrared images of the target object at different time points; Determining the potential motion trajectory of the target object in the infrared image sequence by using spatiotemporal correlation analysis; Generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features; combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence; A multi-hypothesis tracking strategy is adopted to locate the target object in real time in the enhanced infrared image sequence, and the target object is continuously tracked by updating the dynamic weight map.

2. The method according to claim 1, characterized in that The method of determining the potential motion trajectory of the target object in the infrared image sequence by using spatiotemporal correlation analysis includes: Performing inter-frame difference calculation on the continuous infrared image sequence to obtain change information between adjacent frames; constructing an optical flow field based on the change information, wherein the optical flow field represents the movement of pixels in the infrared image sequence from a previous frame to a next frame; Removing noise and outliers in the optical flow field by filtering technology to obtain an optical flow estimate; Based on the optical flow estimation, a dynamic clustering algorithm is applied to identify motion regions associated with the target object; A correspondence between the motion region and the continuous frames is established to form a potential motion trajectory of the target object.

3. The method according to claim 1, characterized in that The generating a dynamic weight map based on the potential motion trajectory comprises: Extracting a region associated with the target object from the potential motion trajectory as a region of interest; Performing feature analysis on the region of interest to identify key feature points representing the target object; Calculating the importance coefficient of each pixel point to the target object according to the position of the key feature point and the change of the key feature point over time; In combination with the background noise level in the current frame, a noise suppression factor is assigned to each pixel to reflect the degree to which the pixel is affected by the background noise; The importance coefficient is combined with the noise suppression factor to generate a dynamic weight map.

4. The method according to claim 1, characterized in that: The step of combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area to obtain an enhanced infrared image sequence includes: For each pixel point in each frame of the infrared image sequence, a weight value in a dynamic weight map corresponding to the position of the pixel point is applied to enhance or weaken the intensity value of the pixel point to form an enhanced infrared image sequence; Among them, the process of enhancing or weakening the intensity value of the pixel point includes: for the pixel points in the target area, using a first preset weight value to enhance the intensity value of the pixel point to improve the visibility of the target area; for the pixel points in the non-target area, using a second preset weight value to weaken the intensity value of the pixel point to reduce the visibility of the target area.

5. The method according to claim 1, characterized in that The method adopts a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and keeps tracking the target object by updating the dynamic weight map, including: For each frame of enhanced infrared image sequence, multiple hypotheses about the position and state of the target object are constructed; Predicting the position of the target object in each hypothesized next frame using a prediction model; In the next frame of enhanced infrared image sequence, the likelihood score corresponding to each hypothesis is calculated according to the predicted position; The hypothesis with the highest likelihood score is selected as the current best estimate, and a preset number of suboptimal hypotheses are retained to cope with uncertainty; According to the result of the current optimal estimation, the weight distribution in the dynamic weight map is updated so that the weight value is more conducive to highlighting the target object and suppressing the background noise, so as to maintain continuous tracking of the target object.

6. The method according to claim 3, characterized in that The calculating the importance coefficient of each pixel point to the target object includes: The importance coefficient of each pixel point to the target object is calculated by the following calculation formula: Where I(x, y, t) represents the importance coefficient of the pixel at position (x, y) at time t, N is the number of key feature points, ω i (t) is the adaptive weight factor of the i-th key feature point at time t, x i (t), y i (t) are the horizontal and vertical coordinates of the i-th key feature point at time t, σ i (t) is the spatial scale parameter associated with the i-th key feature point, It is a two-dimensional Gaussian function used to measure the current position (x, y) and the key feature point (x i (t), y i The influence of the distance between (t)) on the importance coefficient, is the gradient information of key feature point i at time t, It is a nonlinear function based on the gradient information of key feature points Direction parameter θ i , local texture characteristics ψ i (t) and attention weight A t to adjust the importance factor.

7. The method according to claim 5, characterized in that The updating of the weight distribution in the dynamic weight map according to the result of the current optimal estimation includes: The updated weight distribution in the dynamic weight graph is calculated by the following calculation formula: Where W′(x, y, t+1) is the weight value of the updated dynamic weight map at the position (x, y) at time t+1, W(x, y, t) is the weight value at the current time t, α(t) is the learning rate that varies with time, β(t) is the target importance gain coefficient that varies with time, I(x, y, t) is the importance coefficient of the position as defined above, γ(t) is the noise suppression gain coefficient that varies with time, N(x, y, t) is a function that reflects the background noise level at the position (x, y) at time t, and μ N (t) is the mean value of the background noise level of the entire image at time t, λ is the weight coefficient of the smoothing term, ΔW(x,y,t) is the Laplace operator result of the weight map at the (x,y) position, η is the weight coefficient of the additional adjustment term, It is a function based on convolutional neural network (CNN) and attention mechanism, which receives the current weight map W(x, y, t), context information C t and the hidden state H t , output a correction term for further optimizing the weight graph.

8. An anti-interference tracking system based on target infrared imaging features, characterized in that: include: An acquisition module is used to acquire a sequence of continuous infrared images of a target object at different time points; An analysis and determination module, used for determining a potential motion trajectory of a target object in the infrared image sequence by using spatiotemporal correlation analysis; A generating and adjusting module is used to generate a dynamic weight map based on the potential motion trajectory, wherein the dynamic weight map adaptively adjusts the importance of each pixel according to changes in background noise and target features; A combining module, used for combining the dynamic weight map with the infrared image sequence to enhance the target area and suppress the non-target area, thereby obtaining an enhanced infrared image sequence; The real-time positioning and tracking module is used to adopt a multi-hypothesis tracking strategy to locate the target object in real time in the enhanced infrared image sequence, and to keep tracking the target object by updating the dynamic weight map.

9. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement an anti-interference tracking method and system based on target infrared imaging characteristics as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, an anti-interference tracking method and system based on target infrared imaging characteristics as described in any one of claims 1 to 7 is implemented.

Images