Laser seeker vision intelligent detection method based on computer vision

Abstract

The invention relates to the technical field of image processing, in particular to a laser seeker vision intelligent detection method based on computer vision, and the method comprises the steps: obtaining a laser seeker grey-scale map; determining the side length of the light spot detection window of each pixel point; determining a suspected light spot index of each pixel point according to the confusion degree of the gray values of all the pixel points in the light spot detection window of each pixel point and the difference between the gray value of each pixel point and the gray values of all the pixel points in the light spot detection window, and determining a light spot definite index of each pixel point in combination with the suspected light spot index, therefore, the laser seeker can be detected. According to the invention, the adaptive light spot detection window is arranged for each pixel point, and the interference of the light spots on the detection of the laser seeker is eliminated in combination with the light spot confirmation index, so that the detection precision of the laser seeker is improved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and specifically to a visual intelligent detection method for a laser seeker based on computer vision. Background Art

[0002] Laser guidance, as a precise guidance method, has been widely applied in actual combat. The laser seeker processes the laser signals obtained on the detector to acquire the position information of the target. Therefore, the optical imaging performance of the seeker will directly affect the ability of the laser seeker to track the target.

[0003] As one of the important components of the laser seeker, the radome is located at the very front of the seeker and plays a role in protecting the internal components from external environmental erosion and rectification. During the optical processing, due to reasons such as equipment and technology, there will be fine defects such as scratches and pockmarks on the surface of the radome. Currently, when using computer vision algorithms to detect the defects of the radome of the laser seeker, the light spots formed by the reflection on the surface of the radome will seriously interfere with the capture of fine defects in the image, resulting in a decrease in the defect detection accuracy of the radome. Therefore, how to reduce the influence of the reflection light spots on the surface of the laser seeker radome on the detection and improve the detection accuracy of fine defects is an urgent problem to be solved currently. Summary of the Invention

[0004] In order to solve the above technical problems, this application provides a visual intelligent detection method for a laser seeker based on computer vision to solve the existing problems.

[0005] The visual intelligent detection method for a laser seeker based on computer vision in this application adopts the following technical solutions: An embodiment of this application provides a visual intelligent detection method for a laser seeker based on computer vision, and this method includes the following steps: Obtain the grayscale images of the laser seeker at different light source angles; Determine the side length of the light spot detection window for each pixel point according to the distribution of the grayscale values of each pixel point in the grayscale image of the laser seeker; Determine the suspected light spot index for each pixel point according to the degree of chaos of the grayscale values of all pixel points in the light spot detection window of each pixel point and the difference between the grayscale value of each pixel point and the grayscale values of all pixel points in its light spot detection window; Based on the side length of the light spot detection window of each pixel point and the distribution of all pixel points around the central pixel point inside it, combine the suspected light spot index to determine the light spot certainty index for each pixel point to detect the laser seeker.

[0006] Preferably, the expression for the side length of the light spot detection window for each pixel point is: ; where denotes the side length of the light spot detection window for the \(i\)-th pixel point in the grayscale image of the laser seeker; denotes the grayscale value of the \(i\)-th pixel point in the grayscale image of the laser seeker; and respectively denote the preset reference window side length and the preset minimum window side length; ceil( ) represents the ceiling function; norm( ) represents the normalization function.

[0007] Preferably, the determination process of the suspected light spot index for each pixel point is as follows: Determine the chaos degree of each pixel point according to the chaos degree of all pixel points in the light spot detection window of each pixel point; Determine the suspected light spot degree of each pixel point according to the difference between the grayscale value of each pixel point and the grayscale values of all pixel points in its light spot detection window; Based on the chaos degree and the suspected light spot degree, determine the suspected light spot index of each pixel point.

[0008] Preferably, the chaos degree of each pixel point is the image entropy of the light spot detection window where each pixel point is located.

[0009] Preferably, the suspected light spot degree of each pixel point is the ratio of the grayscale value of each pixel point to the minimum grayscale value of all pixel points in its light spot detection window.

[0010] Preferably, the suspected light spot index of each pixel point is the product of the chaos degree and the suspected light spot of each pixel point.

[0011] Preferably, the determination process of the light spot certainty index for each pixel point is as follows: Based on the side length of the light spot detection window of each pixel point and the distribution of all pixel points around the central pixel point inside it, determine the light spot area ratio of each pixel point; Based on the light spot area ratio and the suspected light spot index, determine the light spot certainty index of each pixel point.

[0012] Preferably, the determination process of the light spot area ratio of each pixel point is as follows: Cluster all pixel points in the light spot detection window of each pixel point to obtain multiple clusters; The light spot area ratio of each pixel point is the ratio of the number of all pixel points in the cluster where the central pixel point in the light spot detection window of each pixel point is located to the square of the side length of the light spot detection window.

[0013] Preferably, the light spot certainty index of each pixel point is the product of the suspected light spot index and the light spot area of each pixel point.

[0014] Preferably, the detection of the laser seeker includes: The spot confidence index of all pixel points in the grayscale image of the laser seeker is used as the input of the threshold segmentation algorithm to obtain the segmentation threshold; The pixel points in the grayscale image of the laser seeker whose spot confidence index is less than or equal to the segmentation threshold are marked as valid pixel points; The mean value of the grayscale values of all valid pixel points in the spot detection window of each pixel point is used as the fusion weight of each pixel point; The grayscale value of each pixel point in the optimized grayscale image of the laser seeker is the sum of the products of the grayscale values of the pixel points at the same positions in all grayscale images of the laser seeker and the fusion weights; The U²-Net neural network model is used to detect defects in the optimized grayscale image of the laser seeker.

[0015] This application has at least the following beneficial effects: This application determines the side length of the spot detection window for each pixel point according to the distribution of the grayscale values of each pixel point in the grayscale image of the laser seeker, avoiding misjudging subsequent defects as spots for segmentation; according to the degree of chaos of the grayscale values of all pixel points in the spot detection window of each pixel point, and the difference between the grayscale value of each pixel point and the grayscale values of all pixel points in its spot detection window, the suspected spot index of each pixel point is determined; based on the side length of the spot detection window of each pixel point and the distribution of all pixel points around the central pixel point therein, combined with the suspected spot index, the spot confidence index of each pixel point is determined; to enhance the distinguishability between spot pixel points and other pixel points and improve the accuracy of the spot segmentation on the fairing surface. This application sets an adaptive spot detection window for each pixel point, combines the spot confidence index to exclude the interference of spots on the laser seeker detection, and improves the accuracy of the laser seeker detection. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a flow chart of the steps of a laser seeker vision intelligent detection method based on computer vision; Figure 2 It is a schematic diagram for extracting the spot confidence index; Figure 3 It is a flow chart of the laser seeker defect recognition process; Figure 4 It is a schematic diagram of the laser seeker defect intelligent detection process. Detailed Implementation Manner

[0018] In order to further elaborate on the technical means and effects adopted by this application to achieve the intended invention purpose, the following combines the accompanying drawings and preferred embodiments to detail the specific implementation manner, structure, features, and effects of the computer vision-based visual intelligent detection method for a laser seeker proposed according to this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs.

[0020] The following specifically describes the specific solution of the computer vision-based visual intelligent detection method for a laser seeker provided by this application with reference to the accompanying drawings.

[0021] A computer vision-based visual intelligent detection method for a laser seeker provided in an embodiment of this application. Specifically, the following computer vision-based visual intelligent detection method is provided. Please refer to Figure 1 , and the method includes the following steps: Step S1: Use an industrial CCD camera to photograph the fairing of the laser seeker, perform denoising processing on the photographed grayscale image, and adjust the angle of the annular light source to obtain grayscale images of the laser seeker under different angles of illumination.

[0022] Irradiate the fairing of the laser seeker with the annular light source, and use an industrial CCD camera to capture a grayscale image with a resolution of Q*H. In order to eliminate the influence of noise during the shooting process and enhance the accuracy of subsequent analysis, a filtering algorithm is used to perform denoising processing on the photographed grayscale image, and the denoised grayscale image is denoted as the grayscale image of the laser seeker. By adjusting the illumination angle of the annular light source, the position where the light spot appears in each image is changed, and a preset number of grayscale images of the laser seeker under different light source angles of illumination are obtained.

[0023] It should be noted that the values of Q, H, the preset number, and the light source angle are set artificially. In this embodiment, Q and H are respectively set to 3088 and 2064, the preset number is set to 4, and the illumination angles of the annular light source are respectively selected as 30°, 45°, 60°, and 90°. Implementers can set and adjust according to the actual situation, and this embodiment does not make special restrictions.

[0024] It should be understood that there are many common filtering algorithms. In this embodiment, the Gaussian filtering algorithm is used to denoise the grayscale image of the laser seeker. Implementers can also use filtering algorithms such as median filtering or bilateral filtering to denoise the grayscale image of the laser seeker. This embodiment does not make special restrictions.

[0025] Step S2: Determine the side length of the spot detection window for each pixel point according to the distribution of the grayscale values of each pixel point in the grayscale image of the laser seeker.

[0026] As a key component of the laser seeker, the radome has a great influence on the laser signal obtained by the laser seeker. During the optical processing, due to reasons such as equipment and technology, there will be fine defects such as scratches and pockmarks on the surface of the radome, resulting in optical errors when the laser seeker images, affecting the quality of the optical imaging of the laser seeker and the accuracy of target detection. Compared with other optical defects, the scratches and pockmarks on the surface of the radome have a small existence range and are randomly distributed. At the same time, the spots generated by the light reflection on the surface of the radome have grayscale values in the image close to those of the defects. When the spots coincide with the fine defects on the surface, it will seriously interfere with the detection of surface defects, resulting in a decrease in detection accuracy and the omission of some fine defects.

[0027] In order to reduce the influence of the spots on the surface of the radome on defect detection, it is necessary to segment the spots in the grayscale image of the laser seeker. Since the change range of the grayscale values of the pixel points at the edge of the spot in the image is large, if the entire image is directly segmented, it is difficult to obtain a good segmentation effect at the edge of the spot, thereby affecting the subsequent defect detection at the edge of the spot.

[0028] Therefore, a window of size r×r is established with each pixel point in the grayscale image of the laser seeker as the center, and it is denoted as the spot detection window for each pixel point, where r is the side length of the spot detection window for each pixel point.

[0029] Determine the side length of the spot detection window for each pixel point according to the distribution of the grayscale values of each pixel point in the grayscale image of the laser seeker. The expression for the side length of the spot detection window for each pixel point is: ; In the formula, represents the side length of the spot detection window for the i-th pixel point in the grayscale image of the laser seeker; represents the grayscale value of the i-th pixel point in the grayscale image of the laser seeker; and respectively represent the preset reference window side length and the preset minimum window side length; ceil( ) represents the ceiling function; norm( ) represents the normalization function. and respectively represent the preset reference window side length and the preset minimum window side length. In the present invention, The value is 200, The value is 10.

[0030] It should be noted that the values of the preset reference window side length and the preset minimum window side length are set artificially. In this embodiment, the preset reference window side length has a value of 200, and the preset minimum window side length has a value of 10. Implementers can set them according to specific circumstances, and this embodiment does not have special restrictions.

[0031] The larger the gray value of the pixel point in the gray-scale image of the laser seeker, the larger the side length of its spot detection window. On the one hand, the gray value at the center of the spot is the highest. Using a larger spot detection window avoids the situation where only spot pixels are included in the window; at the same time, a smaller spot detection window is used for the normal pixels near the spot, reducing the excessive expansion of the subsequent spot segmentation area. On the other hand, the gray values of defects such as scratches and pockmarks are relatively high. A larger spot detection window is used to reduce the proportion of scratches, pockmarks and other defects in the window, increasing the distinguishability between scratches, pockmarks and other defects and spot pixels, and avoiding misjudging defects as spots for subsequent segmentation.

[0032] Step S3: Determine the suspected spot index of each pixel point according to the degree of chaos of the gray values of all pixel points in the spot detection window of each pixel point, and the difference between the gray value of each pixel point and the gray values of all pixel points in its spot detection window.

[0033] The gray values of the pixel points where defects such as pockmarks and scratches are located are relatively high, and their corresponding spot detection windows are larger. However, due to the small area of the defects, the proportion in the window is not large, and the degree of chaos of the gray value distribution in the window is small; the gray values of the pixel points where the spots are located are relatively high, and the corresponding spot detection windows are larger. However, the area of the spots is large and the range of change of the gray value at the spot is large, corresponding to a large degree of chaos of the gray value distribution in its window.

[0034] The gray values of the normal pixel points on the surface of the laser seeker fairing are relatively low, and the spot detection windows are relatively small. When the normal pixel points are far from the spot on the fairing surface, only the nearby normal pixel points are included in their spot detection windows, and the gray value distribution is uniform. When the normal pixel points approach the spot, spot pixels will appear in the corresponding spot detection windows, resulting in an increase in the degree of chaos of the gray value distribution of the pixel points in the spot detection windows. Moreover, the farther the normal pixel points are from the spot, the smaller the degree of chaos of the gray value distribution of the pixel points in the spot detection windows.

[0035] Therefore, according to the degree of chaos of the gray values of all pixels in the light spot detection window of each pixel, and the difference between the gray value of each pixel and the gray values of all pixels in its light spot detection window, a suspected light spot index for each pixel is constructed to exclude the influence of light spot pixels on the detection of defects such as pockmarks and scratches. Specifically: Analyze the image entropy of the light spot detection window where each pixel is located, denoted as the chaos degree of each pixel; Determine the ratio of the gray value of each pixel to the minimum gray value of all pixels in its light spot detection window, denoted as the suspected light spot degree of each pixel; Based on the chaos degree and the suspected light spot degree, determine the suspected light spot index of each pixel. The suspected light spot index of each pixel is the product of the chaos degree and the suspected light spot degree of each pixel.

[0036] It should be noted that the calculation process of image entropy is a well-known technology, and the specific calculation process will not be elaborated here.

[0037] The defects on the surface of the laser seeker fairing and the pixels at the light spots have relatively high gray values, and the ratio to the minimum gray value in their respective light spot detection windows is relatively large; while the gray values of normal pixels are relatively low, and the ratio to the minimum gray value in their light spot detection windows is relatively small, and the calculated suspected light spot index is relatively small. In addition, when the degree of chaos of the pixel distribution in the light spot detection window is greater, the image entropy of the light spot detection window where each pixel is located is greater, and the calculated suspected light spot index of the central pixel is greater. Therefore, the suspected light spot index of the pixels at the light spots is greater than the suspected light spot indices of the pixels at the surface defects and normal positions of the laser seeker.

[0038] Step S4: Based on the side length of the light spot detection window of each pixel and the distribution of all pixels around the central pixel within it, and in combination with the suspected light spot index, determine the light spot confidence index of each pixel.

[0039] When the distance between the surface defect of the laser seeker and the light spot is relatively close, the difference between the suspected light spot index of the defective pixel and the window gray value of the light spot pixel in the optimized gray map of the laser seeker is relatively small, and it is easy to mis-segment the defective pixels close to the light spot in the subsequent image segmentation process, reducing the detection accuracy of the surface defects of the laser seeker. However, the areas of defects such as pockmarks and scratches on the surface of the laser seeker fairing are relatively small, and the difference from the area of the light spot is relatively large.

[0040] In order to further improve the accuracy of light spot segmentation on the surface of the fairing, all pixels in the light spot detection window of each pixel are used as the input of the clustering algorithm to obtain multiple clustering clusters.

[0041] It should be understood that there are many common clustering algorithms. In this embodiment, the DBSCAN algorithm is used to cluster all the pixels in the spot detection window of each pixel point. The neighborhood radius of the DBSCAN algorithm is set to a, and the number of clustering clusters is set to m. The values of the neighborhood radius and the number of clustering clusters are both set artificially. In this embodiment, the value of a is 2 and the value of m is 10. Implementers can use other algorithms such as the k-means clustering algorithm according to specific situations, or can also set the values of the neighborhood radius and the number of clustering clusters by themselves according to specific situations. This embodiment does not make special restrictions.

[0042] Among them, the DBSCAN algorithm is a well-known technology, and the specific process of its clustering will not be elaborated here.

[0043] Furthermore, based on the side length of the spot detection window of each pixel point and the distribution of all the pixels around the central pixel point therein, combined with the suspected spot index, the spot confidence index of each pixel point is determined. Specifically: Analyze that the spot area ratio of each pixel point is the ratio of the number of all the pixels in the clustering cluster where the central pixel point in the spot detection window of each pixel point is located to the square of the side length of the spot detection window, denoted as the spot area ratio of each pixel point; Based on the spot area ratio and the suspected spot index, the spot confidence index of each pixel point is determined. The spot confidence index of each pixel point is the product of the suspected spot index of each pixel point and the spot area ratio.

[0044] Compared with the normal pixels and defective pixels in the gray-scale image of the laser seeker, the suspected spot index at the spot on the surface of the laser seeker is larger, and the calculated spot confidence index is larger. In addition, at the defective parts such as pockmarks and scratches on the surface of the fairing, the ratio of the clustering cluster where the central pixel point in its spot detection window is located to the window is smaller, while the ratio at the spot is larger. Therefore, the spot confidence index of the pixels at the spot is significantly larger than that of the normal pixels and defective pixels.

[0045] Preferably, as an embodiment of the present application, the schematic diagram for extracting the spot confidence index is as Figure 2 shown.

[0046] Step S5: Based on the spot confidence index of each pixel point, the laser seeker is detected.

[0047] (1) Taking the spot confidence indices of all the pixels in the gray-scale image of the laser seeker as the input, the segmentation threshold is calculated by using the maximum inter-class variance method.

[0048] (2) Denote the pixel points in the grayscale image of the laser seeker with a spot confidence index greater than the segmentation threshold as spot pixel points, representing the pixel points interfered by the spot on the radome surface; denote the pixel points with a spot confidence index not exceeding the segmentation threshold as valid pixel points, representing the pixel points not interfered by the spot.

[0049] (3) Further, obtain the grayscale images of the laser seeker in a preset number for the same laser seeker. For the pixel points at the same position in the preset number of grayscale images of the laser seeker, count the number of valid pixel points in the spot detection window of each pixel point, and take the mean value of the grayscale values of all valid pixel points in the spot detection window of each pixel point as the fusion weight of each pixel point.

[0050] (4) The grayscale value of each pixel point in the optimized grayscale image of the laser seeker is the sum of the products of the grayscale values of the pixel points at the same positions in all grayscale images of the laser seeker and the fusion weights.

[0051] (5) Feed the optimized grayscale image of the laser seeker into the trained U²-Net network that can identify defects such as pockmarks and scratches. The loss function uses the binary cross-entropy loss function, and the optimizer in the network structure uses the Adam optimizer to output the defect segmentation image of the laser seeker containing defect labels such as pockmarks and scratches.

[0052] (6) When defects such as pockmarks and scratches appear in the defect segmentation image of the laser seeker, the corresponding radome of the laser seeker is determined to be unqualified; when no defects appear in the defect segmentation image of the laser seeker, the corresponding radome of the laser seeker is determined to be qualified, and the visual intelligent detection of the laser seeker is completed.

[0053] Preferably, as an embodiment of the present application, the flowchart of the laser seeker defect recognition process is as Figure 3 shown.

[0054] It should be noted that the U²-Net network and the binary cross-entropy loss function are both well-known technologies, and their specific contents and operation processes will not be elaborated.

[0055] Preferably, as an embodiment of the present application, the schematic diagram of the laser seeker defect intelligent detection process is as Figure 4 shown.

[0056] It should be noted that: the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; modifying the technical solutions recorded in the foregoing embodiments, or equivalently replacing some of the technical features, does not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included within the protection scope of the present application.

Claims

1. A laser seeker visual intelligent detection method based on computer vision, characterized in that: The method comprises the following steps: Get the grayscale image of the laser seeker under different light source angles; Determine the side length of the spot detection window of each pixel point according to the distribution of the gray value of each pixel point in the gray image of the laser seeker; Determine the suspected spot index of each pixel point according to the degree of disorder of the grayscale values ​​of all pixels in the spot detection window of each pixel point and the difference between the grayscale value of each pixel point and the grayscale values ​​of all pixels in the spot detection window; Based on the side length of the spot detection window of each pixel point and the distribution of all pixels around the central pixel point therein, the spot confidence index of each pixel point is determined in combination with the suspected spot index to detect the laser guide head.

2. The computer vision-based laser seeker visual intelligent detection method according to claim 1, characterized in that: The expression of the side length of the spot detection window of each pixel is: ; In the formula, Represents the side length of the spot detection window of the i-th pixel in the grayscale image of the laser seeker; Represents the gray value of the i-th pixel in the gray image of the laser seeker; and They represent the preset reference window length and the preset minimum window length respectively; ceil() represents the upward rounding function; norm() represents the normalization function.

3. The computer vision-based laser seeker visual intelligent detection method according to claim 1, characterized in that: The process of determining the suspected spot index of each pixel is as follows: Determine the disorder degree of each pixel point according to the disorder degree of the grayscale values ​​of all pixels in the spot detection window of each pixel point; Determine the suspected light spot degree of each pixel point according to the difference between the gray value of each pixel point and the gray value of all pixels in its light spot detection window; The suspected light spot index of each pixel is determined based on the disorder degree and the suspected light spot degree.

4. The computer vision-based laser seeker visual intelligent detection method according to claim 3, characterized in that: The chaos degree of each pixel is the image entropy of the light spot detection window where each pixel is located.

5. The computer vision-based laser seeker visual intelligent detection method according to claim 3, characterized in that: The suspected light spot degree of each pixel is the ratio of the gray value of each pixel to the minimum gray value of all pixels in its light spot detection window.

6. The computer vision-based laser seeker visual intelligent detection method according to claim 1, characterized in that: The suspected light spot index of each pixel is the product of the disorder degree of each pixel and the suspected light spot.

7. The computer vision-based laser seeker visual intelligent detection method according to claim 1, characterized in that: The determination process of the light spot confidence index of each pixel point is as follows: Determine the spot area ratio of each pixel based on the side length of the spot detection window of each pixel and the distribution of all pixels around the central pixel within the window; The spot confidence index of each pixel is determined based on the spot area ratio and the suspected spot index.

8. The computer vision-based laser seeker visual intelligent detection method according to claim 7, characterized in that: The process of determining the spot area ratio of each pixel is as follows: Clustering all the pixels in the spot detection window of each pixel to obtain multiple clusters; The spot area ratio of each pixel is the ratio of the number of all pixels in the cluster where the central pixel in the spot detection window of each pixel is located to the square of the side length of the spot detection window.

9. The computer vision-based laser seeker visual intelligent detection method according to claim 7, characterized in that: The light spot confidence index of each pixel is the product of the suspected light spot index of each pixel and the light spot area.

10. The computer vision-based laser seeker visual intelligent detection method according to claim 1, characterized in that: The method of detecting the laser guide head comprises: The spot confidence index of all pixels in the grayscale image of the laser seeker is used as the input of the threshold segmentation algorithm to obtain the segmentation threshold; The pixels in the laser seeker grayscale image whose light spot confidence index is less than or equal to the segmentation threshold are recorded as valid pixels; The mean of the grayscale values ​​of all valid pixels in the spot detection window of each pixel is used as the fusion weight of each pixel; The gray value of each pixel in the laser seeker optimization gray image is the cumulative sum of the gray values ​​of the pixels at the same position in all the laser seeker gray images and the product of the fusion weight; The U²-Net neural network model is used to detect defects in the optimized grayscale image of the laser seeker.

Images