High-dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method

Abstract

The application provides a high-dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method, comprising: monitoring an infrared image sequence through a state machine of a memory window, freezing the memory window and extracting an absolute cold background anchor point when a mutation feature appears; when a target area pixel saturation ratio is greater than a high-risk threshold, assigning a single-frame concealment degree to a limit value; otherwise, when optical vignetting overflows, a nonlinear exponential penalty is applied to the concealment degree; performing multi-peak dynamic clustering on a local background, combining the absolute cold background anchor point to determine the number of effective wave peaks, and when the number is greater than a limited value, determining a comprehensive confusion degree according to a high-order amplification multiplier and a basic single-frame confusion degree; fusing the single-frame concealment degree and the comprehensive confusion degree into a comprehensive complexity, and performing target tracking defense based on the comprehensive complexity. The application can provide an objective, real and robust anti-interference performance measurement scale in a high-dynamic infrared countermeasure scene, effectively supporting the defense decision of an automatic target tracking algorithm.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, specifically relating to a method for anti-distortion feature reconstruction and target tracking defense in high dynamic infrared adversarial scenarios. Background Technology

[0002] In highly dynamic combat scenarios, the ability of an infrared detection system to continuously lock onto a target directly depends on the accuracy of complexity assessment of the infrared image sequence. Accurately measuring the confusion (similarity between the target and the decoy) and occlusion (the degree to which the target is occluded) of a single frame image is a core prerequisite for evaluating the anti-jamming performance of Automatic Target Recognition (ATR) algorithms. However, highly dynamic infrared combat environments present extreme conditions such as strong decoy flare combustion, multiple waves of continuous thermal contamination, and superposition of ultra-low-altitude sea clutter flares, placing extremely high demands on the robustness of complexity assessment.

[0003] In related technologies, a dual-track architecture is usually adopted: extracting the texture, shape and statistical variance of the target as features, and calculating the confusion degree by comparing it with the local background statically; at the same time, calculating the occlusion degree by using the contrast and grayscale change rate of the target relative to the global background, and finally representing the image complexity by simply adding the two weights or by arithmetic average. The scheme has the following drawbacks: extremely poor resistance to physical distortion: when the brightness of the decoy exceeds the dynamic range of the detector, causing large-area overexposure and white spots, the gray-scale variance of the target area approaches zero. The existing algorithm draws the absurd conclusion that "the image has no texture and the features are extremely simple", leading to the collapse of the evaluation system; lack of temporal benchmark isolation mechanism: the existing technology directly uses the background state of the current frame to calculate the local background complexity. When multiple waves of decoys explode in succession, the background mean is raised as a whole, resulting in a smaller relative contrast between the decoys that explode later and the background. The algorithm mistakenly believes that "the subsequent interference is weaker"; unable to cope with complex clustering: for multiple independent interference sources formed by the superposition of decoys and strong reflections from the sea surface, the traditional single "mean-variance" model produces serious distribution distortion, and the true threat intensity is seriously underestimated.

[0004] Therefore, how to establish an active physical distortion prevention and multi-peak dynamic clustering mechanism when the detector encounters extreme interference phenomena, actively take over and reconstruct features at the forefront of the feature extraction process, and provide an objective and realistic anti-interference performance measurement scale has become an urgent technical problem to be solved. Summary of the Invention

[0005] To address the aforementioned problems in existing technologies, this invention provides a method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios. The technical problem to be solved by this invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios, comprising: A time-adaptive memory window is constructed, and the infrared image sequence is monitored through the state machine of the time-adaptive memory window. When the infrared image sequence shows abrupt changes, the time-adaptive memory window is frozen, and absolute cold background anchor points and healthy pure target anchor points are extracted from the frozen time-adaptive memory window. Based on the total number of target region pixels in the current frame of the infrared image sequence, the saturation ratio of the target region pixels is determined. If the saturation ratio is greater than a preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned as the extreme single-frame occlusion degree. When the saturation ratio is not greater than the preset high-risk threshold, the edge gradient decay rate of the current frame is determined based on the healthy pure target anchor point. The target bounding box area change rate of the current frame is cross-validated with the edge gradient decay rate to obtain the optical vignetting judgment result. When the optical vignetting judgment result indicates sensor optical vignetting overflow, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain the updated occlusion. The single-frame occlusion of the current frame is then assigned to the updated occlusion. Local multi-peak dynamic clustering is performed on the local background region of the current frame, and the number of effective peaks is determined in combination with the absolute cold background anchor point. If the number of effective peaks is greater than a limit value, a higher-order amplification multiplier is generated based on the number of effective peaks. The comprehensive confusion of the current frame is determined based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The overall complexity of the current frame is determined based on the single-frame occlusion degree and the overall confusion degree of the current frame, and target tracking defense is performed based on the overall complexity.

[0006] In one embodiment of the present invention, the construction of the time-adaptive memory window involves monitoring an infrared image sequence through the state machine of the time-adaptive memory window, freezing the time-adaptive memory window when abrupt changes occur in the infrared image sequence, and extracting absolute cold background anchor points and healthy pure target anchor points from the frozen time-adaptive memory window, including: A fixed-length sliding cache queue based on first-in-first-out is constructed in memory as the time-adaptive memory window; Real-time monitoring of the grayscale extreme values ​​of the local background region of the current frame, the energy centroid offset rate, the inter-frame reduction of the structural similarity index, or the disorder of the global motion vector distribution; If the grayscale extreme value experiences a nonlinear surge, or if the energy centroid offset rate experiences centroid tearing, or if the inter-frame decrease of the structural similarity index exceeds the decrease threshold, or if the disorder of the global motion vector distribution exceeds the disorder threshold, the temporal adaptive memory window is frozen. The absolute cold background anchor and the healthy pure target anchor are extracted from the frozen temporal adaptive memory window using median filtering.

[0007] In one embodiment of the present invention, the target region pixels of the current frame are the effective pixels of the current frame within the target ground truth box; determining the saturation ratio of the target region pixels based on the total number of target region pixels in the current frame of the infrared image sequence, and assigning the single-frame occlusion degree of the current frame as the limit single-frame occlusion degree when the saturation ratio is greater than a preset high-risk threshold, includes: Extract the total number of valid pixels in the target ground truth box of the current frame; Based on the preset detector overexposure grayscale threshold, the total number of overexposed pixels within the target ground truth box is determined, and the ratio of the total number of overexposed pixels to the total number of effective pixels is calculated to obtain the saturation ratio of the pixels in the target area. If the saturation ratio is greater than the preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned the limit single-frame occlusion degree.

[0008] In one embodiment of the present invention, the process of determining the edge gradient decay rate of the current frame based on the healthy pure target anchor point, performing cross-validation based on the target bounding box area change rate of the current frame and the edge gradient decay rate to obtain an optical vignetting judgment result; and when the optical vignetting judgment result indicates sensor optical vignetting overflow, applying a nonlinear exponential penalty to the single-frame occlusion of the current frame to obtain an updated occlusion, includes: Adaptive Gaussian smoothing is performed on the target region pixels of the current frame. An edge extraction operator is used to perform convolution operation on the target region pixels after adaptive Gaussian smoothing to obtain the first-order partial derivative of the target region pixels. The sum of edge gradients of the current frame is determined based on the first-order partial derivative of the target region pixels. The edge gradient decay rate of the current frame is determined based on the sum of the edge gradients of the current frame and the reference gradient of the healthy pure target anchor point. When the edge gradient decay rate is less than a preset retention threshold and the target bounding box area change rate of the current frame is greater than a preset change threshold, the optical vignetting judgment result characterizes the sensor optical vignetting overflow. A nonlinear exponential penalty multiplier factor is constructed based on the edge gradient decay rate of the current frame. A nonlinear exponential penalty is applied to the single-frame occlusion of the current frame based on the nonlinear exponential penalty multiplier factor to obtain the updated occlusion.

[0009] In one embodiment of the present invention, performing local multi-peak dynamic clustering on the local background region of the current frame and determining the number of effective peaks in combination with the absolute cold background anchor points includes: The local background region of the current frame is flattened into a one-dimensional vector, and extreme pulse noise filtering is performed on the one-dimensional vector to obtain a cleaned one-dimensional vector. A clustering model with a preset number of groups is initialized in parallel. The cleaned one-dimensional vector is dynamically clustered to obtain each group of clustering models after convergence. The best-fitting model is determined from each group of clustering models after convergence according to the model selection criteria. Valid peaks with pixel weights greater than a preset lower limit and mean values ​​higher than the absolute cold background anchor points are selected from the best-fit model, and the number of valid peaks is counted.

[0010] In one embodiment of the present invention, the clustering model employs a Gaussian mixture model, a density-based clustering algorithm, or a mean-shift algorithm; When the clustering model adopts the Gaussian mixture model, the step of dynamically clustering the cleaned one-dimensional vector to obtain converged clustering models, and determining the best-fitting model from the converged clustering models according to the model selection criteria, includes: The cleaned one-dimensional vector is dynamically clustered using the expectation-maximization algorithm to obtain converged clustering models. The best-fit model is then determined from the converged clustering models based on the Bayesian information criterion.

[0011] In one embodiment of the present invention, determining the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame includes: The single-frame occlusion degree of the current frame and the overall confusion degree of the current frame are fused using the harmonic mean paradigm to obtain the overall complexity of the current frame.

[0012] Secondly, the present invention provides a high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense device, comprising: An anchor point extraction module is used to construct a time-adaptive memory window, monitor the infrared image sequence through the state machine of the time-adaptive memory window, freeze the time-adaptive memory window when the infrared image sequence shows abrupt changes, and extract absolute cold background anchor points and healthy pure target anchor points from the frozen time-adaptive memory window. The occlusion assignment module is used to determine the saturation ratio of the target region pixels based on the total number of target region pixels in the current frame of the infrared image sequence, and to assign the single-frame occlusion of the current frame to the limit single-frame occlusion when the saturation ratio is greater than a preset high-risk threshold. The occlusion update module is used to determine the edge gradient decay rate of the current frame based on the healthy pure target anchor point when the saturation ratio is not greater than the preset high-risk threshold, and to perform cross-validation based on the target bounding box area change rate of the current frame and the edge gradient decay rate to obtain the optical vignetting judgment result; when the optical vignetting judgment result indicates sensor optical vignetting overflow, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain the updated occlusion, and the single-frame occlusion of the current frame is assigned as the updated occlusion; The confusion determination module is used to perform local multi-peak dynamic clustering on the local background region of the current frame, determine the number of effective peaks in combination with the absolute cold background anchor point, generate a higher-order amplification multiplier based on the number of effective peaks when the number of effective peaks is greater than a limit value, and determine the comprehensive confusion of the current frame based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The complexity determination module is used to determine the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame, and to perform target tracking defense based on the overall complexity.

[0013] Thirdly, the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the high dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method provided in the first aspect.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the steps of the high dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method provided in the first aspect.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention provides a method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios. This method monitors infrared image sequences using a state machine constructed with a temporal adaptive memory window. When abrupt changes occur in the infrared image sequence, the temporal adaptive memory window is frozen, and absolute cold background anchor points and healthy pure target anchor points are extracted from the frozen window. Based on the total number of target region pixels in the current frame of the infrared image sequence, the saturation ratio of the target region pixels is determined. If the saturation ratio is greater than a preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned as the extreme single-frame occlusion degree. If the saturation ratio is not greater than the preset high-risk threshold, cross-validation is performed based on the target bounding box area change rate and edge gradient decay rate of the current frame. The optical vignetting judgment result is obtained. When the optical vignetting judgment result indicates that the sensor's optical vignetting overflows, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain the updated occlusion. The single-frame occlusion of the current frame is then assigned as the updated occlusion. Local multi-peak dynamic clustering is performed on the local background region of the current frame. The number of effective peaks is determined by combining the absolute cold background anchor point. If the number of effective peaks is greater than the limit, a higher-order amplification multiplier is generated based on the number of effective peaks. The comprehensive confusion of the current frame is determined based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The comprehensive complexity of the current frame is determined based on the single-frame occlusion and the comprehensive confusion of the current frame. Target tracking defense is then performed based on the comprehensive complexity. This invention achieves complete isolation of background reference drift caused by continuous thermal contamination by constructing a temporally adaptive memory window and extracting absolute cold background anchor points and healthy pure target anchor points as fixed benchmarks; through a pixel saturation hard truncation mechanism, it directly outputs the ultimate occlusion degree under extreme conditions of detector overexposure and white spots, avoiding misjudgment by conventional algorithms in the feature-devouring state; by cross-validating the target bounding box area change rate and edge gradient decay rate to identify optical halo overflow and applying a nonlinear exponential penalty, it achieves accurate quantification and occlusion degree reconstruction of high-frequency texture structural damage; through local... Multi-peak dynamic clustering combined with absolute cold background anchor points is used to screen the number of effective peaks and generate higher-order amplification multipliers to separate multi-source independent interference from the chaotic background thermal pollution and reconstruct the comprehensive confusion degree. Finally, the single-frame occlusion degree and the comprehensive confusion degree are fused into the comprehensive complexity using the harmonic mean paradigm. This ensures that the complexity evaluation only shows a non-linear spike when the confusion risk of bright deception and the occlusion risk of optical occlusion are superimposed. Thus, it provides an objective, realistic and robust anti-interference performance metric in high-dynamic infrared countermeasures scenarios, effectively supporting the defense decision of automatic target tracking algorithms.

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] Figure 1This is a schematic diagram illustrating the implementation process of a high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a state machine switching based on a time-adaptive memory window provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of peak stripping in local multi-peak dynamic clustering provided by an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense device provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the hardware entity of an electronic device to which embodiments of the present invention are applied. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0019] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating the implementation process of a high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense method provided by an embodiment of the present invention. Figure 1 As shown in the embodiments of the present invention, the method for anti-distortion feature reconstruction and target tracking defense in high dynamic infrared countermeasure scenarios may include the following steps: Step S101: Construct a time-adaptive memory window. Monitor the infrared image sequence through the state machine of the time-adaptive memory window. When the infrared image sequence shows abrupt changes, freeze the time-adaptive memory window and extract the absolute cold background anchor point and the healthy pure target anchor point from the frozen time-adaptive memory window.

[0020] See Figure 2 A fixed-length sliding buffer queue (e.g., 30 frames) based on First In First Out (FIFO) can be constructed in memory as a time-adaptive memory window. After system initialization, it enters the update state by default. In this state, the system can perform safe cruise and continuously accumulate features: the memory window (corresponding to the time-adaptive memory window in the above implementation) is adaptively updated with the input image sequence, dynamically pushing and continuously accumulating the health features of the most recent 30 frames (e.g., statistical benchmarks of pure targets and local backgrounds). At the same time, the FIFO buffer formation is dynamically resident, and multi-dimensional mutation index monitoring is dynamically executed (corresponding to monitoring the infrared image sequence through a state machine in the above implementation).

[0021] The monitoring hierarchy can monitor in real time the grayscale extremes of the local background region in the current frame, the energy centroid shift rate, the inter-frame reduction of the structural similarity index, or the disorder of the global motion vector distribution. For example... Figure 2 As shown, dual mutation monitoring can be performed. When any of the set mutation threshold conditions are met (i.e., a nonlinear surge occurs in the local background grayscale extreme value, or there is obvious centroid tearing in the energy centroid offset rate, or the inter-frame drop of the Structural Similarity Index (SSIM) is greater than the drop threshold (such as a value between 0.3 and 0.5, or other values ​​set according to the actual situation), or the disorder of the global motion vector distribution determined based on the optical flow method is greater than the disorder threshold (such as a value between 0.5 and 0.7, or other values ​​set according to the actual situation)), the system determines that a sudden high dynamic interference occurs in the field of view. The mechanism gating triggers the interceptor to be opened, and the state machine instantly switches from the update state to the locked state.

[0022] Upon entering the locked state, the sliding memory window (corresponding to the time-adaptive memory window in the above implementation) is forcibly frozen (suspended), completely blocking the contamination of cached samples in memory by continuous hot contamination data during subsequent adversarial cycles, i.e., isolating continuous hot contamination interference from the environment. Subsequently, the system extracts dual absolute reference anchor points, i.e., absolute cold background anchor points, from the frozen time-adaptive memory window by performing median filtering in the time dimension. Anchor point for pure health goals Subsequently, throughout the entire adversarial cycle, the decay calculations of all current frames are forcibly compared with these two fixed absolute anchor points to achieve temporal isolation of continuous thermal pollution in the environment.

[0023] Further, see Figure 2 The bottom loop can continuously monitor the overall energy dissipation state of the local background in the current frame while in the locked state. When the average gray level and energy variance of the local background gradually decrease and meet the adaptive reset mechanism trigger condition of approaching the tolerance range of the absolute cold background anchor point for M consecutive frames (such as 10 frames), it is determined that the current high dynamic interference dissipation has ended. The system unlocks and resets and clears the feature cache queue. The state machine returns to the "update state" to prepare data for the next round of feature capture.

[0024] Step S102: Based on the total number of target region pixels in the current frame of the infrared image sequence, determine the saturation ratio of the target region pixels. If the saturation ratio is greater than the preset high-risk threshold, assign the single-frame occlusion degree of the current frame to the limit single-frame occlusion degree.

[0025] In one possible implementation, the target region pixels of the current frame are the valid pixels within the target ground truth box of the current frame. The total number of valid pixels within the target ground truth box of the current frame can be extracted. Based on the preset detector overexposure grayscale threshold, determine the total number of overexposed pixels within the target ground truth box. The ratio of overexposed pixels to effective pixels is calculated to obtain the saturation ratio of pixels in the target area. ; at saturation ratio If the percentage exceeds a preset high-risk threshold (such as 80% or other values ​​set according to actual conditions), a physical feature swallowing phenomenon is determined to have occurred, and a pixel saturation hard truncation mechanism is triggered: the normal feature distance Euclidean calculation trunk is directly interrupted, and the single-frame occlusion of the current frame is forcibly reduced. The value is assigned to 0.99, a preset extreme value representing the maximum occlusion of a single frame.

[0026] Step S103: When the saturation ratio is not greater than the preset high-risk threshold, determine the edge gradient decay rate of the current frame based on the healthy pure target anchor point, and perform cross-validation based on the target bounding box area change rate and the edge gradient decay rate of the current frame to obtain the optical vignetting judgment result; when the optical vignetting judgment result indicates that the sensor optical vignetting overflows, apply a nonlinear exponential penalty to the single-frame occlusion degree of the current frame to obtain the updated occlusion degree, and assign the single-frame occlusion degree of the current frame to the updated occlusion degree.

[0027] In one possible implementation, when the saturation percentage is no greater than a preset high-risk threshold, adaptive Gaussian smoothing can be applied to the target region pixels of the current frame. An edge extraction operator (such as the Sobel operator, Canny edge detector, or Laplacian second derivative operator) is then used to convolve the adaptively Gaussian smoothed target region pixels to obtain the first-order partial derivatives of the target region pixels. The sum of edge gradients in the current frame is determined based on these first-order partial derivatives. The edge gradient decay rate of the current frame is then determined based on the sum of edge gradients in the current frame and the baseline gradient of the healthy pure target anchor point. When the edge gradient decay rate is less than a preset retention threshold (e.g., 30% or other values ​​set according to actual conditions), and the target bounding box area change rate of the current frame is greater than a preset change threshold (e.g., a value between 10% and 20%, or other values ​​set according to actual conditions), the optical vignette judgment result characterizes sensor optical vignette overflow. A nonlinear exponential penalty multiplier factor is constructed based on the edge gradient decay rate of the current frame. A nonlinear exponential penalty is applied to the single-frame occlusion of the current frame based on the nonlinear exponential penalty multiplier factor to obtain the updated occlusion. The single-frame occlusion of the current frame is then assigned as the updated occlusion. The nonlinear exponential penalty multiplier factor can be determined by the following formula: ; in, This is a non-linear exponential penalty multiplier factor. The preset coefficients, This represents the edge gradient decay rate. It can be seen that the lower the edge gradient decay rate, the larger the nonlinear exponential penalty multiplier factor, resulting in a greater proportion of the current frame's single-frame occlusion being pushed up. This forcibly pushes up the reconstructed single-frame occlusion of the current frame to obtain the updated occlusion, and then assigns the current frame's single-frame occlusion as the updated occlusion.

[0028] Step S104: Perform local multi-peak dynamic clustering on the local background region of the current frame, determine the number of effective peaks by combining the absolute cold background anchor point, and generate a higher-order magnification multiplier based on the number of effective peaks when the number of effective peaks is greater than the limit value. Determine the comprehensive confusion of the current frame based on the higher-order magnification multiplier and the basic single-frame confusion of the current frame.

[0029] The basic single-frame obfuscation is the initial obfuscation of a single-frame image calculated by an image processing algorithm.

[0030] For example, an attentional gravitational field can be applied in a two-dimensional Gaussian space. Under global constraints, independent interference blocks are extracted, and after spatial weighting, the top-3 threat sources are aggregated to obtain the static feature confusion level. As shown in the following formula, static feature confusion can be used as a basis. Difference in extreme attenuation values and normalized weighted gravitational energy center of mass deviation rate Determine the base single-frame obfuscation of the current frame. : ; When faced with boundary distortion conditions such as extreme pixel overexposure, high dynamic range deception, or multimodal coupling, the "anti-distortion feature reconstruction" mechanism described in this invention will proactively take over and adjust the basic single-frame confusion degree obtained from the above formula. The overall ambiguity of the current frame is obtained by implementing forced coverage or exponential amplification based on a higher-order amplification multiplier. It is understandable that, in cases not involving boundary distortion, the basic single-frame ambiguity of the current frame can be used as the overall ambiguity of the current frame.

[0031] Local multi-peak dynamic clustering is performed on the local background region of the current frame. The number of effective peaks is determined by combining the absolute cold background anchor point. If the number of effective peaks is greater than the limit, a higher-order amplification multiplier is generated based on the number of effective peaks. The comprehensive confusion of the current frame is determined based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame.

[0032] The specific mathematical stripping morphology and algorithmic mechanism of the local multi-peak dynamic clustering in this step are referenced. Figure 3As shown. First, the two-dimensional grayscale matrix of the local background region of the current frame, which contains clutter and interfering halo, can be flattened into a one-dimensional vector (i.e., Figure 3 The local background pixel grayscale values ​​on the horizontal axis are used to perform extreme percentile filtering on the one-dimensional vector to remove pulse noise, resulting in a cleaned one-dimensional vector (i.e., the local background pixel grayscale values ​​on the horizontal axis). Figure 3 (Normalized domain of the horizontal axis). Figure 3 The horizontal axis represents the normalized local background pixel grayscale value, and the vertical axis represents the probability density of pixels with different grayscale values ​​appearing in the local background region.

[0033] Subsequently, Gaussian Mixture Models (GMMs) with a preset number of groups (e.g., 1 to 5 groups) can be initialized in parallel as clustering models. The Expectation-Maximization (EM) algorithm is used to dynamically iteratively fit the cleaned one-dimensional vectors, calculate the weight components, mean, and variance of each Gaussian component, and automatically determine the best-fit model from the converged clustering models based on the Bayesian Information Criterion (BIC).

[0034] like Figure 3 As shown, the system establishes a physical qualification review mechanism to filter the components output by the best-fit model: first, it removes tiny noise pixels with pixel weights less than a preset lower limit (e.g., 0.05), then compares the mean values ​​of each component, and filters out components with mean values ​​lower than the absolute cold background anchor point benchmark, retaining only those with mean values ​​significantly higher than the absolute cold background anchor point benchmark locked in the previous step (corresponding to the absolute cold background anchor point in the above implementation method). The composition, and the number of remaining independent effective Gaussian peaks. . Figure 3 The original local background composite distribution curve in the text is the curve corresponding to the one-dimensional vector after cleaning; Component 1: the background Gaussian distribution of sea clutter corresponds to the intercepted area belonging to the background clutter, that is, the area with a mean value lower than the absolute cold background anchor point; Component 2: independent effective interference peak 1 and Component 3: independent effective interference peak 2 are both effective peaks with a mean value higher than the absolute cold background anchor point and which have passed the qualification review.

[0035] like Figure 3 The GMM clustering stripping mechanism in the text is based on the condition of boundary distortion and the number of effective interference peaks (corresponding to the effective peaks in the above implementation method). When a complex multimodal coupling threat is detected in the current environment, the system will trigger a higher-order nonlinear ambiguity amplification to achieve anti-distortion feature reconstruction: based on the number of effective peaks. Generate higher-order multipliers based on the natural logarithm normal form Finally, the higher-order amplification multiplier is used to perform product correction on the previous basic single-frame confusion and apply hard truncation constraint to output the reconstructed high-fidelity comprehensive confusion (i.e. the comprehensive confusion of the current frame), which fundamentally isolates the statistical distribution distortion caused by multimodal interference coupling.

[0036] Step S105: Determine the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame, and perform target tracking defense based on the overall complexity.

[0037] In one possible implementation, the overall complexity of the current frame can be obtained by fusing the single-frame occlusion degree and the overall confusion degree of the current frame using the harmonic mean normal form, as shown by the following formula: ; in, The overall complexity of the current frame is... This represents the single-frame occlusion level of the current frame. The overall obfuscation level of the current frame. It is a very small constant.

[0038] By implementing the above method, it can be ensured that the overall complexity of the current frame only exhibits a non-linear spike when the dual risks of highlight deception and optical occlusion are superimposed.

[0039] In one possible implementation, a line-of-sight adjustment command can be generated based on the overall complexity, and the line-of-sight adjustment command can be output to the photoelectric servo gimbal of the detection system to drive the photoelectric servo gimbal to deflect and avoid high-risk interference blind spots, thereby achieving target tracking and defense.

[0040] In one possible implementation, the overall complexity of the current frame can be... Expanding into both time and space dimensions to achieve target tracking and defense: For time-domain dynamic sequence defense, a normalized remaining target hit time can be extracted to construct an end-point approach penalty factor. This factor, combined with Sigmoid temporal gating activation, generates a composite corrected weighted coefficient. Since the overload required to correct trajectory deviations of the tracking aircraft is inversely proportional to the square of the remaining time, the system outputs a comprehensive sequence tracking failure score. This score is coupled with the current frame's overall complexity. When the overall complexity exhibits a non-linear spike due to the combined effects of highlight deception and optical occlusion, and the comprehensive sequence tracking failure score simultaneously triggers saturation truncation, the system determines that the conventional tracking algorithm will inevitably fail and directly issues a warning to the tracking control center system.

[0041] For omnidirectional defense line-of-sight planning in the spatial domain, a uniform viewpoint array can be generated using Fibonacci Sphere Sampling (FSS) with the target as the origin. The complexity of each viewpoint is calculated in batches using the virtual camera projection matrix. The calculation results for each viewpoint are compared and mapped with the overall complexity of the corresponding frame, and an omnidirectional 3D vulnerability heatmap is generated. The high-risk blind spots in this 3D vulnerability heatmap directly correspond to the frame time intervals where the overall complexity nonlinearly spikes.

[0042] Finally, the comprehensive complexity score can be tactically integrated with the 3D vulnerability heatmap to empower military engineering. On the one hand, comprehensive complexity can be used as a quantitative benchmark to guide the quantitative evaluation and parameter optimization of anti-jamming tracking algorithms for infrared detection systems; on the other hand, the surge range of comprehensive complexity can be used as an avoidance constraint to guide the optimal approach path and autonomous line-of-sight avoidance planning of tracking aircraft at the terminal stage, thereby achieving target tracking and defense at both the tactical and algorithmic levels in highly dynamic infrared countermeasures scenarios.

[0043] The present invention provides a method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios. This method monitors infrared image sequences using a state machine constructed with a temporal adaptive memory window. When abrupt changes occur in the infrared image sequence, the temporal adaptive memory window is frozen, and absolute cold background anchor points and healthy pure target anchor points are extracted from the frozen window. Based on the total number of target region pixels in the current frame of the infrared image sequence, the saturation ratio of the target region pixels is determined. If the saturation ratio is greater than a preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned as the extreme single-frame occlusion degree. If the saturation ratio is not greater than the preset high-risk threshold, cross-validation is performed based on the target bounding box area change rate and edge gradient decay rate of the current frame. The optical vignetting judgment result is obtained. When the optical vignetting judgment result indicates that the sensor's optical vignetting overflows, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain the updated occlusion. The single-frame occlusion of the current frame is then assigned as the updated occlusion. Local multi-peak dynamic clustering is performed on the local background region of the current frame. The number of effective peaks is determined by combining the absolute cold background anchor point. If the number of effective peaks is greater than the limit, a higher-order amplification multiplier is generated based on the number of effective peaks. The comprehensive confusion of the current frame is determined based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The comprehensive complexity of the current frame is determined based on the single-frame occlusion and the comprehensive confusion of the current frame. Target tracking defense is then performed based on the comprehensive complexity. This invention achieves complete isolation of background reference drift caused by continuous thermal contamination by constructing a temporally adaptive memory window and extracting absolute cold background anchor points and healthy pure target anchor points as fixed benchmarks; through a pixel saturation hard truncation mechanism, it directly outputs the ultimate occlusion degree under extreme conditions of detector overexposure and white spots, avoiding misjudgment by conventional algorithms in the feature-devouring state; by cross-validating the target bounding box area change rate and edge gradient decay rate to identify optical halo overflow and applying a nonlinear exponential penalty, it achieves accurate quantification and occlusion degree reconstruction of high-frequency texture structural damage; through local... Multi-peak dynamic clustering combined with absolute cold background anchor points is used to screen the number of effective peaks and generate higher-order amplification multipliers to separate multi-source independent interference from the chaotic background thermal pollution and reconstruct the comprehensive confusion degree. Finally, the single-frame occlusion degree and the comprehensive confusion degree are fused into the comprehensive complexity using the harmonic mean paradigm. This ensures that the complexity evaluation only shows a non-linear spike when the confusion risk of bright deception and the occlusion risk of optical occlusion are superimposed. Thus, it provides an objective, realistic and robust anti-interference performance metric in high-dynamic infrared countermeasures scenarios, effectively supporting the defense decision of automatic target tracking algorithms.

[0044] Please see Figure 4 , Figure 4This is a schematic diagram of a high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense device provided by an embodiment of the present invention. The high-dynamic infrared countermeasure scenario anti-distortion feature reconstruction and target tracking defense device provided by the present invention includes an anchor point extraction module 401, an occlusion assignment module 402, an occlusion update module 403, an confusion determination module 404, and a complexity determination module 405.

[0045] Among them, the anchor point extraction module 401 is used to construct a time-adaptive memory window, monitor the infrared image sequence through the state machine of the time-adaptive memory window, freeze the time-adaptive memory window when the infrared image sequence shows abrupt changes, and extract the absolute cold background anchor point and the healthy pure target anchor point from the frozen time-adaptive memory window. The occlusion assignment module 402 is used to determine the saturation ratio of the target region pixels based on the total number of target region pixels in the current frame of the infrared image sequence. If the saturation ratio is greater than the preset high-risk threshold, the single-frame occlusion of the current frame is assigned as the extreme single-frame occlusion. The occlusion update module 403 is used to determine the edge gradient decay rate of the current frame based on the healthy pure target anchor point when the saturation ratio is not greater than the preset high-risk threshold, and to perform cross-validation based on the target bounding box area change rate and the edge gradient decay rate of the current frame to obtain the optical vignetting judgment result; when the optical vignetting judgment result indicates that the sensor optical vignetting overflows, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain the updated occlusion, and the single-frame occlusion of the current frame is assigned as the updated occlusion. The confusion determination module 404 is used to perform local multi-peak dynamic clustering on the local background region of the current frame, determine the number of effective peaks by combining the absolute cold background anchor point, generate a higher-order amplification multiplier based on the number of effective peaks when the number of effective peaks is greater than the limit value, and determine the comprehensive confusion of the current frame based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The complexity determination module 405 is used to determine the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame, and to perform target tracking defense based on the overall complexity.

[0046] Please see Figure 5 , Figure 5 This is a schematic diagram of the hardware entity of an electronic device applicable to an embodiment of the present invention. The electronic device provided in this embodiment includes a processor 501, a communication interface 502, a memory 503, and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other via the communication bus 504. Memory 503 is used to store computer programs; When the processor 501 executes the program stored in the memory 503, it implements the steps in the above method embodiments.

[0047] The communication bus 504 mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus 504 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, it is represented by only one thick line in the figure, but this does not indicate that there is only one bus or one type of bus.

[0048] Communication interface 502 is used for communication between the above-mentioned electronic device and other devices.

[0049] The memory 503 may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory 503 may also be at least one storage device located remotely from the aforementioned processor.

[0050] The processor 501 mentioned above can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0051] The method provided in this invention can be applied to electronic devices. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc. No limitation is made herein; any electronic device that can implement this invention falls within the protection scope of this invention.

[0052] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps provided in the above-described method embodiments.

[0053] For the embodiments of the device / electronic device / storage medium, since they are basically similar to the method embodiments, the description is relatively simple. For specific details and beneficial effects, please refer to the description of the method embodiments.

[0054] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0055] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0056] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus (devices), or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects, all of which are collectively referred to herein as "modules" or "systems." Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The computer program may be stored / distributed in a suitable medium, provided with or as part of other hardware, or may take other distribution forms, such as via the Internet or other wired or wireless telecommunications systems.

[0057] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0058] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0059] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0060] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios, characterized in that, include: A time-adaptive memory window is constructed, and the infrared image sequence is monitored through the state machine of the time-adaptive memory window. When the infrared image sequence shows abrupt changes, the time-adaptive memory window is frozen, and absolute cold background anchor points and healthy pure target anchor points are extracted from the frozen time-adaptive memory window. Based on the total number of target region pixels in the current frame of the infrared image sequence, the saturation ratio of the target region pixels is determined. If the saturation ratio is greater than a preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned as the extreme single-frame occlusion degree. When the saturation ratio is not greater than the preset high-risk threshold, the edge gradient decay rate of the current frame is determined based on the healthy pure target anchor point, and the target bounding box area change rate of the current frame is cross-validated with the edge gradient decay rate to obtain the optical vignetting judgment result. When the optical vignetting judgment result indicates that the sensor's optical vignetting overflows, a nonlinear exponential penalty is applied to the single-frame occlusion degree of the current frame to obtain an updated occlusion degree, and the single-frame occlusion degree of the current frame is assigned as the updated occlusion degree. Local multi-peak dynamic clustering is performed on the local background region of the current frame, and the number of effective peaks is determined in combination with the absolute cold background anchor point. If the number of effective peaks is greater than a limit value, a higher-order amplification multiplier is generated based on the number of effective peaks. The comprehensive confusion of the current frame is determined based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The overall complexity of the current frame is determined based on the single-frame occlusion degree and the overall confusion degree of the current frame, and target tracking defense is performed based on the overall complexity.

2. The method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios according to claim 1, characterized in that, The construction of the time-adaptive memory window involves monitoring the infrared image sequence through its state machine. When abrupt changes occur in the infrared image sequence, the time-adaptive memory window is frozen. From the frozen time-adaptive memory window, absolute cold background anchor points and healthy pure target anchor points are extracted, including: A fixed-length sliding cache queue based on first-in-first-out is constructed in memory as the time-adaptive memory window; Real-time monitoring of the grayscale extreme values ​​of the local background region of the current frame, the energy centroid offset rate, the inter-frame reduction of the structural similarity index, or the disorder of the global motion vector distribution; If the grayscale extreme value experiences a nonlinear surge, or if the energy centroid offset rate experiences centroid tearing, or if the inter-frame decrease of the structural similarity index exceeds the decrease threshold, or if the disorder of the global motion vector distribution exceeds the disorder threshold, the temporal adaptive memory window is frozen. The absolute cold background anchor and the healthy pure target anchor are extracted from the frozen temporal adaptive memory window using median filtering.

3. The method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios according to claim 1, characterized in that, The target region pixels of the current frame are the valid pixels within the target ground truth box of the current frame; determining the saturation ratio of the target region pixels based on the total number of target region pixels in the current frame of the infrared image sequence, and assigning the single-frame occlusion degree of the current frame as the limit single-frame occlusion degree when the saturation ratio is greater than a preset high-risk threshold, includes: Extract the total number of valid pixels in the target ground truth box of the current frame; Based on the preset detector overexposure grayscale threshold, the total number of overexposed pixels within the target ground truth box is determined, and the ratio of the total number of overexposed pixels to the total number of effective pixels is calculated to obtain the saturation ratio of the pixels in the target area. If the saturation ratio is greater than the preset high-risk threshold, the single-frame occlusion degree of the current frame is assigned the limit single-frame occlusion degree.

4. The method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios according to claim 1, characterized in that, The edge gradient decay rate of the current frame is determined based on the healthy pure target anchor point, and the target bounding box area change rate of the current frame is cross-validated with the edge gradient decay rate to obtain the optical vignetting judgment result. When the optical vignetting determination result indicates sensor optical vignetting overflow, a nonlinear exponential penalty is applied to the single-frame occlusion of the current frame to obtain an updated occlusion, including: Adaptive Gaussian smoothing is performed on the target region pixels of the current frame. An edge extraction operator is used to perform convolution operation on the target region pixels after adaptive Gaussian smoothing to obtain the first-order partial derivative of the target region pixels. The sum of edge gradients of the current frame is determined based on the first-order partial derivative of the target region pixels. The edge gradient decay rate of the current frame is determined based on the sum of the edge gradients of the current frame and the reference gradient of the healthy pure target anchor point. When the edge gradient decay rate is less than a preset retention threshold and the target bounding box area change rate of the current frame is greater than a preset change threshold, the optical vignetting judgment result characterizes the sensor optical vignetting overflow. A nonlinear exponential penalty multiplier factor is constructed based on the edge gradient decay rate of the current frame. A nonlinear exponential penalty is applied to the single-frame occlusion of the current frame based on the nonlinear exponential penalty multiplier factor to obtain the updated occlusion.

5. The method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios according to claim 1, characterized in that, The step of performing local multi-peak dynamic clustering on the local background region of the current frame, and determining the number of effective peaks in combination with the absolute cold background anchor points, includes: The local background region of the current frame is flattened into a one-dimensional vector, and extreme pulse noise filtering is performed on the one-dimensional vector to obtain a cleaned one-dimensional vector. A clustering model with a preset number of groups is initialized in parallel. The cleaned one-dimensional vector is dynamically clustered to obtain each group of clustering models after convergence. The best-fitting model is determined from each group of clustering models after convergence according to the model selection criteria. Valid peaks with pixel weights greater than a preset lower limit and mean values ​​higher than the absolute cold background anchor points are selected from the best-fit model, and the number of valid peaks is counted.

6. The method for anti-distortion feature reconstruction and target tracking defense in high-dynamic infrared countermeasure scenarios according to claim 5, characterized in that, The clustering model employs a Gaussian mixture model, a density-based clustering algorithm, or a mean-shift algorithm. When the clustering model adopts the Gaussian mixture model, the step of dynamically clustering the cleaned one-dimensional vector to obtain converged clustering models, and determining the best-fitting model from the converged clustering models according to the model selection criteria, includes: The cleaned one-dimensional vector is dynamically clustered using the expectation-maximization algorithm to obtain converged clustering models. The best-fit model is then determined from the converged clustering models based on the Bayesian information criterion.

7. The method for anti-distortion feature reconstruction and target tracking defense in high dynamic infrared countermeasure scenarios according to claim 1, characterized in that, The step of determining the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame includes: The single-frame occlusion degree of the current frame and the overall confusion degree of the current frame are fused using the harmonic mean paradigm to obtain the overall complexity of the current frame.

8. A high-dynamic infrared anti-distortion feature reconstruction and target tracking defense device for high-dynamic infrared countermeasure scenarios, characterized in that, include: An anchor point extraction module is used to construct a time-adaptive memory window, monitor the infrared image sequence through the state machine of the time-adaptive memory window, freeze the time-adaptive memory window when the infrared image sequence shows abrupt changes, and extract absolute cold background anchor points and healthy pure target anchor points from the frozen time-adaptive memory window. The occlusion assignment module is used to determine the saturation ratio of the target region pixels based on the total number of target region pixels in the current frame of the infrared image sequence, and to assign the single-frame occlusion of the current frame to the limit single-frame occlusion when the saturation ratio is greater than a preset high-risk threshold. The occlusion update module is used to determine the edge gradient decay rate of the current frame based on the healthy pure target anchor point when the saturation ratio is not greater than the preset high-risk threshold, and to perform cross-validation based on the target bounding box area change rate of the current frame and the edge gradient decay rate to obtain the optical vignetting judgment result. When the optical vignetting judgment result indicates that the sensor's optical vignetting overflows, a nonlinear exponential penalty is applied to the single-frame occlusion degree of the current frame to obtain an updated occlusion degree, and the single-frame occlusion degree of the current frame is assigned as the updated occlusion degree. The confusion determination module is used to perform local multi-peak dynamic clustering on the local background region of the current frame, determine the number of effective peaks in combination with the absolute cold background anchor point, generate a higher-order amplification multiplier based on the number of effective peaks when the number of effective peaks is greater than a limit value, and determine the comprehensive confusion of the current frame based on the higher-order amplification multiplier and the basic single-frame confusion of the current frame. The complexity determination module is used to determine the overall complexity of the current frame based on the single-frame occlusion degree and the overall confusion degree of the current frame, and to perform target tracking defense based on the overall complexity.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the steps of the high dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the high-dynamic infrared countermeasure scene anti-distortion feature reconstruction and target tracking defense method according to any one of claims 1-7.

Images