A noise-based digital camouflage generation method

Abstract

The application provides a noise-based digital camouflage generation method, comprising the following steps: obtaining a scene color set by means of a color extraction algorithm through five environmental images from different directions of a camouflage target; re-customizing and optimizing a genetic algorithm bottom operator aiming at noise generation and superposition process parameter optimization problems; optimizing key parameters of a Perlin noise superposition generation texture process based on a newly defined camouflage comprehensive evaluation factor as fitness; limiting parameter optimization range aiming at different types of camouflage objects; applying the optimized parameters to generate optimal camouflage texture of the scene; digitally generating digital camouflage; and further adjusting the digital camouflage according to national standards to obtain final camouflage. The application greatly improves the generation efficiency of the camouflage and the self-adaptive camouflage effect of various types of camouflage objects in multiple scenes.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of digital camouflage generation, and particularly relates to a digital camouflage generation method based on noise. BACKGROUND

[0002] In modern wars and military activities, camouflage is of great importance, and the design purpose is to minimize the detection and identification ability of the enemy to our armored vehicles in different environments.

[0003] The camouflage design is divided into two parts of background tone extraction and pattern texture design as a whole, and the traditional camouflage design relies on manual color selection against the background, and then relies on experience to draw camouflage patterns in the scene. With the development of computer vision and image processing technology and the improvement of computing power, it is possible to automatically generate camouflage textures highly adapted to specific environments.

[0004] There are many generation methods at the present stage, mainly focusing on style transfer generation camouflage based on deep learning, specific scene camouflage based on adversarial neural networks, image recognition and environment clustering, etc. However, these existing researches and related methods still have certain defects, such as: low texture generation efficiency, long time consumption in early optimization, and being extremely limited by computing resources, poor adaptability to multiple scene environments, and certain defects in the verification and evaluation standards, resulting in that the actual camouflage effect of the generated camouflage is not as expected. The above are problems to be solved in the field of camouflage generation, and are also the starting point of the original method of the present application. SUMMARY

[0005] The purpose of the present application is to provide a digital camouflage generation method based on noise, which improves the method of generating multiple Perlin noise textures by a customized improved genetic algorithm, redesigns and optimizes the genetic algorithm bottom operators, especially the probability calculation of each operator, proposes a new multi-dimensional texture evaluation factor as the fitness, and improves the interpolation algorithm in the texture generation framework, so that the texture generation parameter optimization efficiency is improved, the texture generation effect after optimization is also improved, and better camouflage texture can be stably output, which can better integrate into the background environment and significantly improve the camouflage effect compared with other algorithms.

[0006] The embodiments of the present application can be implemented as follows:

[0007] The present application provides a digital camouflage generation method based on noise, which is used for generating camouflage digital camouflage suitable for a target scene according to the target scene, and the generation method comprises the following steps:

[0008] In the case of non-rain and snow weather and sufficient natural light, environment images are shot in the front, back, left, right and down directions of the target scene;

[0009] extracting a final main color set member under the target scene from the five environment images, for color distribution planning of background and main spots, decoration spots, etc. in the camouflage texture;

[0010] Selecting multiple Perlin noise superposition to generate camouflage texture and optimizing the internal algorithm, further filtering the parameters involved in the noise generation texture algorithm process, obtaining the process key parameters to be optimized;

[0011] According to the parameter optimization problem in the noise generation texture process, the bottom operator of the genetic algorithm is reconstructed, and a more comprehensive camouflage evaluation factor is designed as the fitness of the genetic algorithm, which is used for the optimization process of the process key parameters, wherein the optimization direction is limited for different types of camouflage objects to adapt to the camouflage needs of the camouflage object in the current target scene background, and the environment background is integrated as much as possible;

[0012] The optimized parameters are stored locally with the environment images, and are evaluated and compared with the locally stored images every time a new image is obtained in the target scene to determine whether to update the parameters or directly use the parameters to generate a deformed camouflage, and then a digital camouflage texture module is generated;

[0013] According to the camouflage camouflage national standard of segmentation attribute, hierarchical attribute, etc. Adjust the layout of the digital camouflage texture module, the relationship between main spots, spot decoration effect, etc.

[0014] Further, in an optional embodiment, the specific process of extracting colors to obtain the final main color set and plan color distribution is:

[0015] The color data in the five environment images is respectively subjected to a clustering algorithm, and the pixel colors with similar colors are aggregated into the same category, and the representative color of the category is output and recorded as a color;

[0016] In this way, the top four colors of each of the environment images are obtained and stored in a color set C, and five color sets C are obtained from five pictures i , i ∈ {1, 2, 3, 4, 5}, c mn represents the color set C obtained from the mth picture m The color occupies the nth most in the color set C

[0017] If for c mn ∈ C i , c mn is put into the final color set G, and the final color set G with x elements is finally obtained, if x < 4, take

[0018]

[0019] The color g n It is also stored in the final color set G, so that the final color set G contains 4 color elements, as the final color set of the target scene;

[0020] The color ranked 1st in the final color set G is used as the main background color of the camouflage texture, the colors ranked 2nd and 3rd are used as optional colors for generating the main spots, and the color ranked 4th is reserved as the color to be used when adding additional decorative spots.

[0021] Furthermore, in an optional embodiment, the specific steps for generating camouflage textures by superimposing multiple Perlin noises after optimizing and adjusting the internal algorithm and then filtering to obtain key process parameters are as follows:

[0022] After the preliminary noise calculation steps, such as grid network definition, gradient vector assignment, distance vector calculation, and dot product calculation, when combining the dot product values ​​of adjacent grid points through an interpolation function, this invention introduces a texture adjustment factor α into conventional bilinear interpolation. The noise calculation for conventional bilinear interpolation is as follows:

[0023] N(x,y)=lerp(u,lerp(v,dot i,j ,dot i+1,j ),lerp(v,dot i,j+1 ,dot i+1,j+1 ))

[0024] The improved interpolation process after introducing a texture adjustment factor is as follows:

[0025]

[0026] Where lerp(t,a,b) represents a linear interpolation function, dot i,j dot i+1,j dot i,j+1 dot i+1,j+1 Let represent the dot products of the gradient vectors of the four vertices (bottom left vertex, bottom right vertex, top left vertex, and top right vertex) and the distance vectors from those vertices to the interpolation point (x, y). These represent the noise values ​​obtained after adjusting the dot product values ​​of each vertex. u and v are the smoothing function values ​​in the x and y directions based on the input point coordinates, respectively. The smoothing function uses an improved 5th-order Hermite smoothing function to ensure better interpolation smoothness.

[0027] By adjusting the value of α, richer and more delicate effects can be created in different texture areas, thereby better simulating the variability and complexity of camouflage textures in nature. α will also be an important parameter to be optimized.

[0028] Then, a complex and diverse texture is generated by superimposing multiple layers of Perlin noise with different frequencies and amplitudes. After the noise superposition is completed, color mapping is performed according to the final color distribution planning of claim 2 to generate the final camouflage texture.

[0029] The final to-be-optimized parameters after screening are: the amplitude decay rate between different frequency layers in the noise texture, i.e., persistence, the texture adjustment factor a in the improved interpolation algorithm, the size of the feature in the noise texture, i.e., the noise texture detail ratio Scale, frequency Frequency, amplitude Amplitude, noise layer coefficient Octaves, lacunarity Lacunarity, offset offset, noise gain gain, threshold Threshold for noise value binarization, maximum range value Maximum limiting noise value, and minimum range value Minimum limiting noise value.

[0030] Further, in an optional embodiment, an improved genetic algorithm is applied to the problem, and the specific content is:

[0031] The fitness of the improved genetic algorithm is a new camouflage evaluation factor, and the calculation method is:

[0032] The histograms of the final camouflage texture image and the environment image are calculated, and the similarity between them is compared using the Bhattacharyya coefficient. The larger this coefficient, the higher the color similarity between the two images;

[0033] GLCM is used to evaluate the texture similarity of the final camouflage texture image and the environment image. GLCM is a statistical method for analyzing image texture features, which describes the texture features of an image by calculating the gray level difference of pixel points in different directions. The GLCM comparison function is used to calculate the similarity between two groups of GLCMs. The larger this value, the higher the texture similarity between the two images;

[0034] Canny edge detection is used to evaluate the confusion effect of the final camouflage texture image and the environment image. Canny edge detection is a commonly used edge detection algorithm that can detect obvious edges in an image. The two images are respectively subjected to Canny edge detection, and then their edge information is compared. The smaller this value, the better the confusion effect of the camouflage texture image and the environment background image on the edge;

[0035] The scores of the three dimensions are averaged to obtain a comprehensive score, and the range is controlled to be between 0 and 1, which is used as a camouflage evaluation factor, i.e., the fitness of the improved genetic algorithm.

[0036] The selection operator of the improved genetic algorithm is improved based on the ordinary roulette selection algorithm, the new evaluation factor is taken as the fitness, and it is found through a large number of experiments that the excellent texture evaluation factor is usually above 0.8, so the individual with the fitness higher than 0.8 is marked as an excellent individual, the probability of the excellent individual is doubled, so that it is easier to be selected and the excellent features are passed down, that is, the probability P of each chromosome being selected is calculated each time i , there are M chromosomes, f i is the fitness of the chromosome, P i can be expressed as

[0037]

[0038] The crossover mode of the crossover operator of the improved genetic algorithm is single-point crossover, the mutation mode of the mutation operator is single-point mutation, and the crossover probability and the mutation probability are dynamically adjusted according to the colony state:

[0039] Crossover probability:

[0040] In the formula, P c is the crossover probability, P c_max is the maximum crossover probability in the current population, f' is the larger fitness value of the two parent individuals in the crossover operation, f max is the maximum fitness value in the entire population, and f min is the minimum fitness value, the accuracy rate is used as the fitness function in the present application, so f [0,1], f max -f min [0,1]

[0041] Mutation probability:

[0042] In the formula, P m is the mutation probability, P m_max is the maximum mutation probability at present, f is the fitness value of the individual, and the others are the same as the crossover probability part above.

[0043] Further, in the optional embodiment, the tuning range limit of different types of camouflage objects is specifically:

[0044] For the personal equipment and uniform camouflage of the relevant personnel, such as uniform, helmet, equipment, weapon, etc., the range limit of the Amplitude, Scale, alpha and other parameters in claim 3 is increased, specifically: Amplitude is between 0.5 and 2.0, Scale is between 10 and 100, and alpha is between 1 and 5;

[0045] For the camouflage of medium and long-range equipment such as armored vehicles, non-armored support vehicles, artillery systems, surface ships, fixed-wing aircraft, etc., the range limits of the Scale, Octaves, Frequency, etc. parameters described in claim 3 are increased, specifically: Scale takes 50 to 500, Octaves takes 4 to 8, and Frequency takes 0.01 to 0.1;

[0046] For the camouflage of fixed military facilities such as command posts, defensive works, logistics facilities, bridges, and important traffic nodes, the range limits of the Scale, Octaves, a, etc. parameters described in claim 3 are increased, specifically: Scale takes 100 to 1000, Octaves takes 6 to 10, and a takes 2 to 6;

[0047] For the camouflage of underwater equipment such as submarines, the range limits of the Scale, Octaves, Persistence, etc. parameters described in claim 3 are increased, specifically: Scale takes 20 to 200, Octaves takes 5 to 9, and Persistence takes 0.4 to 0.7;

[0048] For the camouflage of tactical deception such as decoys and false targets, the range limits of the Octaves, a, etc. parameters described in claim 3 are increased, specifically: Octaves takes 3 to 7, and a takes 3 to 8;

[0049] The above range intervals are all closed intervals.

[0050] Further, in optional embodiments, the parameter updating in actual scenarios and specific applications, the specific steps are:

[0051] The new environment image collected by the target scene is input, the first two evaluation dimensions described in claim 6 are used to perform color similarity evaluation and texture matching evaluation with the stored image of the scene, the average value of the two is taken, and the final result range is 0-1; if there is no stored image, i.e. the first time running in the target scene, the evaluation result is 0;

[0052] When the evaluation result is less than 0.5, the improved genetic algorithm of claim 7 is used to re-optimize and obtain a new parameter set, the new parameter set is used to update the parameter set of the target scene, and the newly provided image of the current environment is used to update the local stored image of the target scene;

[0053] After the evaluation, the parameter set is used to generate a noise texture according to the specific steps in claim 4, to produce a final camouflage pattern texture, i.e., a morphing camouflage pattern, and then a grid of appropriate size is constructed on the basis of the morphing camouflage pattern, and when the pixels of a main pattern color in a certain grid are more than 50%, the grid is set to the main pattern color, and finally a digital camouflage pattern texture is obtained;

[0054] Further, in an optional embodiment, the digital camouflage module is adjusted according to the camouflage pattern related specification standards, and the specific content is as follows:

[0055] According to the segmentation requirement, if the distance between a main spot on a surface of a camouflage object and an edge is less than 10% of the width of the surface perpendicular to the edge, the whole is extended to the edge in the direction perpendicular to the edge, and is rotated clockwise until the whole of the spot pattern is at an angle of 30 degrees with the edge;

[0056] According to the multi-level requirement, for a main spot on a surface of a camouflage object, the area of the maximum circumscribed rectangle thereof is calculated, and if the area is greater than 15% of the area of the surface on which the main spot is located, a periphery modification spot of the same color is added outside the main spot, and an inner package modification spot of a different color is added inside the main spot; for a recessed part of the camouflage object, if the area of the recessed part is greater than 20% of the area of the surface on which the recessed part is located, a main spot is added to the recessed part and a color member with a light bias in the final color set in claim 2 is used; for a protruding part of the camouflage object, if the area of the protruding part is greater than 20% of the area of the surface on which the protruding part is located, a main spot is added to the protruding part and a color member with a dark bias in the final color set in claim 2 is used;

[0057] For an independent main spot on a camouflage object, if there is a continuous rectangular area without any main spot and modification spot in the horizontal direction, the vertical direction and the diagonal direction of the maximum circumscribed rectangle of the main spot, and the area of the blank rectangular area is greater than 80% of the area of the maximum circumscribed rectangle of the main spot, other main spots and corresponding modification spots constituting staggered and juxtaposed positional relationships are added to the blank rectangular area, i.e., the original independent main spot.

[0058] Compared with the prior art, the present application has the following beneficial effects:

[0059] The application selects a genetic algorithm with less parameter quantity requirement, and improves the genetic algorithm from three bottom operators of the genetic algorithm, and when parameters such as Amplitude, scale, octaves, Persistence, Frequenc and a small amount of parameters such as an interpolation texture adjustment factor alpha are selected and optimized quickly, the gradient is more easily approximated to the global optimal noise generation parameter set, the convergence speed is greatly accelerated, and the final generation effect and parameter optimization efficiency after optimization are improved; through comparison experiments of various generation camouflage texture methods, it is found that the noise generation camouflage texture is not only suitable for high complexity environment, but also has better generation effect on fine granularity environment and better comprehensive camouflage effect, so the noise generation texture is specially optimized for camouflage purpose, and a new interpolation scheme is improved on the basis of bilinear interpolation, the texture detail change space is increased compared with ordinary linear interpolation after introducing the new texture adjustment factor alpha, so that the texture is more delicate and changeable, and is more suitable for fine granularity environment camouflage texture generation. Under the condition of reasonable parameter adjustment, it can be well adapted to the regularity of open natural environment, so the bilinear interpolation is used as the interpolation function to improve the ordinary Perlin noise generation texture and the superimposed noise generation texture, and the improved multi-noise superimposed texture generation more suitable for camouflage texture is formed; the application proposes a new texture comprehensive evaluation factor as the fitness of the genetic algorithm, that is, the optimization index, which solves the defects that most of the existing evaluation methods only pay attention to one aspect of color matching, texture similarity and confusion degree. How to accurately evaluate the effectiveness of the generated camouflage is crucial for the process optimization and the final generation texture effect. BRIEF DESCRIPTION OF DRAWINGS

[0060] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of this application and of the related description are used to explain the application and are not intended to limit the application. In the drawings:

[0061] Figure 1 A flowchart of a noise-based digital camouflage generation method provided for an embodiment of the application;

[0062] Figure 2 A flowchart of color extraction provided for an embodiment of the application;

[0063] Figure 3 A process diagram of noise generation texture and optimization of key parameters using improved genetic algorithm provided for an embodiment of the application DETAILED DESCRIPTION

[0064] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and examples. However, the specific examples described herein are only used to explain the present application and do not limit the present application.

[0065] In one embodiment, as shown in Figure 1 Figure 1 is a flowchart of a noise-based digital camouflage generation method provided by an embodiment of the present application, comprising the following steps:

[0066] S101, in the case of non-rain and snow weather and sufficient natural light, the environment images are captured in the front, back, left, right and down directions of the target scene.

[0067] The common classic camouflage target scene includes desert, woodland, snowfield, grassland, gobi and the like. The position with better field of view and distinctive surrounding environment features in the scene is selected as much as possible in the daylight of sunny day with sufficient light. The camera, mobile phone and other conventional shooting devices are used to capture the environment images in the front, back, left, right and down directions respectively at a height of more than one meter from the ground.

[0068] S102, the final main color set in the target scene is extracted from the five environment images, which is used for color distribution planning of the background and main spots, decorative spots and the like in the camouflage texture.

[0069] First, the colors in each environment image are clustered respectively to obtain a plurality of colors, and five color sets are obtained. Then, the colors appearing in the five color sets are found. If there are, the final color set is added. If the final color set does not reach four colors after this step, the high-frequency colors of the five color sets are used to supplement the average value. The final four colors are further planned for specific use in the camouflage.

[0070] S103, the multi-Perlin noise superposition is selected to generate the camouflage texture and optimize the internal algorithm. The process key parameters to be optimized are obtained by further screening the parameters involved in the noise generation texture algorithm process.

[0071] First, the conventional noise calculation steps are performed: grid network definition, gradient vector allocation, distance vector calculation, dot product calculation, then the improved interpolation method with texture adjustment factor is used to perform the final interpolation calculation of noise, and then multi-layer noise superposition is performed. Here, three-layer noise superposition is taken as an example to show the calculation steps:

[0072]

[0073] wherein, N i represents the Perlin noise of the i-th layer.

[0074] ​Finally, color mapping is performed using the final color set S102 to obtain the final camouflage, and the parameters involved in the process, including the texture adjustment factor, are left for tuning.

[0075] S104, according to the parameter tuning problem in the noise generation texture process, the bottom operator of the genetic algorithm is reconstructed, and a more comprehensive camouflage evaluation factor is designed as the fitness of the genetic algorithm for the tuning process of the key parameters of the process, wherein the optimization direction is limited for different types of camouflage objects to adapt to the camouflage requirements of the camouflage objects in the current target scene, so that they are integrated into the environment background.

[0076] An improved genetic algorithm using redesigned selection operator, crossover operator and mutation operator is used to cover the new evaluation factors of color matching degree, texture similarity and confusion degree as fitness to tune the parameters of S103 process.

[0077] The camouflage objects are classified for the tuning process, and the parameters that are more sensitive to each class, i.e. the key parameters for the tuning of the camouflage objects of the class, are selected, and the range is particularly limited to prevent the camouflage texture from not matching the camouflage requirements of the camouflage objects, resulting in poor actual camouflage effect.

[0078] As an example, when the camouflage object is personal equipment and uniform, personal equipment needs to be effectively camouflaged in close range, so it needs highly detailed and variable texture, so the following parameters are particularly limited in range:

[0079] Amplitude needs to be low in contrast to avoid being too conspicuous in close range, taking 0.5 to 2.0; Scale adjusts the detail level and size of the pattern to adapt to the size of the personal equipment, taking 10 to 100; Alpha (texture adjustment factor in the interpolation stage) is used to create fine texture to effectively break the human body contour in close range, taking 1 to 5.

[0080] As an example, when the camouflage object is an armored vehicle, an artillery system, a surface ship, a fixed-wing aircraft, etc. Medium and long-range equipment, the demand is more focused on improving the confusion effect, integrating with the background environment, and reducing the possibility of being discovered by long-range observers, so the following parameters are particularly limited in range:

[0081] Scale needs to be medium scale to adapt to the observation of medium and long range, taking 50 to 500; Octaves need to increase the number of noise layers to increase the complexity and detail of the texture, taking 4 to 8; Frequency adjustment to match the environmental texture characteristics under medium and long distance, taking 0.01 to 0.1.

[0082] S105. The optimized parameters are stored locally together with the environmental image. Each time a new image is acquired in the target scene, the parameters are evaluated and compared with the locally stored image to determine whether to update the parameters or directly use the parameters to generate a deformed camouflage. Then, a digital camouflage texture is generated digitally.

[0083] After each generation process is completed, the initial image and the optimized parameters will be automatically determined to be saved locally and to overwrite the locally saved version, so as to facilitate the generation of camouflage for the scene when no new image is provided.

[0084] S106. Adjust the layout, main spot relationships, and spot modification effects of the digital camouflage texture module according to camouflage-related specifications and standards such as segmentation attributes and hierarchical attributes.

[0085] The adjustments to the segmentation requirements mainly aim to prevent the overall orientation, shape, size, and positional distribution of the generated digital camouflage modules from exhibiting regularity, and to address the insufficient standardization in the handling of the boundaries between camouflage object surfaces. The former is primarily addressed by adding or modifying smaller spots on the main spots of the digital camouflage modules and adjusting the position and orientation of the main spots. The latter mainly targets main spots near or touching the edge, extending them to pass through the boundary and intersect it obliquely. Specifically, in implementation, when the distance between a main spot and the boundary is less than 10% of the horizontal or vertical length spanned by the main spot, the main spot is extended to the adjacent surface, and the overall orientation and the angle between it and the boundary are controlled to be as close to 30 degrees as possible.

[0086] The adjustments to the multi-layered requirements mainly involve adding decorative spots and special treatment of large areas of unevenness on the camouflaged object. The former primarily involves adding inner spots inside larger main spots and outer spots around their perimeter. The inner spots are selected from the three final colors and are a lighter shade than the main spot, while the outer spots are the same color as the main spot. The latter mainly involves adding main spots to large areas of depression and protrusion on the camouflaged object. The depressions are selected from the three final colors and the protrusions from the three final colors.

[0087] In an optional embodiment, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the color extraction process provided in an embodiment of the present invention. This embodiment relates to how to extract the main color set of a scene using multi-angle environmental images. Based on the above embodiment, it includes the following steps:

[0088] S201. Five environmental images are clustered by color to obtain five color sets. Each color set retains four color elements based on their proportion within the image.

[0089] The color of all the pixels appearing in each environment image is relatively large, and the color is clustered to obtain several colors representing the image. The representative color mapped in each image color clustering is consistent.

[0090] As an example, the colors a, b, and c clustered together in the environment image 1 and the colors d, e, f, and g clustered together in the environment image 2 are relatively similar colors, and the representative colors mapped after clustering are the same.

[0091] Then, each color set is sorted according to the proportion of the color in the image before clustering from large to small, and the first four color members are retained.

[0092] S202, the color appearing in the five color sets is selected and put into the final color set.

[0093] The color located in the five color sets at the same time is selected, that is, the color occupying the top four in the five environment image color clustering.

[0094] S203, the high-frequency color of the five color sets is used for neutralization, and the supplementary color is calculated to obtain the final color set member number of 4.

[0095] If the five color sets are completely the same, that is, four colors are selected and put into the final color set, no supplementary color needs to be calculated. Otherwise, the supplementary color is obtained by neutralizing the high-frequency color, such as S202, two colors are selected and put into the final color set, that is, two colors exist in the five color sets at the same time, then two supplementary colors need to be calculated, the first supplementary color is obtained by neutralizing and averaging the color ranked first in the five color sets and put into the final color set, and the second supplementary color is obtained by neutralizing and averaging the color ranked second in the five color sets and put into the final color set, so that the final color set has four color members.

[0096] S204, the four color set members are planned to be used in specific camouflage according to the order of entering.

[0097] As an example, assuming that the four color members are color x, y, z, and w in order, color x is the background color when generating camouflage, colors y and z are the main spot colors, and color w is used for subsequent additional modification spots. In special cases such as internal package spots, camouflage object concave-convex position spots, etc., colors y, z, and w may be set as main spot colors or modification spot colors according to the relative brightness.

[0098] In an optional embodiment, as shown in Figure 3 , the four color members are color x, y, z, and w in order, color x is the background color when generating camouflage, colors y and z are the main spot colors, and color w is used for subsequent additional modification spots. In special cases such as internal package spots, camouflage object concave-convex position spots, etc., colors y, z, and w may be set as main spot colors or modification spot colors according to the relative brightness. Figure 3is a process schematic diagram of noise generation texture provided by the embodiment of the present application and using improved genetic algorithm to optimize key parameters. The embodiment relates to how to use the improved genetic algorithm to optimize the noise generation texture process, especially how to optimize the key parameters of the process. On the basis of the above embodiment, the following contents are included:

[0099] The whole texture generation process can be divided into Perlin noise calculation, noise superposition and color mapping. There are many key parameters in the first two processes, which constitute the parameter set to be optimized. The process of improving the genetic algorithm to optimize is to regard each noise parameter set as an individual, then let the individual be selected to continuously iterate through the crossover and mutation operations. The excellent parameter individual generated after a certain number of iterations can be used to generate camouflage effect better camouflage texture, and the newly designed texture evaluation factor is the fitness used to evaluate the goodness of the individual in this process. In summary, it is to let the parameter set continuously iterate towards a higher score of the texture evaluation factor.

[0100] For the improved selection operator, the parameter individual with the texture evaluation factor greater than 0.8 is regarded as a better individual in advance, and a greater selection probability is given when the selection probability is calculated, so that the excellent individual is more likely to be selected. For the improved crossover operator and mutation operator, the new probability calculation algorithm makes the individual with a fitness higher than the average level of the population to reduce the crossover and mutation probability as much as possible when the population is relatively dispersed, so as to ensure that the excellent parameter set individual can be reserved into the next generation, so that the final result is better.

[0101] As an example, suppose that there are currently many parameter set individuals in the colony, wherein the parameters of parameter set individual 1 are Scale = 50.0, Amplitude = 3.5, Octaves = 3, the fitness, i.e., the texture evaluation factor score of the generated camouflage texture, is 0.81, the parameters of parameter set individual 2 are Scale = 100.0, Amplitude = 2.0, Octaves = 2, the fitness is 0.75, and the parameters of parameter set individual 3 are Scale = 40.0, Amplitude = 2.0, Octaves = 2, the fitness is 0.68, then parameter set individual 1 and individual 2 are more likely to be selected than individual 3, and the probability of individual 1 being selected is much greater than that of individual 2 and individual 3, at this time, if individual 1 and individual 2 are selected, the next time the crossover operation is calculated, on the one hand, the overall fitness of the two individuals is relatively large, so there is a positive addition to the probability of occurrence, but at the same time, it will depend on the overall fitness of the current colony, if the overall colony is dispersed, in order to preserve the relatively excellent camouflage generation effect of the parameter set of individual 1, the crossover operation may not occur; if the overall colony is concentrated, the two are more likely to occur crossover operation, and a new parameter set individual such as Scale = 50.0, Amplitude = 2.0, Octaves = 3, whose fitness is 0.84, may be generated, that is, a better parameter set individual is generated in this process.

[0102] The application provides a noise-based digital camouflage generation method, and the above description is only a preferred embodiment of the application, and it should be pointed out that, for ordinary skilled persons in the technical field, some improvements and refinements can be made without departing from the principles of the application, and these improvements and refinements should also be regarded as the protection scope of the application. The components not explicitly described in the embodiment can be implemented by using the prior art.

Claims

1. A noise-based digital camouflage generation method, used to generate a camouflage digital pattern suitable for a target scene, characterized in that, Includes the following steps: Step 1: In non-rainy or snowy weather and with sufficient natural light, take environmental images from five directions (front, back, left, right, and down) at the shooting point in the target scene. Step 2: Extract the final main color set of the target scene from the environmental images captured from the five directions, and use it for color distribution planning of the background, main spots, and modification spots in the camouflage texture; Step 3: Use multiple Perlin noise superposition to generate camouflage textures and optimize the internal algorithm. Further screen the parameters involved in the noise texture generation algorithm process to obtain the key parameters to be optimized. Specific steps include: After the preliminary steps of grid network definition, gradient vector assignment, distance vector calculation, dot product calculation, and noise calculation, when combining the dot product values ​​of adjacent grid points through an interpolation function, this invention introduces a texture adjustment factor α into conventional bilinear interpolation. The noise calculation for conventional bilinear interpolation is as follows: N(x,y)=lerp(u,lerp(v,dot i，j ,dot i+1,j ),lerp(v,dot i，j+1 ,dot i+1,j+1 )) The improved interpolation process after introducing the texture adjustment factor α is as follows: Where lerp(t, a, b) represents a linear interpolation function, dot i，j dot i+1，j dot i，j+1 dot i+1，j+1 Let represent the dot products of the gradient vectors of the four vertices (bottom left vertex, bottom right vertex, top left vertex, and top right vertex) and the distance vectors from those vertices to the interpolation point (x, y), respectively. These represent the noise values ​​obtained after adjusting the dot product values ​​of each vertex. u and v are the smoothing function values ​​in the x and y directions based on the input point coordinates, respectively. The smoothing function adopts the improved 5th order Hermite smoothing function to ensure better interpolation smoothness. By adjusting the value of α, richer and more delicate effects can be created in different texture areas, thereby better simulating the variability and complexity of camouflage textures in nature. α will also be an important parameter to be optimized. The final parameters to be tuned are: the amplitude attenuation rate between different frequency layers in the noise texture (Persistence), the texture adjustment factor α in the improved interpolation algorithm, the size of the features in the noise texture (Scale), frequency, amplitude, noise layer coefficients (Octaves), gap degree (Lacunarity), offset, noise gain, threshold for noise value binarization, maximum value for limiting noise values, and minimum value for limiting noise values. Step 4: Reconstruct the underlying operators of the genetic algorithm based on the parameter tuning problem in the noise-generated texture process, and design a more comprehensive camouflage evaluation factor as the fitness of the genetic algorithm for the tuning of key parameters in the process. The optimization direction is specifically restricted for different types of camouflage objects to adapt to the camouflage requirements of the camouflage objects in the current target scene background, so that they can blend into the environmental background. Step 5: Store the optimized parameters locally along with the environmental image. When a new image is acquired in the target scene, evaluate and compare it with the locally stored image to determine whether to update the parameters or directly use the parameters to generate a deformed camouflage. Then, digitally generate a digital camouflage texture module. Step 6: Adjust the layout, main spot relationships, and spot modification effects of the digital camouflage texture module according to the national standard for camouflage based on the segmentation and hierarchical attributes.

2. The noise-based digital camouflage generation method according to claim 1, characterized in that, The specific process of extracting colors to obtain the final master color set and planning the color distribution in step 2 is as follows: Clustering algorithms are applied to the color data in the environmental images captured from the five directions to group pixels with similar colors into the same category, and the representative color of the category is output and recorded as a color. By analogy, the first four colors of each environmental image are obtained and stored in color set C, resulting in five color sets C from the five images. q q∈{1,2,3,4,5},c mn C represents the color set obtained from the m-th image. m The color with the highest percentage in the middle; If for c mn ∈C q Then c mn Place them into the final color set G, and you will get the final color set G containing x elements. If x < 4, take... The color g n It is also stored in the final color set G, so that the final color set G contains 4 color elements, as the final color set of the target scene; The color ranked 1st in the final color set G is used as the main background color of the camouflage texture, the colors ranked 2nd and 3rd are used as optional colors for generating the main spots, and the color ranked 4th is reserved as the color to be used when adding additional decorative spots.

3. The noise-based digital camouflage generation method according to claim 2, characterized in that, Step 3 also includes: Then, by superimposing multiple layers of Perlin noise with different frequencies and amplitudes, complex and diverse textures are generated. After the noise superposition is completed, color mapping is performed according to the final color distribution plan in step 2 to generate the final camouflage texture.

4. The noise-based digital camouflage generation method according to claim 1, characterized in that, The improved genetic algorithm in step 4 is as follows: The fitness of the improved genetic algorithm is a new camouflage evaluation factor, calculated as follows: Calculate the histograms of the final camouflage texture image and the environment image, and compare their similarity using the Bhattacharyya coefficient. The larger the Bhattacharyya coefficient, the higher the color similarity between the two images. GLCM is used to evaluate the texture similarity between the final camouflage texture image and the environment image. GLCM is a statistical method for analyzing image texture features. It describes the texture features of an image by calculating the gray level difference of pixels in different directions. The GLCM comparison function is used to calculate the similarity between two sets of GLCM. The larger this value is, the higher the similarity between the two images in terms of texture. Canny edge detection is used to evaluate the confusion effect between the final camouflage texture image and the environment image. Canny edge detection is a commonly used edge detection algorithm that can detect obvious edges in an image. Canny edge detection is performed on the two images respectively, and then their edge information is compared. The smaller the value, the better the confusion effect between the camouflage texture image and the environment background image on the edge. The average score of the scores in the three dimensions is used to obtain the comprehensive score, which is controlled within the range of 0 to 1. This comprehensive score is used as the camouflage evaluation factor, which is the fitness of the improved genetic algorithm. The improved genetic algorithm's selection operator is based on the ordinary roulette wheel selection algorithm, using a new camouflage evaluation factor as fitness. Extensive experiments have shown that individuals with excellent camouflage evaluation factors typically have a fitness factor above 0.

8. Therefore, individuals with a fitness higher than 0.8 are marked as superior individuals, doubling their probability of selection and passing on superior traits. This involves calculating the probability P of each chromosome being selected each time. e There are M chromosomes in total, f e P represents the fitness of this chromosome. e Represented as The improved genetic algorithm uses a single-point crossover operator and a single-point mutation operator, with both crossover and mutation probabilities dynamically adjusted based on the community state. Crossover probability: In the formula P c Let P be the crossover probability. c_max f is the maximum crossover probability in the current population, and f′ is the larger fitness value among the two parent individuals in the crossover operation. max f is the maximum fitness value in the entire population. min To find the minimum fitness value, and since the fitness function in this application is accuracy, f∈[0,1], f max -f min ∈[0,1]; Mutation probability: In the formula P m Let P be the mutation probability. m_max is the current maximum mutation probability, f is the fitness value of the individual, and the rest is the same as the crossover probability part above.

5. The noise-based digital camouflage generation method according to claim 1, characterized in that, The specific details of the optimization range restrictions for different types of masquerading objects in step 4 are as follows: For the uniforms, helmets, carrying equipment, weapons, personal equipment, and uniform camouflage of relevant personnel, the range restrictions of the parameters Amplitude, Noise Texture Detail Scale, and Texture Adjustment Factor α mentioned in step 3 are added, specifically: Amplitude is between 0.5 and 2.0, Noise Texture Detail Scale is between 10 and 100, and Texture Adjustment Factor α is between 1 and 5. For camouflage of armored vehicles, unarmored support vehicles, artillery systems, surface ships, and fixed-wing aircraft at medium and long ranges, the range restrictions for the noise texture detail scale, noise layer coefficient Octaves, and frequency parameters mentioned in step 3 are added. Specifically, the noise texture detail scale is between 50 and 500, the noise layer coefficient Octaves is between 4 and 8, and the frequency is between 0.01 and 0.

1. For the camouflage of fixed military facilities such as command posts, fortifications, logistics facilities, bridges and important transportation nodes, the range restrictions of the noise texture detail scale, noise layer coefficient Octaves and texture adjustment factor α parameters mentioned in step 3 are added. Specifically, the noise texture detail scale is between 100 and 1000, the noise layer coefficient Octaves is between 6 and 10, and the texture adjustment factor α is between 2 and 6. For submarine underwater equipment camouflage, the range restrictions of the noise texture detail scale, noise layer coefficient Octaves, and persistence parameters mentioned in step 3 are added. Specifically, the noise texture detail scale is between 20 and 200, the noise layer coefficient Octaves is between 5 and 9, and the persistence is between 0.4 and 0.

7. For decoy and false target tactical deception and camouflage, the range restrictions of the noise layer coefficient Octaves and texture adjustment factor α parameters mentioned in step 3 are added, specifically: the noise layer coefficient Octaves is between 3 and 7, and the texture adjustment factor α is between 3 and 8. All the ranges mentioned above are closed intervals.

6. The noise-based digital camouflage generation method according to claim 4, characterized in that, The parameter update and specific application in the actual scenario in step 5 are as follows: The new environmental image acquired from the target scene is input, and the first two evaluation dimensions of the camouflage evaluation factor in step 4 are used to perform color similarity evaluation and texture matching evaluation with the image already stored in the scene. The average of the two is taken, and the final result ranges from 0 to 1. If no image is stored, that is, the first run in the target scene, the evaluation result is returned as 0. When the evaluation result is less than 0.5, the improved genetic algorithm in step 4 is used to re-fine the algorithm to obtain a new parameter set. The new parameter set is used to overwrite and update the parameter set of the target scene, and the newly provided image of the current environment is used to overwrite and update the locally stored image of the target scene. After the evaluation, the parameter set is used to generate noise texture according to the specific steps in step 3 to produce the final camouflage texture, i.e. deformed camouflage texture. Then, a grid of appropriate size is constructed on the deformed camouflage. When the number of pixels of the main pattern color in a grid is more than 50%, the grid is set to the main pattern color to obtain the digital camouflage texture.

7. The noise-based digital camouflage generation method according to claim 2, characterized in that, In step 6, the digital camouflage module is adjusted according to relevant camouflage standards and specifications. The specific details are as follows: According to the segmentation requirements, if the distance between a main spot on a certain surface of the camouflaged object and the edge is less than 10% of the width of the surface perpendicular to the edge, then the entire spot extends through the edge in a direction perpendicular to the edge and rotates clockwise until the overall direction of the spot pattern forms a 30-degree angle with the edge. Based on the multi-layered requirements, for the main spot on a certain face of the camouflage object, calculate the area of ​​its maximum bounding rectangle. If it is greater than 15% of the area of ​​the face where the main spot is located, add an outer decorative spot of the same color to its outside and an inner decorative spot of a different color to its inside. For the recessed part of the camouflage object, if the area exceeds 20% of the area of ​​the face where the recessed part is located, add a main spot to the recessed part and use a brighter color member from the final color set G described in step 2. For the protruding part of the camouflage object, if the area exceeds 20% of the area of ​​the face where the recessed part is located, add a main spot to the protruding part and use a darker color member from the final color set G described in step 2. For a specific main spot on a camouflaged object, if there are continuous rectangular areas without any main spots or decorative spots in the horizontal, vertical, and diagonal directions of the largest bounding rectangle of the main spot, and the area of ​​the blank rectangular area exceeds 80% of the area of ​​the largest bounding rectangle of the main spot, then other main spots and corresponding decorative spots that form an interlaced and juxtaposed positional relationship are added to the blank rectangular area, i.e., next to the original independent main spot.

Images