A high-speed aircraft terminal lightweight target identification method and system

Abstract

This invention relates to the field of target recognition technology, specifically providing a lightweight target recognition method and system for high-speed aircraft terminals. The method includes: extracting a first infrared image sequence when target occlusion is detected; calculating aerosol static distribution index and aerosol dynamic distribution index by analyzing the first infrared image sequence; fusing the two into an aerosol spatial information index; selecting a first target image sequence using a genetic algorithm; calculating optimized parameters for an infrared imaging system by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; obtaining a second target image sequence using the optimized parameters; and tracking and identifying the target based on the second target image sequence. This invention reduces computational load while maintaining target recognition accuracy, enabling efficient target recognition by a high-speed aircraft terminal in dusty aerosol environments.

Description

Technical Field

[0001] This invention relates to the field of target recognition technology, specifically a lightweight target recognition method and system for high-speed aircraft terminals. Background Technology

[0002] Infrared imaging systems, as an important target identification method for modern high-speed aircraft terminals, are characterized by fast identification and high accuracy. However, in desert environments, the abundant aerosol effect can absorb and scatter the infrared radiation of targets, thus interfering with the detection path of the target identification system and severely affecting its detection capabilities. Existing infrared imaging target identification systems typically require complex image enhancement algorithms and multi-frame fusion techniques to cope with sand and dust aerosol interference. These processing methods often involve large-scale computations, increasing the terminal load. For high-speed aircraft terminals with severely limited resources, this high-load operating mode severely restricts the parallel execution of other tasks, affecting the overall system performance.

[0003] Therefore, there is an urgent need for a lightweight target identification method and system for high-speed aircraft terminals that can effectively suppress aerosol interference while maintaining the accuracy of target identification. Summary of the Invention

[0004] (1) Technical problems to be solved

[0005] The purpose of this invention is to provide a lightweight target recognition method and system for high-speed aircraft terminals, in order to solve the problem that existing infrared imaging target recognition systems suffer from increased terminal load and reduced target recognition accuracy when facing sand and dust aerosol interference. This is due to the use of complex image enhancement algorithms and multi-frame fusion technology in traditional high-speed aircraft terminals.

[0006] (2) Technical solution

[0007] To achieve the above objectives, in one aspect, the present invention provides a lightweight target identification method for high-speed aircraft terminals, the method comprising:

[0008] When the high-speed aircraft terminal determines that the target is blocked, it extracts the first infrared image sequence; the first infrared image sequence is processed into grayscale and the binarized edge image of each frame is obtained through the edge detection algorithm; the infrared radiation absorption blocking area and the infrared radiation release blocking area are extracted based on the binarized edge image.

[0009] The region grayscale contrast and edge gradient magnitude are calculated using the infrared radiation absorption and emission blocking regions. An aerosol static distribution index is calculated based on the region grayscale contrast and edge gradient magnitude. An aerosol dynamic distribution index is obtained by analyzing consecutive frames in the first infrared image sequence. An aerosol spatial information index is calculated based on the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence. Finally, an aerosol spatial information index is calculated by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence.

[0010] The aerosol spatial information index is used to obtain the first target image sequence through a genetic algorithm.

[0011] The infrared imaging system optimization parameters are calculated by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; the infrared imaging system is optimized according to the infrared imaging system optimization parameters, and a second infrared image sequence is extracted; the second infrared image sequence is processed through the same steps to obtain a second target image sequence; the high-speed aircraft terminal performs target tracking and identification based on the second target image sequence.

[0012] Furthermore, the method for extracting the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image includes:

[0013] The binarized edge image is segmented to obtain the initial occlusion region; the texture features of the initial occlusion region are extracted; the initial infrared radiation occlusion region is identified based on the texture features; the initial infrared radiation occlusion regions of adjacent frames in the first infrared image sequence are obtained; the gray-scale mean of the initial infrared radiation occlusion regions in the adjacent frames is calculated by analyzing the initial infrared radiation occlusion regions in the adjacent frames; the difference between the gray-scale mean values ​​of the initial infrared radiation occlusion regions in the adjacent frames is used to calculate the region stability; the initial infrared radiation occlusion region is divided into the final infrared radiation occlusion region based on the region stability.

[0014] The initial infrared radiation value of the target identified by the high-speed aircraft is obtained and converted into a grayscale threshold for the target image. When the average grayscale value of the final infrared radiation blocking area is greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for releasing infrared radiation blocking. When the average grayscale value of the infrared radiation blocking area is not greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for absorbing infrared radiation blocking.

[0015] Furthermore, the method for calculating the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption blocking region and the infrared radiation emission blocking region includes:

[0016] The maximum and minimum grayscale values ​​of the infrared radiation absorption and emission areas are obtained, and the grayscale contrast of the areas is calculated. The edge information of the infrared radiation absorption and emission areas is obtained through an edge detection algorithm. The edge gradient magnitude is obtained by feature extraction based on the edge information.

[0017] Further, the method for calculating the aerosol static distribution index based on the region's grayscale contrast and edge gradient magnitude; obtaining the aerosol dynamic distribution index by analyzing consecutive frames in the first infrared image sequence; and calculating the aerosol spatial information index based on the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence includes:

[0018] The static distribution index of aerosols is obtained by normalizing the grayscale contrast and edge gradient magnitude and then calculating the geometric mean.

[0019] Obtain grayscale images of motion-compensated consecutive frames from the first infrared image sequence region, and calculate the optical flow motion vector based on the grayscale images of the motion-compensated consecutive frames. The optical flow motion vector The calculation formula is:

[0020] ;

[0021] in, For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image.

[0022] Obtain the optical flow motion vector set of the infrared radiation blocking region; calculate the aerosol dynamic distribution index by analyzing the optical flow motion vector set.

[0023] Furthermore, the method for calculating the aerosol dynamic distribution index by analyzing the optical flow motion vector set includes:

[0024] Obtain the optical flow motion vector of the pixel within the final infrared radiation blocking area. ;in, The speed at which a pixel moves in the horizontal direction; The vertical velocity of a pixel is given by the optical flow vector of the pixel within the final infrared radiation occlusion region; the average velocity of the region is calculated from this vector using the optical flow vector of the pixel within the final infrared radiation occlusion region. ;in, The number of pixels within the final infrared radiation blocking area; based on the optical flow vector of the pixels within the final infrared radiation blocking area. The average motion vector of the pixels within the final infrared radiation blocking area is obtained through calculation. The motion consistency of the final infrared radiation blocking area is calculated by analyzing the average motion vector of pixels within the final infrared radiation blocking area. ; Pixels within the final infrared radiation blocking area Optical flow motion vector The regional divergence of the final infrared radiation blocking area was calculated. The average speed of the final infrared radiation blocking area Motion consistency and regional divergence The aerosol dynamic distribution index was obtained after normalization.

[0025] Further, the method of calculating the aerosol spatial information index by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence, and obtaining the first target image sequence by using the aerosol spatial information index through a genetic algorithm, includes:

[0026] Number of images in the first infrared image sequence Randomly generated A starting frame index is used as the initial population; the starting frame index is... A processing scheme consisting of consecutive frames of images; generating an initial population for the processing scheme, the initial population including... individual ; Evaluate individuals in the initial population sequentially The corresponding aerosol spatial information index under the processing scheme is calculated by weighting the aerosol static distribution index and the aerosol dynamic distribution index; wherein, the aerosol static distribution index is calculated by the processing scheme corresponding to the processing scheme. The aerosol static distribution index is obtained by averaging the values ​​of the aerosol static distribution index from consecutive frames of images; the aerosol dynamic distribution index is obtained by calculation. The average value of the aerosol dynamic distribution index in consecutive frames of images; and the calculation of all individual values. The aerosol spatial information index.

[0027] Ranked by aerosol spatial information index and individuals with the lowest aerosol spatial information index were removed. The winning individuals are selected using a roulette wheel selection method, and the winning individuals are then crossovered and mutated to obtain offspring individuals. The crossover operation performs arithmetic crossover between the starting frame indices of the two winning individuals, while the mutation operation performs a small random perturbation on the starting frame index of the winning individuals; the initial population is then replaced with offspring individuals. The process continues iteratively until the improvement in the aerosol spatial information index of both the offspring and parent individuals falls below a set fitness threshold. The final winning individual obtained after this iterative evolution is then processed according to the given rules. A series of consecutive images are used as the first target image sequence.

[0028] Furthermore, the method for calculating the optimized parameters of the infrared imaging system by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first target image sequence includes:

[0029] By analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence, the corresponding response surface model coefficients are determined from the working condition section coefficient mapping table. According to the coefficients of the response surface model The frame integral time is obtained by calculating the extreme points of the response surface model. and frame sampling time The calculation formula for the response surface model is as follows:

[0030] ;

[0031] in, It is the spatial information index of aerosols.

[0032] Furthermore, the method for constructing the working condition section coefficient mapping table includes:

[0033] The aerosol static distribution index and aerosol dynamic distribution index were divided into multiple operating condition segments. For each operating condition segment, calibration sample data was collected during the frame integration time and frame sampling time through system experiments. A set of response surface model coefficients for each operating condition segment was obtained by analyzing the calibration sample data. Based on a set of response surface model coefficients for each operating condition segment. Generate a coefficient mapping table for operating conditions.

[0034] Based on the same inventive concept, the present invention also provides a lightweight target identification system for high-speed aircraft terminals, the system comprising:

[0035] The infrared radiation occlusion region extraction module is used to extract a first infrared image sequence when the high-speed aircraft terminal determines that the target is occluded; to process the first infrared image sequence into grayscale and to obtain a binarized edge image of each frame image through an edge detection algorithm; and to extract the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image.

[0036] The aerosol spatial information index calculation module is used to calculate the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption and emission shielding regions; calculate the aerosol static distribution index based on the regional grayscale contrast and edge gradient magnitude; obtain the aerosol dynamic distribution index by analyzing consecutive frames in the first infrared image sequence; calculate the aerosol spatial information index based on the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence; and calculate the aerosol spatial information index by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence.

[0037] The target image sequence filtering module is used to obtain the first target image sequence from the aerosol spatial information index through a genetic algorithm.

[0038] The target tracking and recognition optimization module is used to calculate the infrared imaging system optimization parameters by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; optimize the infrared imaging system according to the infrared imaging system optimization parameters, and extract the second infrared image sequence; obtain the second target image sequence by processing the second infrared image sequence through the same processing steps; and the high-speed aircraft terminal performs target tracking and recognition according to the second target image sequence.

[0039] (3) Beneficial effects

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. Lightweight feature extraction and image filtering replace traditional algorithms. This reduces computational load while maintaining target recognition accuracy, enabling high-speed aircraft to quickly complete target recognition under resource-constrained conditions.

[0042] 2. By dynamically adjusting the frame integration time and frame sampling time of the infrared imaging system, the system can automatically adapt to aerosol interference environments to obtain clearer images, effectively improving the accuracy and reliability of target recognition. Attached Figure Description

[0043] Figure 1 This is a flowchart of a lightweight target identification method for a high-speed aircraft terminal according to Embodiment 1 of the present invention;

[0044] Figure 2 This is a schematic diagram of the module composition of a lightweight target recognition system for a high-speed aircraft terminal according to Embodiment 2 of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0046] Before giving examples, it is necessary to describe the application scenario of the present invention, which is applied to the problem of target identification of high-speed aircraft in complex aerosol environments.

[0047] Example 1: As Figure 1 As shown in the figure, this embodiment provides a lightweight target identification method for high-speed aircraft terminals, the method comprising:

[0048] When the high-speed aircraft terminal determines that the target is blocked, it extracts the first infrared image sequence; the first infrared image sequence is processed into grayscale and a binary edge image of each frame is obtained through an edge detection algorithm; the infrared radiation absorption blocking area and the infrared radiation release blocking area are extracted based on the binary edge image.

[0049] The region grayscale contrast and edge gradient magnitude are calculated using the infrared radiation absorption and emission blocking regions. An aerosol static distribution index is calculated based on the region grayscale contrast and edge gradient magnitude. An aerosol dynamic distribution index is obtained by analyzing consecutive frames in the first infrared image sequence. An aerosol spatial information index is calculated based on the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence. Finally, an aerosol spatial information index is calculated by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence.

[0050] The aerosol spatial information index is used to obtain the first target image sequence through a genetic algorithm.

[0051] The infrared imaging system optimization parameters are calculated by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; the infrared imaging system is optimized according to the infrared imaging system optimization parameters, and a second infrared image sequence is extracted; the second infrared image sequence is processed through the same steps to obtain a second target image sequence; the high-speed aircraft terminal performs target tracking and identification based on the second target image sequence.

[0052] It should be noted that the same processing steps for obtaining the second target image sequence from the second infrared image sequence are: generating an aerosol spatial information index based on the second infrared image sequence, and then using a genetic algorithm to select the second target image sequence.

[0053] For example, when a high-speed aircraft flies in a desert environment, the target (such as a vehicle) is obscured by sand aerosols. This results in the vehicle's thermal radiation edges being partially covered by sand aerosols in the infrared image, leading to a blurred outline after edge detection and hindering effective target localization and tracking. The high-speed aircraft continuously acquires initial images of the vehicle obscured by sand aerosols using its infrared imaging system, obtaining a first infrared image sequence. By analyzing this first infrared image sequence, the high-speed aircraft terminal calculates the static and dynamic distribution indices of the sand aerosols and fuses them to generate an aerosol spatial information index. Based on the aerosol spatial information index, the high-speed aircraft terminal optimizes the infrared imaging parameters using a genetic algorithm to obtain a first target image sequence with reduced sand aerosol interference. This further guides the infrared imaging system to perform real-time parameter adjustments, ultimately obtaining a second infrared image sequence with a clear target outline and significantly suppressed sand aerosol interference. The high-speed aircraft terminal can extract the vehicle's contour and feature points based on the second target image sequence, and stably estimate and predict its trajectory using optical flow and Kalman filtering based on the vehicle's contour and feature points to achieve continuous tracking of the target.

[0054] The method for extracting the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image includes:

[0055] The binarized edge image is segmented to obtain the initial occlusion region; the texture features of the initial occlusion region are extracted; the initial infrared radiation occlusion region is identified based on the texture features; the initial infrared radiation occlusion regions of adjacent frames in the first infrared image sequence are obtained; the gray-scale mean of the initial infrared radiation occlusion regions in the adjacent frames is calculated by analyzing the initial infrared radiation occlusion regions in the adjacent frames; the difference between the gray-scale mean values ​​of the initial infrared radiation occlusion regions in the adjacent frames is used to calculate the region stability; the initial infrared radiation occlusion region is divided into the final infrared radiation occlusion region based on the region stability.

[0056] The initial infrared radiation value of the target identified by the high-speed aircraft is obtained and converted into a grayscale threshold for the target image. When the average grayscale value of the final infrared radiation blocking area is greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for releasing infrared radiation blocking. When the average grayscale value of the infrared radiation blocking area is not greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for absorbing infrared radiation blocking.

[0057] It should be noted that in desert environments, sand aerosols are more disordered than other background environmental elements. For example, large areas of sand in the desert environment exhibit certain texture features in infrared images. Therefore, the initial occlusion area is segmented based on the texture features of sand aerosols. However, other relatively unique background environments in the desert also have complex texture features, such as rough rock surfaces, cracked soil, and shrubs; for example, sandy areas that are slowly cooling after being exposed to sunlight during hot weather in the desert will also exhibit complex texture features in infrared images; therefore, further segmentation of these areas is necessary. These areas are generally more stable than sand aerosols, so the grayscale value changes between adjacent frames are extracted to reflect the stability of the regions, thereby achieving the purpose of further segmenting the infrared radiation occlusion areas.

[0058] It should also be noted that, when there is no obstruction, the infrared imaging system locks onto the image region where the target is located and simultaneously calculates the initial infrared radiation value of the image region. Based on the initial infrared radiation value, the average grayscale value of the target image is obtained through infrared imaging system parameters and infrared radiation path compensation.

[0059] For example, a high-speed aircraft stably tracks a vehicle in a desert environment. In the initial phase of the mission, when the vehicle is not obscured by sand aerosols, the average grayscale value of the vehicle's main image is calculated and set as the initial target grayscale threshold of 150. When encountering sand aerosols, the high-speed aircraft terminal initiates a processing flow: first, morphological closing operations are performed on the binarized edge image of the first frame to connect broken edges, and noise points with an area less than 50 pixels are removed through connected component analysis, initially obtaining candidate regions such as sand aerosols, vehicles, and rocks. Next, based on the initial classification of texture features: the entropy value of each region is calculated, and an entropy threshold of 0.7 is set as the discrimination criterion. The entropy value of the sand aerosol region is 0.8 (0.8 > 0.7), so it is classified as an initial infrared radiation obscured region; the entropy value of the rock region is 0.4 (0.4 ≤ 0.7), so it is classified as a background region and removed. Subsequently, based on the final determination of regional stability: in three consecutive frames of images, the grayscale change values ​​of the above candidate regions are calculated, and a regional stability threshold of 10 is set. The grayscale change value of the dust aerosol region is 15 (15>10), so it is finally determined as the final infrared radiation blocking region; the rock region is 3 (3≤10), so it is determined as a stable background region and is removed. Finally, the average grayscale value of the final infrared radiation blocking region is compared with the target initial grayscale threshold of 150. When the average grayscale value of a region is 170 (170>150), it is determined as an infrared radiation releasing blocking region; when the average grayscale value of a region is 120 (120≤150), it is determined as an infrared radiation absorbing blocking region.

[0060] The method for calculating the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption and infrared radiation emission areas includes:

[0061] The maximum and minimum grayscale values ​​of the infrared radiation absorption and emission areas are obtained, and the grayscale contrast of the areas is calculated. The edge information of the infrared radiation absorption and emission areas is obtained through an edge detection algorithm. The edge gradient magnitude is obtained by feature extraction based on the edge information.

[0062] For example, for the infrared radiation absorption blocking area, the maximum gray value is obtained by statistically analyzing its gray-level histogram. Minimum grayscale value The formula for calculating regional grayscale contrast is adopted. Calculate the grayscale contrast of the region The value is 0.2. This relatively high value reflects the non-uniform absorption and scattering of infrared radiation by dust aerosols as discrete particulate media. For areas that block infrared radiation (such as vehicle engines), the maximum gray value is obtained by statistically analyzing their gray-level histogram. Minimum grayscale value The formula for calculating regional grayscale contrast is adopted. The calculated regional grayscale contrast is approximately 0.143. This low value reflects the stable and consistent radiation intensity within a uniform solid heat source. Subsequently, [further details are needed]. Sobel The edge detection operator was used to process the edge gradient amplitude of the infrared radiation absorption and blocking area, which was 30, which is consistent with the fuzzy boundary characteristics of sand and dust aerosols. The edge gradient amplitude of the infrared radiation release and blocking area was 60, which characterized the sharp contour features of the solid object.

[0063] The method for calculating the aerosol static distribution index based on the region's grayscale contrast and edge gradient magnitude; obtaining the aerosol dynamic distribution index by analyzing consecutive frames in the first infrared image sequence; and calculating the aerosol spatial information index based on the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence includes:

[0064] The static distribution index of aerosols is obtained by normalizing the grayscale contrast and edge gradient magnitude and then calculating the geometric mean.

[0065] For example, to accurately quantify aerosol characteristics, this scheme uses the geometric mean method to construct the aerosol static distribution index, the calculation formula of which is: The method can automatically suppress regions that cannot simultaneously satisfy both internal inhomogeneity and boundary diffusion, thus effectively eliminating interference from most single features. For example, for a dust aerosol region with a gray-level contrast of 0.2 and an edge gradient amplitude of 30, the calculated aerosol static distribution index is approximately 0.374; for a vehicle engine region with a gray-level contrast of 0.143 and an edge gradient amplitude of 60, the calculated aerosol static distribution index is approximately 0.239. The index for dust aerosols is higher than that for vehicle engines. This result clearly indicates that the method can effectively capture the complex physical characteristics of aerosols, providing a basis for accurate target identification and stable tracking.

[0066] Obtain grayscale images of motion-compensated consecutive frames from the first infrared image sequence region, and calculate the optical flow motion vector based on the grayscale images of the motion-compensated consecutive frames. The optical flow motion vector The calculation formula is:

[0067] ;

[0068] in, For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image.

[0069] It should be noted that changes in the flight attitude of a high-speed aircraft cause pixel displacement in the entire image background. If the optical flow vector is directly calculated on this image, the resulting optical flow vector has a low signal-to-noise ratio and cannot be used for analysis. Therefore, motion compensation is necessary first. Specifically, this involves first estimating the transformation matrix representing global motion by analyzing consecutive frames, and then mapping the previous frame to the coordinate system of the current frame, thus obtaining motion-compensated consecutive frame images. In this invention, the optical flow method is used to quantify the dynamic physical characteristics of the dust aerosol interference field. Although the transient characteristics of dust aerosols do not conform to the ideal assumptions of optical flow calculation, the macroscopic laws exhibited by its motion vector field at the statistical level constitute the key characteristics that distinguish it from other motion interferences. By analyzing the statistical characteristics of the optical flow vector set, the system can effectively quantify the dynamic behavior of dust aerosols.

[0070] For example, the gray-level curvature matrix is ​​the image function within the local neighborhood of a pixel. Matrix. Both the gray-level curvature matrix and the gray-level gradient vector can be obtained through... It is obtained by calculating using the isodifferential operator. The local neighborhood of a pixel in a frame image is ;No. The local neighborhood of a pixel in a frame image is ;use The algorithm calculates for and for ;use Algorithm calculation for and for Then, the optical flow motion vector is obtained through the formula for calculating the optical flow motion vector. That is, the optical flow motion vector is zero, indicating that no significant relative motion occurs in this region between frames. The local neighborhood of a pixel in a frame image is ;No. The local neighborhood of a pixel in a frame image is ;use The algorithm calculates for and for ;use The algorithm calculates for and for Then, the optical flow motion vector is obtained through the formula for calculating the optical flow motion vector. This result indicates the existence of a local motion of approximately -0.45 pixels / frame along the negative x-axis, and such motion features can be used to identify the drift of dust aerosols.

[0071] Obtain the optical flow motion vector set of the infrared radiation blocking region; calculate the aerosol dynamic distribution index by analyzing the optical flow motion vector set.

[0072] The method for calculating the aerosol dynamic distribution index by analyzing the optical flow motion vector set includes:

[0073] Obtain the optical flow motion vector of the pixel within the final infrared radiation blocking area. ;in, The speed at which a pixel moves in the horizontal direction; The vertical velocity of a pixel is given by the optical flow vector of the pixel within the final infrared radiation occlusion region; the average velocity of the region is calculated from this vector using the optical flow vector of the pixel within the final infrared radiation occlusion region. ;in, The number of pixels within the final infrared radiation blocking area; based on the optical flow vector of the pixels within the final infrared radiation blocking area. The average motion vector of the pixels within the final infrared radiation blocking area is obtained through calculation. The motion consistency of the final infrared radiation blocking area is calculated by analyzing the average motion vector of pixels within the final infrared radiation blocking area. ; Pixels within the final infrared radiation blocking area Optical flow motion vector The regional divergence of the final infrared radiation blocking area was calculated. The average speed of the final infrared radiation blocking area Motion consistency and regional divergence The aerosol dynamic distribution index was obtained after normalization.

[0074] For example, the optical flow vector of the dust aerosol region The values ​​are: (1.0, 0.5), (0.8, 0.7), (1.2, 0.3), (0.9, 0.6). The average speed is calculated. Approximately 1.125, motion consistency The regional divergence is approximately 0.983. It is 0.25. For comparison, the optical flow motion vector in the vehicle region... Given the values ​​(2.0, 0.1), (2.0, 0.1), (2.0, 0.0), and (2.0, 0.0), the average velocity is calculated using the same process. The motion consistency is 2.001. Approximately 0.999, regional divergence The value is 0.002. The comparison results indicate that the regional dispersity of dust aerosols is... Vehicle area divergence above rigid motion This difference effectively reveals the unique dynamic and divergent nature of dust aerosols as a dispersion medium, while the divergence in the vehicle area approaches zero due to near-pure translational motion. Through this quantitative comparison, dust aerosol interference and targets can be reliably distinguished based on aerosol dynamic characteristics such as regional divergence.

[0075] The method of calculating the aerosol spatial information index by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence, and obtaining the first target image sequence by using the aerosol spatial information index through a genetic algorithm, includes:

[0076] Number of images in the first infrared image sequence Randomly generated A starting frame index is used as the initial population; the starting frame index is... A processing scheme consisting of consecutive frames of images; generating an initial population for the processing scheme, the initial population including... individual ; Evaluate individuals in the initial population sequentially The corresponding aerosol spatial information index under the processing scheme is calculated by weighting the aerosol static distribution index and the aerosol dynamic distribution index; wherein, the aerosol static distribution index is calculated by the processing scheme corresponding to the processing scheme. The aerosol static distribution index is obtained by averaging the values ​​of the aerosol static distribution index from consecutive frames of images; the aerosol dynamic distribution index is obtained by calculation. The average value of the aerosol dynamic distribution index in consecutive frames of images; and the calculation of all individual values. aerosol spatial information index;

[0077] Ranked by aerosol spatial information index and individuals with the lowest aerosol spatial information index were removed. The winning individuals are selected using a roulette wheel selection method, and the winning individuals are then crossovered and mutated to obtain offspring individuals. The crossover operation performs arithmetic crossover between the starting frame indices of the two winning individuals, while the mutation operation performs a small random perturbation on the starting frame index of the winning individuals; the initial population is then replaced with offspring individuals. The process continues iteratively until the improvement in the aerosol spatial information index of both the offspring and parent individuals falls below a set fitness threshold; the final winning individual obtained after the iterative evolution is then processed according to the corresponding treatment plan. A series of consecutive images are used as the first target image sequence.

[0078] For example, suppose an infrared imaging system captures a sequence of 100 infrared images. The system needs to automatically select the 5 consecutive images with the least aerosol interference as the first target image sequence. To this end, the infrared imaging system initializes a population containing 6 candidate schemes, each scheme representing 5 consecutive images with a starting frame index; for example, the scheme with starting frame 10 corresponds to frames 10 to 14. The formula for calculating the aerosol spatial information index is: ;in, This represents the average aerosol static distribution index of 5 consecutive frames corresponding to the scheme. This is the average value of the aerosol dynamic distribution index of 5 consecutive images corresponding to the scheme. Weights for the static information index of aerosols; As the weight of the aerosol dynamic information index; and This example It is 0.6. The initial value is 0.4. In the first-generation evaluation, the system calculated the aerosol spatial information index for each scheme, finding that the scheme with a starting frame of 30 had the highest index of 0.88, while the scheme with a starting frame of 45 had the lowest index of 0.58. The system sorted by fitness, first eliminating the individual with the lowest fitness (starting frame 45). Then, using a roulette wheel selection method (selection probability proportional to fitness), four individuals were chosen from the remaining individuals as parents, and the individual with the highest fitness in the current generation (starting frame 30) was forcibly retained as an elite to directly enter the offspring generation. Next, crossover and mutation operations were performed on the selected parents. The four selected parents were randomly paired, and arithmetic crossover was performed with a crossover probability of 0.8. The starting frame indices of the two parents were arithmetically weighted and rounded to obtain the starting frame index of the offspring individual. For example, crossing individuals with starting frames 30 and 25 generated a new scheme with a starting frame of 27. Mutate offspring generated by crossover and parent individuals not selected for crossover (excluding elites) with a mutation probability of 0.1. Apply a mutation probability of 0.1 to the individual's starting frame index. A uniform random integer perturbation within a 3-frame range is applied. For example, an individual starting at frame 75 is mutated to generate a new scheme starting at frame 73. If the starting frame index generated by the genetic operation does not meet the boundary (because 5 consecutive images are required, with a maximum starting frame of 96), it is corrected to the nearest valid boundary value; that is, if the starting frame index is greater than 96, it is corrected to 96, and if the starting frame index is less than 1, it is corrected to 1. After the above operations are completed, elite individuals are merged with the new individuals generated through crossover and mutation to form a progeny population containing 6 new schemes. In subsequent iterations, the system repeats the evolutionary process of evaluation, selection, crossover, and mutation, and sets a maximum of 100 iterations as one of the termination conditions. After five generations of evolution, the index of the optimal scheme increases from 0.88 to 0.93, and the improvement in the optimal fitness of the fifth generation population relative to the fourth generation is 0, which is lower than the preset fitness threshold of 0.01, thus meeting the termination condition and ending the evolution. Finally, the optimal solution obtained during the output evolution process, namely the continuous image sequence corresponding to the starting frame 31, is used as the first target image sequence. The first target image sequence has the best overall performance in terms of aerosol static distribution and dynamic characteristics, providing a target image sequence with the least interference and the highest quality for subsequent processing.

[0079] The method for calculating the optimized parameters of the infrared imaging system by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first target image sequence includes:

[0080] By analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence, the corresponding response surface model coefficients are determined from the working condition section coefficient mapping table. According to the coefficients of the response surface model The frame integral time is obtained by calculating the extreme points of the response surface model. and frame sampling time The calculation formula for the response surface model is as follows:

[0081] ;

[0082] in, It is the spatial information index of aerosols.

[0083] It should be noted that the aerosol static distribution index and aerosol dynamic distribution index are chosen as independent inputs in this step, rather than directly using the fused aerosol spatial information index, because of the physical correlation between these two parameters and the imaging parameters to be optimized. The static distribution index characterizes the spatial texture complexity of aerosols, and its value directly affects the image signal-to-noise ratio (SNR), thus being strongly correlated with the optimization decision of the frame integration time. For example, a higher aerosol static distribution index usually requires extending the integration time to improve the SNR, while a lower aerosol static distribution index allows for a shorter integration time to avoid saturation. The aerosol dynamic distribution index quantifies the intensity of aerosol temporal motion, and its value determines the optimization direction of the frame sampling time. That is, a higher aerosol dynamic distribution index requires precise selection of the frame sampling time to extract rapidly changing transients, while a lower aerosol dynamic distribution index is insensitive to the frame sampling time. If a comprehensive aerosol spatial information index is directly used as a macroscopic evaluation index, it is impossible to distinguish the different physical mechanisms mentioned above, resulting in the inability to provide a targeted adjustment basis for two imaging parameters with different physical meanings. Therefore, decoupling and using the aerosol static distribution index and the aerosol dynamic distribution index in this key step is the choice to achieve the transition from scene perception to parameter control.

[0084] For example, firstly, the aerosol static distribution index and aerosol dynamic distribution index are calculated based on the obtained first target image sequence. Then, based on the two indices, it is determined that the current operating condition is in a medium-interference zone. Finally, the response surface model coefficients of the medium-interference zone are obtained from a preset mapping table. Input the coefficients of the response surface model into the response surface model, and solve the system of equations composed of the partial derivatives of the response surface model. and The spatial information index of aerosols was calculated. The optimal solution that maximizes this is the frame integral time. Frame sampling time Integrate the frame time Frame sampling time As an optimization parameter for the infrared imaging system, the second infrared image sequence acquired after adjusting the infrared imaging system accordingly was verified to have an aerosol spatial information index. The image quality was improved by increasing the value from 0.82 to 0.91.

[0085] The method for constructing the working condition section coefficient mapping table includes:

[0086] The aerosol static distribution index and aerosol dynamic distribution index were divided into multiple operating condition segments. For each operating condition segment, calibration sample data was collected during the frame integration time and frame sampling time through system experiments. A set of response surface model coefficients for each operating condition segment was obtained by analyzing the calibration sample data. Based on a set of response surface model coefficients for each operating condition segment. Generate a coefficient mapping table for operating conditions.

[0087] For example, firstly, through the analysis of a large amount of experimental data, based on the numerical distribution characteristics of the aerosol static distribution index and dynamic distribution index, the continuous aerosol state space is divided into several typical operating condition segments with clear boundaries. For example, it can be divided into three segments: slight disturbance, moderate disturbance, and severe disturbance, to achieve a reasonable mapping from actual operating conditions to a finite standard mode. Subsequently, for each defined operating condition segment, in a laboratory or controlled typical environment, a systematic experimental design method is used to systematically collect infrared image sequences covering different parameter combinations within the full parameter space of frame integration time and frame sampling time, as calibration sample data, to ensure the comprehensiveness and representativeness of the data. Next, for the calibration sample dataset collected under each operating condition segment, with frame integration time and frame sampling time as independent variables and the calculated aerosol spatial information index as the dependent variable, a set of response surface model coefficients that can accurately describe the parameter performance relationship within the operating condition segment is independently fitted using a multiple regression analysis method. Finally, a coefficient mapping table for each operating condition segment is generated by associating all operating condition segments with their corresponding unique set of model coefficients and pre-loaded into the memory of the high-speed aircraft terminal. Through this complete offline modeling and solidification process, the terminal can quickly obtain the optimal imaging parameter optimization scheme based on the real-time calculated aerosol index by simply performing a table lookup operation during the online identification phase. This ensures extreme processing efficiency while enabling the infrared imaging system to intelligently adapt to complex aerosol environments.

[0088] Example 2: Based on the same inventive concept, such as Figure 2 As shown, this embodiment also provides a lightweight target recognition system for high-speed aircraft terminals, including:

[0089] The infrared radiation occlusion region extraction module is used to extract a first infrared image sequence when the high-speed aircraft terminal determines that the target is occluded; to process the first infrared image sequence into grayscale and to obtain a binarized edge image of each frame image through an edge detection algorithm; and to extract the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image.

[0090] The aerosol spatial information index calculation module is used to calculate the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption and emission blocking areas; calculate the aerosol static distribution index based on the regional grayscale contrast and edge gradient magnitude; obtain the aerosol dynamic distribution index by analyzing consecutive frames in the first infrared image sequence; calculate the aerosol spatial information index based on the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence; and calculate the aerosol spatial information index by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence.

[0091] The target image sequence filtering module is used to obtain the first target image sequence from the aerosol spatial information index through a genetic algorithm;

[0092] The target tracking and recognition optimization module is used to calculate the infrared imaging system optimization parameters by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; optimize the infrared imaging system according to the infrared imaging system optimization parameters, and extract the second infrared image sequence; obtain the second target image sequence by processing the second infrared image sequence through the same processing steps; and the high-speed aircraft terminal performs target tracking and recognition according to the second target image sequence.

[0093] It should be noted that the specific ways in which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0094] Finally, it should be noted that although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lightweight target identification method for a high-speed aircraft terminal, characterized in that, The method includes: When the high-speed aircraft terminal determines that the target is blocked, it extracts the first infrared image sequence; the first infrared image sequence is processed into grayscale and a binary edge image of each frame is obtained through an edge detection algorithm; the infrared radiation absorption blocking area and the infrared radiation release blocking area are extracted based on the binary edge image. The grayscale contrast and edge gradient magnitude of the regions are calculated using the regions that absorb and emit infrared radiation. The static distribution index of aerosols is calculated based on the grayscale contrast and edge gradient magnitude of the regions. The dynamic distribution index of aerosols is obtained by analyzing consecutive frames in the first infrared image sequence. The spatial information index of aerosols is calculated based on the static distribution index and the dynamic distribution index of aerosols in the first infrared image sequence. The aerosol spatial information index is used to obtain the first target image sequence through a genetic algorithm; The infrared imaging system optimization parameters are calculated by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; the infrared imaging system is optimized according to the infrared imaging system optimization parameters, and a second infrared image sequence is extracted; the second infrared image sequence is processed through the same steps to obtain a second target image sequence; the high-speed aircraft terminal performs target tracking and identification based on the second target image sequence.

2. The lightweight target identification method for a high-speed aircraft terminal according to claim 1, characterized in that, The method for extracting the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image includes: The binarized edge image is segmented to obtain the initial occlusion region; the initial occlusion region is extracted to obtain its texture features; the initial infrared radiation occlusion region is identified based on its texture features; the initial infrared radiation occlusion regions of adjacent frames in the first infrared image sequence are obtained; the gray-scale mean of the initial infrared radiation occlusion regions in the adjacent frames is calculated by analyzing them; the difference between the gray-scale mean values ​​of the initial infrared radiation occlusion regions in the adjacent frames is used to calculate the region stability; and the initial infrared radiation occlusion region is divided into the final infrared radiation occlusion region based on the region stability. The initial infrared radiation value of the target identified by the high-speed aircraft is obtained and converted into a grayscale threshold for the target image. When the average grayscale value of the final infrared radiation blocking area is greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for releasing infrared radiation blocking. When the average grayscale value of the infrared radiation blocking area is not greater than the grayscale threshold of the target image, the final infrared radiation blocking area is used as a region for absorbing infrared radiation blocking.

3. The lightweight target identification method for a high-speed aircraft terminal according to claim 2, characterized in that, The method for calculating the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption and infrared radiation emission areas includes: The maximum and minimum gray values ​​of the infrared radiation absorption and emission blocking regions are obtained, and the region gray-scale contrast is calculated. The region edge information is obtained by using an edge detection algorithm on the infrared radiation absorption and emission blocking regions. The edge gradient magnitude is obtained by feature extraction of the region edge information.

4. The lightweight target identification method for a high-speed aircraft terminal according to claim 3, characterized in that, The aerosol static distribution index is calculated based on the region's grayscale contrast and edge gradient magnitude; the aerosol dynamic distribution index is obtained by analyzing consecutive frames in the first infrared image sequence. The method for calculating the aerosol spatial information index based on the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence includes: The static distribution index of aerosols is obtained by normalizing the grayscale contrast and edge gradient magnitude and then calculating the geometric mean. Obtain grayscale images of motion-compensated consecutive frames from the first infrared image sequence region, and calculate the optical flow vector based on the grayscale images of the motion-compensated consecutive frames. The optical flow motion vector The calculation formula is: ； in, For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level curvature matrix of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image; For the first The gray-level gradient vector of a local neighborhood of a pixel in a frame image; Obtain the optical flow motion vector set of the infrared radiation blocking region; calculate the aerosol dynamic distribution index by analyzing the optical flow motion vector set.

5. The lightweight target identification method for a high-speed aircraft terminal according to claim 4, characterized in that, The method for calculating the aerosol dynamic distribution index by analyzing the optical flow motion vector set includes: Obtain the optical flow motion vector of the pixel within the final infrared radiation blocking area. ;in, The speed at which a pixel moves in the horizontal direction; The vertical velocity of a pixel is given by the optical flow vector of the pixel within the final infrared radiation occlusion region; the average velocity of the region is calculated from this vector using the optical flow vector of the pixel within the final infrared radiation occlusion region. ;in, The number of pixels within the final infrared radiation blocking area; based on the optical flow vector of the pixels within the final infrared radiation blocking area. The average motion vector of the pixels within the final infrared radiation blocking area is obtained through calculation. The motion consistency of the final infrared radiation blocking area is calculated by analyzing the average motion vector of pixels within the final infrared radiation blocking area. ; Pixels within the final infrared radiation blocking area Optical flow motion vector The regional divergence of the final infrared radiation blocking area was calculated. The average speed of the final infrared radiation blocking area Motion consistency and regional divergence The aerosol dynamic distribution index was obtained after normalization.

6. The lightweight target identification method for a high-speed aircraft terminal according to claim 5, characterized in that, The method of calculating the aerosol spatial information index by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first infrared image sequence, and obtaining the first target image sequence by using the aerosol spatial information index through a genetic algorithm, includes: Number of images in the first infrared image sequence Randomly generated A starting frame index is used as the initial population; the starting frame index is... A processing scheme consisting of consecutive frames of images; generating an initial population for the processing scheme, the initial population including... individual ; Evaluate individuals in the initial population sequentially The corresponding aerosol spatial information index under the processing scheme is calculated by weighting the aerosol static distribution index and the aerosol dynamic distribution index; wherein, the aerosol static distribution index is calculated by the processing scheme corresponding to the processing scheme. The aerosol static distribution index is obtained by averaging the values ​​of the aerosol static distribution index from consecutive frames of images; the aerosol dynamic distribution index is obtained by calculation. The aerosol dynamic distribution index is obtained by averaging the values ​​of the aerosol dynamic distribution index in consecutive frames of images; and the values ​​of all individuals are calculated. aerosol spatial information index; Ranked by aerosol spatial information index and individuals with the lowest aerosol spatial information index were removed. The winning individuals are selected using a roulette wheel selection method, and the winning individuals are then crossovered and mutated to obtain offspring individuals. The crossover operation performs arithmetic crossover between the starting frame indices of the two winning individuals, while the mutation operation performs a small random perturbation on the starting frame index of the winning individuals; the initial population is then replaced with offspring individuals. The process continues iteratively until the improvement in the aerosol spatial information index of both the offspring and parent individuals falls below a set fitness threshold; the final winning individual obtained after the iterative evolution is then processed according to the corresponding treatment plan. A series of consecutive images are used as the first target image sequence.

7. The lightweight target identification method for a high-speed aircraft terminal according to claim 6, characterized in that, The method for calculating the optimized parameters of the infrared imaging system by analyzing the aerosol static distribution index and the aerosol dynamic distribution index of the first target image sequence includes: By analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence, the corresponding response surface model coefficients are determined from the working condition section coefficient mapping table. According to the coefficients of the response surface model The frame integral time is obtained by calculating the extreme points of the response surface model. and frame sampling time The calculation formula for the response surface model is as follows: ； in, It is the spatial information index of aerosols.

8. The lightweight target identification method for a high-speed aircraft terminal according to claim 7, characterized in that, The method for constructing the working condition section coefficient mapping table includes: The aerosol static distribution index and aerosol dynamic distribution index were divided into multiple operating condition segments. For each operating condition segment, calibration sample data was collected during the frame integration time and frame sampling time through system experiments. A set of response surface model coefficients for each operating condition segment was obtained by analyzing the calibration sample data. Based on a set of response surface model coefficients for each operating condition segment. Generate a working condition section coefficient mapping table.

9. A lightweight target identification system for a high-speed aircraft terminal, characterized in that, The system includes: The infrared radiation occlusion region extraction module is used to extract a first infrared image sequence when the high-speed aircraft terminal determines that the target is occluded; to process the first infrared image sequence into grayscale and to obtain a binarized edge image of each frame image through an edge detection algorithm; and to extract the infrared radiation absorption occlusion region and the infrared radiation release occlusion region based on the binarized edge image. The aerosol spatial information index calculation module is used to calculate the regional grayscale contrast and edge gradient magnitude through the infrared radiation absorption and emission blocking areas; calculate the aerosol static distribution index based on the regional grayscale contrast and edge gradient magnitude; obtain the aerosol dynamic distribution index by analyzing consecutive frames in the first infrared image sequence; calculate the aerosol spatial information index based on the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence; and calculate the aerosol spatial information index by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first infrared image sequence. The target image sequence filtering module is used to obtain the first target image sequence from the aerosol spatial information index through a genetic algorithm; The target tracking and recognition optimization module is used to calculate the infrared imaging system optimization parameters by analyzing the aerosol static distribution index and aerosol dynamic distribution index of the first target image sequence; optimize the infrared imaging system according to the infrared imaging system optimization parameters, and extract the second infrared image sequence; obtain the second target image sequence by processing the second infrared image sequence through the same processing steps; and the high-speed aircraft terminal performs target tracking and recognition according to the second target image sequence.

Images