Strike process simulation method and system

Through the combination of image processing technology and cloud model, the problem of low accuracy in damage effect evaluation in the existing technology is solved, accurate assessment and reasonable decision support for target damage situations are achieved, and the fire strike plan is optimized.

CN119067951AInactive Publication Date: 2024-12-03CHINESE PEOPLES LIBERATION ARMY ARMY ARTILLERY & AIR DEFENSE ACAD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411183859.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, there are problems such as low accuracy in the evaluation of damage effects and unreasonable inference results, especially when dealing with uncertainty and ambiguity, data loss and rule disasters.

Method used

The strike process simulation method based on image processing technology is adopted to obtain the target texture feature change rate and geometric feature change rate by detecting target information, cropping images, inverting color processing, grayscale symbiosis matrix calculation, texture feature analysis and cloud model evaluation. Combined with the pre-set cloud model to evaluate the damage of ground targets, providing more reasonable decision support.

Benefits of technology

It achieves an accurate assessment of the target damage effect, provides strong decision support, optimizes fire strike solutions, and makes the inference results more reasonable, suitable for dealing with ambiguity and randomness, avoiding the loss of data uncertainty and rule disasters.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067951A_ABST
    Figure CN119067951A_ABST
Patent Text Reader

Abstract

The invention provides a striking process simulation method and system. The method comprises the following steps: receiving damage information; cutting the image; overlapping the images; calculating a gray level co-occurrence matrix; target texture features are calculated; calculating a target texture feature change rate; calculating a target geometric feature change rate; and evaluating a target damage effect based on the cloud model. The technical problems that the damage effect evaluation accuracy is low, and the reasoning result is unreasonable are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of visualization of a process of striking a target, and in particular to a method and system for simulating a striking process. Background Art

[0002] The simulation of the striking process is a problem of great concern in many fields, such as military, sports, and game development. Traditional striking process simulation methods mainly rely on physical experiments and theoretical analysis, which is not only costly and time-consuming, but also the accuracy of the results is limited by many factors. However, with the development of computer graphics technology, the striking process simulation method based on graphics has gradually been valued and used.

[0003] Graphics algorithms are a computational method used to process graphics and images. They mainly include graphics generation, transformation, processing, and rendering techniques. Through graphics algorithms, we can create a variety of graphics and perform various complex operations. In the simulation of the striking process, graphics algorithms can be used to generate models of striking objects and simulate various physical phenomena in the striking process, such as collision, rebound, and breakage.

[0004] Strike process simulation is a method of predicting and analyzing the effects of strikes through computer simulation. In this process, we need to consider many factors, such as the shape, mass, speed of the strike object, and the conditions of the strike environment. Through simulation, we can predict the effects of strikes, such as the range, depth, degree of damage, etc., thus providing a scientific basis for decision-making.

[0005] The attack process simulation method based on graphics algorithm is to apply graphics algorithm to the attack process simulation. The main advantage of this method is that it can generate realistic attack effects and improve the authenticity of the simulation. At the same time, due to the powerful functions of graphics algorithm, we can also achieve some complex attack effects, such as explosion, fragmentation, etc.

[0006] In modern science and technology, the development of simulation technology has penetrated into various fields. Especially in the fields of military, aviation, construction, etc., the application of simulation technology has become an indispensable part. Among them, the simulation of the strike process is an important research direction. It can predict and analyze the strike effect through computer simulation and provide a scientific basis for decision-making. The strike process simulation method based on graphics algorithm occupies an important position in this field.

[0007] The fuzzy neural network in the existing technology can effectively deal with nonlinearity, fuzziness and other problems, and plays a huge role in intelligent information processing. The traditional fuzzy neural network algorithm has the learning ability of the traditional neural network and can take into account the fuzziness of data in data mining, but there are problems such as the membership function needs to be determined manually based on expert experience and the inability to handle the randomness of data. In addition, it cannot realize the two-way conversion between quantitative data and qualitative concepts, which easily causes the loss of data uncertainty during data processing and algorithm operation. Cloud theory has great advantages in dealing with uncertainty. It realizes the unified processing of data randomness and fuzziness through three digital features, certainty and uncertainty cloud generator, and ensures the transmission of uncertainty, which is more suitable for the expression of uncertain concepts. The existing cloud neural network model is prone to "rule disaster" when processing large data sets.

[0008] In summary, the existing technology has technical problems such as low accuracy of damage effect assessment and unreasonable reasoning results. Summary of the invention

[0009] The technical problem to be solved by the present invention is to solve the technical problems in the prior art of low accuracy in damage effect evaluation and unreasonable reasoning results.

[0010] The present invention adopts the following technical solutions to solve the above technical problems: a method for simulating a striking process comprises:

[0011] S1. Detecting target information and damage information, wherein the damage information includes: an image before the strike, an image after the strike, a binary image of the changed area, a target mask, and coordinates of a rectangle circumscribing the target in the image;

[0012] S2. According to the coordinates of the circumscribed rectangle of the target in the image, the image before the strike, the image after the strike, the binary image of the changed area and the target mask are respectively cut to obtain the cut image I of the image before the strike. 1 , Image cropping after attack I 2 , Binary image cropping of the changing area I 3 And the target mask clipping image I 4 ;

[0013] S3, cropping image I of the binary image of the changed area 3 Performing color inversion processing and image superposition operations to obtain an inverted color superposition image before striking and an inverted color superposition image after striking;

[0014] S4, obtaining the number of occurrences of specific pixels in the inverted color superposition image before the strike and the inverted color superposition image after the strike, so as to obtain a gray level co-occurrence matrix GLCM;

[0015] S5. According to the gray level co-occurrence matrix GLCM, the recurrence probability of the gray structure is obtained in the inverted color superposition image before the impact and the inverted color superposition image after the impact, so as to describe the target texture characteristics, wherein the target texture characteristics include: entropy value, contrast, correlation and angular second-order moment;

[0016] S6. According to the target texture feature, the texture feature average value is obtained by processing to obtain the target texture feature change rate;

[0017] S7. Calculate the target geometric feature change rate according to the target information;

[0018] S8. Based on the damage level range of the preset cloud model, the damage situation of the ground target is evaluated according to the target texture feature change rate and the target geometric feature change rate, wherein the digital features of the preset cloud model include: expected Ex, entropy En and excess entropy He.

[0019] The present invention uses image processing technology to extract target damage feature information based on the image quality, target type and target characteristics of the acquired ground targets, comprehensively evaluates the target damage effect, provides powerful decision support for specific application scenarios, and optimizes the firepower strike plan. Compared with traditional fuzzy reasoning, the cloud model used in the present invention takes into account the fuzziness and randomness in the decision-making process, and the reasoning results are more reasonable.

[0020] Based on the registration of the images before and after the target is hit, the present invention uses a small amount of target knowledge templates to achieve rapid extraction of key parts of the target. In response to the problems of feature selection and difference measurement in the third type of method, this paper proves through experiments that on the basis of using the grayscale difference method to pre-detect the changed area, the GLCM is used to extract the texture feature vector, and the Mahalanobis distance is used to measure the difference between the feature vectors. This method can quickly and accurately extract the target damage information.

[0021] In a more specific technical solution, step S3 includes:

[0022] S31, cutting image I of binary image of changed area 3 Perform color inversion to obtain an inverted image:

[0023] I 3_new =I 5 -I 3 (1)

[0024] In the formula, I 3_new Represents the inverted image, I 5 A white image in which the color value of each pixel is (255,255,255);

[0025] S32, invert the color of the image I 3_new , respectively, with the image cropped before the attack I1 , Image cropping after attack I 2 Perform the overlay operation to obtain the inverted overlay image before the attack and the inverted overlay image after the attack:

[0026] I 1_new =I 3_new +I 1 (2)

[0027] I 2_new =I 3_new +I 2 (3)

[0028] In the formula, I 1_new ,I 2_new They respectively represent the inverted overlay image before the strike and the inverted overlay image after the strike.

[0029] In a more specific technical solution, step S4 includes:

[0030] S41, take a pixel center point A, take the horizontal direction as the reference line, obtain a specific pixel point B (m, n) at a position with a distance d and an azimuth angle θ from the pixel center point A, and calculate the number of occurrences of the specific pixel point B (m, n):

[0031] P(i,j|d,θ)=num{(k,l)∈I k1×k2 ,(m,n)∈I k1×k2 |d,θ,I(k,l)=i,I(m,n)=j} (4)

[0032] In the formula, i represents the gray level N g In the case of , j represents the pixel grayscale value of point A, j represents the pixel grayscale value of point B, num{·} represents the number of all pixel pairs that meet the above conditions in the area, and m, n represent the pixel coordinates;

[0033] According to the number of occurrences, the initial matrix G is obtained using the following logical processing:

[0034]

[0035] S42. Using the following logic, the initial matrix G is normalized and gray-level co-occurrence is calculated to obtain a gray-level co-occurrence matrix GLCM:

[0036]

[0037] Texture analysis is the extraction and analysis of the spatial distribution pattern of grayscale (shades and shades) in an image. Texture analysis is widely used in the interpretation and processing of remote sensing images, X-ray photographs, and cell images. There is no unified mathematical model for texture. It originated from the concept of texture that characterizes the surface properties of textiles and can be used to describe the arrangement of components of any material, such as lung texture, vascular texture in medical X-ray photographs, and lithological texture in aerospace (or aviation) terrain photographs. Visual texture in image processing is usually understood as a repeated arrangement of a certain basic pattern (tone primitive). Therefore, describing a texture includes determining the tone primitives that make up the texture and determining the relationship between the tone primitives. Texture is a regional characteristic and is therefore related to the size and shape of the region. The boundary between two texture patterns can be determined by observing whether the texture metric has changed significantly. Texture is a reflection of the structure of an object. Analyzing texture can obtain important information about the object in the image and is an important means of image segmentation, feature extraction, and classification and recognition. For spatial domain images or transform domain images (see image transformation), texture analysis can be performed using both statistical and structural methods.

[0038] Statistical texture analysis looks for digital features that characterize textures, and uses these features or other non-texture features to classify regions (rather than individual pixels) in the image. The autocorrelation function, grayscale co-occurrence matrix, grayscale run, and various statistics of grayscale distribution in local areas of the image are commonly used digital texture features. For example, the grayscale co-occurrence matrix uses the spatial distribution of grayscale to characterize texture. Since the grayscale distribution of coarse textures changes much more slowly with distance than that of fine textures, the two have completely different grayscale co-occurrence matrices.

[0039] Structural texture analysis studies the primitives that make up textures and their arrangement rules. A primitive can be the grayscale of a pixel or a connected set of pixels with specific properties. The arrangement rules of primitives are often described using tree grammar.

[0040] The present invention can evaluate the damage effect on the target through the gray-level co-occurrence matrix and cloud model based on the image information obtained by the reconnaissance means in the specific application scenario, accurately evaluate the target damage level, and provide the target damage situation for the command decision-making department in the specific application scenario.

[0041] In a more specific technical solution, step S5 includes:

[0042] S51. Calculate the entropy value using the following logic:

[0043]

[0044] In the formula, C b1_θ represents the first entropy value of the bth image at angle θ;

[0045] S52, using the following logic to calculate the contrast:

[0046]

[0047] In the formula, C b2_θ represents the second contrast of the bth image at angle θ;

[0048] S53. Calculate the correlation using the following logic:

[0049]

[0050] In the formula, C b3_θ represents the third correlation of the bth image at angle θ;

[0051] S54. Calculate the angular second moment using the following logic:

[0052]

[0053] In the formula, C b4_θ It represents the 4th angular second-order moment of the bth image at angle θ.

[0054] In a more specific technical solution, the formula (10) in step S53 is:

[0055]

[0056] In a more specific technical solution, step S6 includes:

[0057] S61, for each inverted color superposition image before the strike and the inverted color superposition image after the strike, take no less than two specific angles, and calculate the average value of the texture feature as the texture feature value:

[0058]

[0059] Where D ba represents the average value of the ath texture feature value of the bth image, where a=1,2,3,4; C ba_0° , C ba_45° , C ba_90° , C ba_135° It represents the ath texture feature value of the bth image when θ takes the values ​​of 0°, 45°, 90°, and 135° respectively;

[0060] S62. Calculate the target texture feature change rate according to the texture feature value:

[0061]

[0062] Where, L a Represents the rate of change of the ath texture feature.

[0063] In a more specific technical solution, in step S7, the target geometric feature change rate is calculated using the following logic:

[0064]

[0065] Where, L 5 Indicates the rate of change of the target geometric features; S change represents the number of white pixels in image I3 in step 2; S mask Represents image I in step 2 4 The number of white pixels.

[0066] In a more specific technical solution, step S8 includes:

[0067] S81. Based on the target characteristic information and expert opinions, the target damage degree is divided into no less than 2 damage level intervals by analyzing the target characteristics;

[0068] S82. Calculate the expectation of the sth damage level interval:

[0069] E x_s =(Z min_s +Z max_s ) / 2 (19)

[0070] Where Ex_s represents the expectation of the sth damage level interval; Z min_s represents the lower limit of the sth damage level interval; Z max_s Indicates the upper boundary value of the sth damage level interval;

[0071] S83. Calculate the entropy of the sth damage level interval using the following logic:

[0072] E n_s =(Z min_s -Z max_s ) / 2.355 (20)

[0073] In the formula, E n_s represents the entropy of the sth damage level interval;

[0074] S84, calculate the oth target texture feature change rate L o , the degree of certainty v in the sth damage level interval os :

[0075]

[0076] S85, using the degree of certainty v os Construct the certainty matrix U o×s,According to the expert opinion information, the weights of the target texture feature change rate and geometric feature change rate are determined to obtain the comprehensive certainty;

[0077] S86. Select the maximum comprehensive certainty from the comprehensive certainty F to obtain the target damage level.

[0078] In a more specific technical solution, in step S85, the comprehensive certainty F is calculated according to the weight:

[0079]

[0080] In the formula, f s represents the certainty at the sth damage level, W 1×5 The weights representing the target texture feature change rate and geometric feature change rate.

[0081] After measuring the distance of the texture feature vector corresponding to each pixel position in the target area of ​​the image before and after the impact, the Mahalanobis distance value of each position is normalized to the range of 0 to 255 to form a distance value image, and then the OTSU threshold segmentation algorithm is used to segment the image into change areas and non-change areas. The segmentation algorithm can be used to obtain the change area of ​​the image after the impact, but it contains some pseudo-change areas. By analyzing and observing the characteristics of the image after the impact, it is found that the crater is generally elliptical or circular with a certain size of area. According to these characteristics, this paper selects structural elements of appropriate size in combination with the resolution of the target image to perform morphological processing on the image change area, eliminating some noise interference areas with inappropriate shapes and areas. After completing the detection of the change area of ​​the image after the impact, the damage information of the change area is described, mainly the physical damage information of the target, including the number of craters N, the damage area S, and the physical damage rate P. The calculation formula is as follows:

[0082]

[0083] Among them, S i represents the area of ​​the crater numbered i, S T is the total area of ​​the target region, and the unit of area is pixel.

[0084] By analyzing the experimental data, we can draw the following conclusions: the grayscale difference method can extract almost all target damage areas, but at the same time detect a large number of pseudo-change areas. This method is not suitable for use alone, so this paper uses it to pre-detect the image. The entropy, contrast and inverse moment methods used alone can correctly detect the number of craters, but there are still certain errors in the damage area and physical damage rate, indicating that a single eigenvalue is difficult to fully describe the image texture features. Therefore, based on the full use of the advantages of the grayscale difference method such as fast speed and high detection rate, this paper selects three texture features with relatively good characterization performance: entropy, contrast, and inverse moment, and then uses the fused feature vector for image change detection. This method can not only accurately detect the number of craters, but also more accurately count the target damage area and physical damage rate, and can effectively extract target damage information.

[0085] In a more specific technical solution, a striking process simulation system according to claim 1 includes:

[0086] A reconnaissance and acquisition module is used to detect target information and damage information, wherein the damage information includes: a pre-strike image, a post-strike image, a binary image of the changed area, a target mask, and the coordinates of the target's circumscribed rectangle in the image;

[0087] The cropping module is used to crop the image before the impact, the image after the impact, the binary image of the changed area and the target mask according to the coordinates of the circumscribed rectangle of the target in the image, and obtain the cropped image I before the impact 1 , Image cropping after attack I 2 , Binary image cropping of the changing area I 3 And the target mask clipping image I 4 , the cropping module is connected with the reconnaissance and collection module;

[0088] Image overlay module, used to crop binary images of changing areas 3 Performing color inversion processing and image superposition operations to obtain an inverted color superposition image before striking and an inverted color superposition image after striking, and the image superposition module is connected to the cropping module;

[0089] A grayscale symbiosis calculation module is used to invert and overlay the image before and after the strike, obtain the number of occurrences of a specific pixel point, and process to obtain a grayscale symbiosis matrix GLCM. The grayscale symbiosis calculation module is connected to the image overlay module;

[0090] A target texture feature description module is used to obtain the recurrence probability of the grayscale structure in the inverted color superposition image before the impact and the inverted color superposition image after the impact according to the grayscale co-occurrence matrix GLCM, so as to describe the target texture features, wherein the target texture features include: entropy value, contrast, correlation and angular second-order moment, and the target texture feature description module is connected with the grayscale co-occurrence calculation module;

[0091] A texture feature change rate processing module is used to process the target texture feature to obtain the texture feature average value so as to obtain the target texture feature change rate. The texture feature change rate processing module is connected to the target texture feature description module.

[0092] A geometric feature change rate processing module is used to calculate the target geometric feature change rate according to the target information. The geometric feature change rate processing module is connected to the reconnaissance and collection module.

[0093] The target damage assessment level acquisition module is used to evaluate the damage of ground targets based on the damage level interval of the preset cloud model, according to the target texture feature change rate and the target geometric feature change rate. The digital features of the preset cloud model include: expected E x 、Entropy E n And the super entropy H e , the target damage assessment level acquisition module is connected with the geometric feature change rate processing module and the texture feature change rate processing module.

[0094] Compared with the prior art, the present invention has the following advantages:

[0095] The present invention uses image processing technology to extract target damage feature information based on the image quality, target type and target characteristics of the acquired ground targets, comprehensively evaluates the target damage effect, provides powerful decision support for specific application scenarios, and optimizes the firepower strike plan. Compared with traditional fuzzy reasoning, the cloud model used in the present invention takes into account the fuzziness and randomness in the decision-making process, and the reasoning results are more reasonable.

[0096] The present invention can evaluate the damage effect on the target through the gray-level co-occurrence matrix and cloud model based on the image information obtained by the reconnaissance means in the specific application scenario, accurately evaluate the target damage level, and provide the target damage situation for the command decision-making department in the specific application scenario.

[0097] The present invention solves the technical problems existing in the prior art of low accuracy in damage effect evaluation and unreasonable reasoning results. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1a A schematic diagram of basic steps of a method for simulating a striking process according to Embodiment 1 of the present invention;

[0099] Figure 1b This is a schematic diagram of a specific data flow processing of a strike process simulation method according to Embodiment 1 of the present invention;

[0100] Figure 2a This is a schematic diagram of a damage image input before the strike according to Embodiment 1 of the present invention;

[0101] Figure 2bThis is a schematic diagram of inputting a damage image after a strike according to Embodiment 1 of the present invention;

[0102] Figure 2c This is a schematic diagram of a target mask image of Example 1 of the present invention;

[0103] Figure 2d This is a schematic diagram of a binary image of a changed region according to Embodiment 1 of the present invention;

[0104] Figure 3a This is a schematic diagram of image cropping before striking according to Embodiment 1 of the present invention;

[0105] Figure 3b This is a schematic diagram of cropping an image after being struck according to Embodiment 1 of the present invention;

[0106] Figure 3c This is a schematic diagram of cropping a binary image in a changing region according to Embodiment 1 of the present invention;

[0107] Figure 3d This is a schematic diagram of target mask cutting in Example 1 of the present invention;

[0108] Figure 4 This is a schematic diagram of specific steps of image superposition in Example 1 of the present invention;

[0109] Figure 5 Schematic diagram of specific steps for calculating the gray level co-occurrence matrix in Example 1 of the present invention;

[0110] Figure 6 Schematic diagram of specific steps for calculating the target texture feature change rate in Example 1 of the present invention;

[0111] Figure 7 This is a schematic diagram of the specific steps for evaluating the target damage effect according to Example 1 of the present invention. DETAILED DESCRIPTION

[0112] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in combination with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0113] Example 1

[0114] like Figure 1a and Figure 1b As shown, a method for simulating a striking process provided by the present invention comprises the following basic steps:

[0115] Step S1, receiving damage information;

[0116] In the task process simulated in this embodiment, target information is obtained through specific application scenario reconnaissance means, and the obtained damage information includes but is not limited to: pre-strike image, post-strike image, change area binary image, target mask, and target circumscribed rectangle coordinates in the image; in this embodiment, the pre-strike image and the post-strike image are visible light images; in this embodiment, the first pixel point in the upper left corner of the image is the coordinate origin O, the horizontal right direction of the image is the positive direction of the x-axis, and the vertical downward direction is the positive direction of the y-axis, and the target circumscribed rectangle coordinates in the image are: (x 1 ,y 1 )、(x 2 ,y 2 )、(x 3 ,y 3 )、(x 4 ,y 4 ), where x 1 、x 2 、x 3 、x 4 ,y 1 ,y 2 ,y 3 ,y 4 Indicates the coordinate value, x 1 =x 2 , x 3 =x 4 , x 1 <x 3 ,y 1 =y 3 ,y 2 =y 4 ,y 1 <y 2 ;

[0117] like Figure 2a , Figure 2b , Figure 2c , Figure 2d As shown, in this embodiment, the coordinates of the circumscribed rectangle of the target in the image are: (x 1 ,y 1 )=(107,137),(x 2 ,y 2 )=(107,250),(x 3 ,y 3 )=(247,137),(x 4 ,y 4 )=(247,250).

[0118] Step S2, image cropping;

[0119] like Figure 3a , Figure 3b , Figure 3c , Figure 3d As shown, in this embodiment, according to the circumscribed rectangle coordinates of the target in the image received in the aforementioned step S1, the image before the strike, the image after the strike, the binary image of the changed area and the target mask received in step S1 are respectively cropped to obtain a cropped image of the image before the strike, which is denoted as I 1 , the cropped image after the attack, denoted as I 2 , the cropped image of the binary image in the changing area, denoted as I 3 , the target mask cropping image, denoted as I 4 ;

[0120] Step S3, image superposition;

[0121] like Figure 4 As shown, in this embodiment, the image superposition step S3 further includes the following specific steps:

[0122] Step S31: The image I outputted in step S2 is 3 Invert color processing;

[0123] In this embodiment, the color inversion process is performed using the following formula:

[0124] I 3_new =I 5 -I 3 (1)

[0125] Among them, I 3_new Represents the image after inversion processing; I 5 A white image in which the color value of each pixel is (255,255,255);

[0126] Step S32: image I 3_new The image I outputted in step S2 is 1 ,I 2 superposition;

[0127] In this embodiment, the following superposition process is performed:

[0128] I 1_new =I 3_new +I 1 (2)

[0129] I 2_new =I 3_new +I 2 (3)

[0130] Among them, I 1_new ,I 2_new Represents image I 3_new Respectively with image I 1 ,I 2 The superimposed image;

[0131] Step S4, calculating the Grey-Level Co-occurrence Matrix (GLCM);

[0132] In this embodiment, the gray level co-occurrence matrix method is also called the joint probability matrix method, which describes the image texture features based on the probability of repeated occurrence of gray structure in the image;

[0133] like Figure 5 As shown, in this embodiment, the specific steps of calculating the Grey-Level Co-occurrence Matrix (GLCM) in the aforementioned step S4 also include:

[0134] Step S41: Set the input image size to k 1 ×k 2 , the image is marked as I, and the gray level is N g , take any pixel point A(k,l), k,l represent the pixel point coordinates, take point A as the center, the horizontal direction as the reference line, the pixel point at a distance d and an azimuth angle θ is B(m,n), m,n represent the pixel point coordinates, and find the number of times point B appears:

[0135] P(i,j|d,θ)=num{(k,l)∈I k1×k2 ,(m,n)∈I k1×k2 |d,θ,I(k,l)=i,I(m,n)=j} (4)

[0136] Where: i represents the gray level N g j represents the pixel grayscale value of point A in the case of ; j represents the pixel grayscale value of point B; num{·} represents the number of all pixel pairs that meet the above conditions in the area;

[0137] According to formula (4), the calculation result is recorded as matrix G:

[0138]

[0139] Step S42: normalize the matrix G:

[0140]

[0141] According to formula (6), the gray level co-occurrence matrix is ​​calculated and recorded as GLCM:

[0142]

[0143] The excess entropy value is 0.003, reflecting the degree of discreteness of cloud droplets.

[0144] Step S5, calculating target texture features;

[0145] In this embodiment, the target texture features include but are not limited to: entropy value, contrast, correlation, and angular second-order moment;

[0146] In this embodiment, the entropy value is calculated using the following formula:

[0147]

[0148] Among them, C b1_θ represents the first texture feature value of the bth image at angle θ, that is, the entropy value;

[0149] In this embodiment, the contrast is calculated using the following logic:

[0150]

[0151] Among them, C b2_θ represents the second texture feature value of the bth image at angle θ, i.e., contrast;

[0152] In this embodiment, the correlation is calculated using the following formula:

[0153]

[0154] Among them, C b3_θ represents the third texture feature value of the bth image at angle θ, i.e., the correlation;

[0155]

[0156] In this embodiment, the angular second-order moment is calculated using the following formula:

[0157]

[0158] Among them, C b4_θ represents the fourth texture eigenvalue of the bth image at angle θ, i.e., the angular second-order moment;

[0159] Step S6, calculating the target texture feature change rate;

[0160] like Figure 6 As shown, in this embodiment, the step S6 of calculating the target texture feature change rate also includes the following specific steps:

[0161] Step S61: for the image I outputted in step S3 1_new ,I 2_new , where the image I output in step 3 is 1_new and image I 2_new The size is 140×113, and the gray level is N g=16, d=1, according to step S4 and step S5, for each target texture feature, the values ​​of θ when it is 0°, 45°, 90°, and 135° are calculated respectively, and then the average value of the calculation results at the four angles is taken as the final value of the texture feature:

[0162]

[0163] Among them, D ba represents the average value of the ath texture feature value of the bth image, where a=1,2,3,4; C ba_0° , C ba_45° , C ba_90° , C ba_135° It represents the ath texture feature value of the bth image when θ takes the values ​​of 0°, 45°, 90°, and 135° respectively;

[0164] Step S62, calculate the change rate of each texture feature:

[0165]

[0166] Among them, L a represents the rate of change of the ath texture feature;

[0167] In this embodiment, the calculation results are shown in Table 1.

[0168] Table 1 Texture feature calculation results

[0169] project Entropy Contrast Relevance Angular second moment Pre-strike image <![CDATA[1.661×10 12 ]]> <![CDATA[3.273×10 4 ]]> 0.999 <![CDATA[8.065×10 7 ]]> After the attack image <![CDATA[2.027×10 12 ]]> <![CDATA[4.580×10 4 ]]> 0.999 <![CDATA[7.128×10 7 ]]> Rate of change 22.077% 39.960% 0 11.622%

[0170] Step S7, calculating the target geometric feature change rate;

[0171] In this embodiment, the target geometric feature change rate is calculated using the following formula:

[0172]

[0173] Among them, L 5 Indicates the rate of change of the target geometric features; S change represents the number of white pixels in image I3 in step 2; S mask Represents image I in step 2 4 The number of white pixels;

[0174] In this embodiment, the target geometric feature change rate is:

[0175]

[0176] Step S8: Evaluate the target damage effect based on the cloud model.

[0177] In this embodiment, after calculating the target texture feature change rate and the target geometric feature change rate, the ground target damage is comprehensively evaluated based on the cloud model; in this embodiment, the digital features of the cloud model include but are not limited to: expected E x 、Entropy E n And the super entropy H e , which is expected to reflect the central value of the quantitative description; entropy is a comprehensive evaluation of qualitative concepts, which can reflect both the value space of cloud droplets and their discreteness; in this embodiment, the super entropy is empirically taken as 0.003, which reflects the discreteness of cloud droplets.

[0178] like Figure 7 As shown, in this embodiment, step S8 of evaluating the target damage effect also includes the following specific steps:

[0179] Step S81: By analyzing the target characteristics and according to the expert opinions, the target damage degree is divided into four levels, and four damage level intervals corresponding to the damage level are divided between [0, 1]:

[0180] Level Ⅰ damage (severe damage, the target is seriously damaged and its function is greatly reduced), [ψ 2 , ψ 1 ];

[0181] Level II damage (moderate damage, the target is severely damaged and some functions are lost), [ψ 4 ,ψ 3 );

[0182] Level III damage (mild damage, slight decrease in target function), [ψ 6 ,ψ 5 );

[0183] Level IV damage (no damage, the target is not damaged), [ψ 8 ,ψ 7 );

[0184] In this embodiment, the aforementioned damage level interval may be set to, for example:

[0185] Among them, ψ e Indicates the interval boundary value, e=1,2,…,7,8;

[0186] Level I damage (severe damage, the target is severely damaged and its function is greatly reduced), [0.4, 1];

[0187] Level II damage (moderate damage, the target is severely damaged and some functions are lost), [0.2, 0.4);

[0188] Level III damage (mild damage, slight decrease in target function), [0.05, 0.2);

[0189] Level IV damage (no damage, the target is not damaged), [0, 0.05).

[0190] Step S82, calculating the expectation of the sth damage level interval;

[0191] In this embodiment, the expectation of the sth damage level interval is calculated using the following formula:

[0192] E x_s =(Z min_s +Z max_s ) / 2 (19)

[0193] Where Ex_s represents the expectation of the sth damage level interval, s = 1, 2, 3, 4; Z min_s represents the lower limit of the sth damage level interval, s = 1, 2, 3, 4; Z max_s Indicates the upper boundary value of the sth damage level interval;

[0194] Step S83, calculating the entropy of the s-th damage level interval;

[0195] In this embodiment, the entropy of the sth damage level interval is calculated using the following formula:

[0196] E n_s =(Z min_s -Z max_s ) / 2.355 (20)

[0197] Among them, E n_s represents the entropy of the sth damage level interval;

[0198] Step S84: Calculate the degree of certainty of the oth characteristic change rate in the sth damage level interval as v os :

[0199]

[0200] Among them, L o Indicates the target texture feature change rate and geometric feature change rate, o = 1, 2, 3, 4, 5;

[0201] Step S85: Construct the degree of certainty matrix U from the degrees of certainty obtained in step S84 o×s , determine the weights of texture feature change rate and geometric feature change rate based on expert opinions, and calculate the comprehensive certainty;

[0202] In this embodiment, the weights of the texture feature (entropy value, contrast, correlation, angular second-order moment) change rate and the geometric feature change rate are determined based on expert opinions as W 1×5 =(w 1 ,w 2 ,w3 ,w 4 ,w 5 ), calculate the comprehensive certainty F:

[0203]

[0204] Among them, f s represents the degree of certainty at the sth damage level;

[0205] Step S86, selecting the maximum value in the comprehensive certainty F, and the damage level where the maximum value is located is the target damage level;

[0206] In this embodiment, according to the above formula (19) to the above formula (21), the certainty matrix U is calculated o×s :

[0207]

[0208] In this embodiment, the weights of the texture feature (entropy value, contrast, correlation, angular second-order moment) change rate and the geometric feature change rate are determined based on expert opinions as W 1×5 =(0.2, 0.2, 0.08, 0.02, 0.5), and then calculate the comprehensive certainty F:

[0209]

[0210] In the comprehensive certainty of this embodiment, the maximum certainty is 0.637, and the damage level corresponding to the maximum value is level I damage, that is, the damage level of the target is level I damage.

[0211] The present invention introduces a cloud model into the fuzzy neural network, and through the cloud model's ability to process uncertainty, the timeliness and accuracy of the algorithm are guaranteed while taking into account the fuzziness and randomness of the data.

[0212] In order to solve the problem that the membership function determination process requires experience and is greatly influenced by human factors, the present invention introduces the certainty of the cloud model into the fuzzy neural network to replace the membership degree. By randomly taking values ​​each time, the influence of human determination and expert experience is avoided. For different types of data sets, it is proposed to use Gaussian fitting or fuzzy clustering methods to determine the digital characteristics of the cloud to improve the accuracy of cloud model calculations; the cloud generator is combined with neurons to construct uncertainty neurons to achieve the transmission of data uncertainty. Since the cloud model is greatly affected by the data dimension and data volume when generating rules, the number of rules generated by multidimensional data grows exponentially. It is proposed to use lattice theory to perform upper and lower bounds on cloud rules, and merge similar cloud rules into cloud rules of a certain interval to achieve the purpose of improving the efficiency of the algorithm.

[0213] In this embodiment, the fuzzy neural network needs to adjust the initial value through error propagation. The determination of the initial value affects the number of error iterations. It is proposed to replace the traditional 0 and 1 with the expected Ex and entropy En of the cloud model to reduce the number of iterations and avoid the local optimal solution. On the basis of constructing the cloud fuzzy neural network model, the above method is used for experimental comparative analysis. Experiments are conducted through UCI data sets and engineering examples. The traditional cloud model, fuzzy neural network, and BP neural network are compared to analyze the operation efficiency, accuracy, and number of rules of the algorithm. Through experimental comparative analysis, the cloud fuzzy neural network can fully consider the randomness and fuzziness of the data, optimize the generation of cloud rules, avoid multi-dimensional rule disasters, and ensure the operation efficiency of the algorithm; the accuracy of the algorithm is improved compared with the traditional technical algorithm, and it is applicable to a variety of data sets.

[0214] In summary, the present invention uses image processing technology to extract target damage feature information based on the image quality, target type and target characteristics of the acquired ground targets, comprehensively evaluates the target damage effect, provides powerful decision support for specific application scenarios, and optimizes the firepower strike plan. Compared with traditional fuzzy reasoning, the cloud model used in the present invention takes into account the fuzziness and randomness in the decision-making process, and the reasoning results are more reasonable.

[0215] The present invention can evaluate the damage effect on the target through the gray-level co-occurrence matrix and cloud model based on the image information obtained by the reconnaissance means in the specific application scenario, accurately evaluate the target damage level, and provide the target damage situation for the command decision-making department in the specific application scenario.

[0216] The present invention solves the technical problems existing in the prior art of low accuracy in damage effect evaluation and unreasonable reasoning results.

[0217] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for simulating a striking process, characterized in that: The method comprises: S1. Detecting target information and damage information, wherein the damage information includes: an image before the strike, an image after the strike, a binary image of the changed area, a target mask, and coordinates of a circumscribed rectangle of the target in the image; S2, according to the coordinates of the circumscribed rectangle of the target in the image, respectively cropping the pre-strike image, the post-strike image, the binary image of the changed area and the target mask to obtain a cropped image I1 of the pre-strike image, a cropped image I2 of the post-strike image, a cropped image I3 of the binary image of the changed area and a cropped image I4 of the target mask; S3, performing color inversion processing and image overlay operation on the changed region binary image cropping image I3 to obtain an inverted color overlay image before striking and an inverted color overlay image after striking; S4, obtaining the number of occurrences of specific pixels in the inverted color superposition image before the strike and the inverted color superposition image after the strike, so as to obtain a gray level co-occurrence matrix GLCM; S5, according to the gray level co-occurrence matrix GLCM, in the inverted color superposition image before the impact and the inverted color superposition image after the impact, obtaining the recurrence probability of the gray structure to describe the target texture characteristics, wherein the target texture characteristics include: entropy value, contrast, correlation and angular second-order moment; S6. According to the target texture feature, processing is performed to obtain an average value of the texture feature to obtain a change rate of the target texture feature; S7, calculating the target geometric feature change rate according to the target information; S8, based on the damage level interval of the preset cloud model, according to the target texture feature change rate and the target geometric feature change rate, evaluate the damage of the ground target, wherein the digital features of the preset cloud model include: expected E x 、Entropy E n And the super entropy H e .

2. A method for simulating a striking process according to claim 1, characterized in that: The step S3 comprises: S31, performing color inversion processing on the cropped binary image I3 of the changed region to obtain a color inversion processed image: I 3_new =I5 -I3 (1) In the formula, I 3_new represents the inverted image, and I5 represents a white image in which the color value of each pixel is (255, 255, 255); S32, the inverted image I 3_new , respectively, are superimposed with the pre-strike image cropping image I1 and the post-strike image cropping image I2 to obtain the pre-strike inverted superimposed image and the post-strike inverted superimposed image: I 1_new =I 3_new +I1 (2) I 2_new =I 3_new +I2 (3) In the formula, I 1_new ,I 2_new They respectively represent the inverted color superimposed image before the strike and the inverted color superimposed image after the strike.

3. A method for simulating a striking process according to claim 1, characterized in that: The step S4 comprises: S41, take a pixel center point A, take the horizontal direction as the reference line, obtain the specific pixel point B (m, n) at a position with a distance d and an azimuth angle θ from the pixel center point A, and calculate the number of occurrences of the specific pixel point B (m, n): P(i,j|d,θ)=num{(k,l)∈I k1×k2 ,(m,n)∈I k1×k2 |d,θ,I(k,l)=i,I(m,n)=j} (4) In the formula, i represents the gray level N g In the case of , j represents the pixel grayscale value of point A, j represents the pixel grayscale value of point B, num{·} represents the number of all pixel pairs that meet the above conditions in the area, and m, n represent the pixel coordinates; According to the number of occurrences, the initial matrix G is obtained by the following logical processing: S42. Using the following logic, the initial matrix G is normalized and gray-level co-occurrence is calculated to obtain the gray-level co-occurrence matrix GLCM:

4. A method for simulating a striking process according to claim 1, characterized in that: The step S5 comprises: S51. Calculate the entropy value using the following logic: In the formula, C b1_θ represents the first entropy value of the b-th image at angle θ; S52, using the following logic to calculate the contrast: In the formula, C b2_θ represents the second contrast of the b-th image at angle θ; S53, using the following logic to calculate the correlation: In the formula, C b3_θ represents the third correlation of the bth image at angle θ; S54, using the following logic, calculate the angular second-order moment: In the formula, C b4_θ It represents the 4th angular second-order moment of the bth image at an angle of θ.

5. A method for simulating a striking process according to claim 4, characterized in that: In step S53, the formula (10) satisfies:

6. A method for simulating a striking process according to claim 1, characterized in that: The step S6 comprises: S61, for each of the inverted color superimposed image before the impact and the inverted color superimposed image after the impact, take no less than two specific angles, and calculate the average value of the texture feature as the texture feature value: Where D ba represents the average value of the ath texture feature value of the bth image, where a=1,2,3,4; C ba_0° , C ba_45° , C ba_90° , C ba_135° It represents the ath texture feature value of the bth image when θ takes the values ​​of 0°, 45°, 90°, and 135° respectively; S62. Calculate the target texture feature change rate according to the texture feature value: Where, L a Represents the rate of change of the ath texture feature.

7. A method for simulating a striking process according to claim 1, characterized in that: In step S7, the target geometric feature change rate is calculated using the following logic: Where, L5 represents the rate of change of target geometric features; S change Indicates the number of white pixels in image I3 in step 2; S mask Indicates the number of white pixels in image I4 in step 2.

8. A method for simulating a striking process according to claim 1, characterized in that: The step S8 comprises: S81, dividing the target damage degree into no less than 2 damage level intervals by analyzing the target characteristics according to the target characteristic information and expert opinions; S82, calculating the expectation of the sth damage level interval: E x_s =(Z min_s +Z max_s ) / 2 (19) Where Ex_s represents the expectation of the sth damage level interval; Z min_s represents the lower limit of the sth damage level interval; Z max_s Indicates the upper boundary value of the sth damage level interval; S83. Calculate the entropy of the sth damage level interval using the following logic: E n_s =(Z min_s -Z max_s ) / 2.355 (20) In the formula, E n_s represents the entropy of the sth damage level interval; S84, calculating the oth target texture feature change rate L o , the degree of certainty v in the sth damage level interval os : S85, using the degree of certainty v os Construct the certainty matrix U o×s , according to the expert opinion information, determining the weights of the target texture feature change rate and the geometric feature change rate to obtain a comprehensive degree of certainty; S86. Select the maximum comprehensive certainty from the comprehensive certainty F to obtain the target damage level.

9. A method for simulating a striking process according to claim 8, characterized in that: In the step S85, the comprehensive certainty F is calculated according to the weight: In the formula, f s represents the certainty at the sth damage level, W 1×5 The weights representing the target texture feature change rate and the geometric feature change rate.

10. The striking process simulation system according to claim 1, characterized in that: The system comprises: A reconnaissance and acquisition module, used to detect target information and damage information, wherein the damage information includes: a pre-strike image, a post-strike image, a binary image of the changed area, a target mask, and the coordinates of a circumscribed rectangle of the target in the image; A cropping module, for cropping the pre-strike image, the post-strike image, the binary image of the changed region and the target mask according to the coordinates of the circumscribed rectangle of the target in the image, to obtain a cropped image I1 of the pre-strike image, a cropped image I2 of the post-strike image, a cropped image I3 of the binary image of the changed region and a cropped image I4 of the target mask, wherein the cropping module is connected to the reconnaissance and acquisition module; An image overlay module, used for performing color inversion processing and image overlay operation on the change region binary image cropping image I3, so as to obtain an inverted overlay image before the strike and an inverted overlay image after the strike, the image overlay module being connected to the cropping module; A grayscale symbiosis calculation module, used for inverting and superimposing the image before the attack and the image after the attack, obtaining the number of occurrences of specific pixels to obtain a grayscale symbiosis matrix GLCM, and the grayscale symbiosis calculation module is connected to the image superposition module; a target texture feature description module, for obtaining the recurrence probability of the grayscale structure in the inverted color superposition image before striking and the inverted color superposition image after striking according to the grayscale co-occurrence matrix GLCM, so as to describe the target texture feature, wherein the target texture feature includes: entropy value, contrast, correlation and angular second-order moment, and the target texture feature description module is connected to the grayscale co-occurrence calculation module; A texture feature change rate processing module, used for processing the target texture feature to obtain a texture feature average value so as to obtain a target texture feature change rate, wherein the texture feature change rate processing module is connected to the target texture feature description module; A geometric feature change rate processing module, used for calculating the target geometric feature change rate according to the target information, the set feature change rate processing module is connected to the reconnaissance acquisition module; The target damage assessment level acquisition module is used to assess the damage of the ground target based on the damage level interval of the preset cloud model and the target texture feature change rate and the target geometric feature change rate, wherein the digital features of the preset cloud model include: expected E x 、Entropy E n And the super entropy H e , the target damage assessment level acquisition module is connected to the geometric feature change rate processing module and the texture feature change rate processing module.

Citation Information

Patent Citations

  • Ground target cloud model damage effect evaluation method based on visible light image processing

    CN113283280A

  • Ground target visible light damage image processing method for change detection and target detection

    CN114037650A

  • Damage assessment method, system and equipment based on image analysis and storage medium

    CN114841956A