A method and system for satellite component contour extraction for a space lightweight platform

Abstract

The present invention discloses a method and system for satellite component contour extraction for a space lightweight platform, which uses a depth map as the data source to avoid non-contour edges introduced by satellite component textures and complex spatial lighting environments; at the same time, the edges extracted by the improved Canny algorithm and the surface normal edges are combined as two clues for contour extraction, so that the obtained satellite component contours have good integrity, solving the problem that the current traditional methods cannot effectively complete the task of satellite component contour extraction; having contour extraction performance comparable to that of learning-based methods, and at the same time having real-time working efficiency under a consumer-grade CPU platform, laying a solid foundation for the deployment of a space lightweight platform. In addition, in complex scenarios such as close-range occlusion of similar satellite components, the present invention has stronger robustness compared with learning-based methods.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and relates to a method and system for satellite component contour extraction for a space lightweight platform. Background Art

[0002] The tasks of satellite on-orbit assembly, maintenance, modeling, and capture are both difficult and expensive. Satellite component contour extraction is the basis for an autonomous mobile platform to complete the above tasks. Therefore, how to achieve real-time extraction of satellite component contours on a lightweight space mobile platform is an important research topic.

[0003] Contour extraction methods can generally be divided into traditional methods and deep learning-based methods. Among traditional methods, pixel-based methods do not require complex initialization settings and require less computational effort, so they have been widely used in practical applications. The core idea of pixel-based methods is to first construct pixel features and then determine whether a pixel belongs to a certain contour based on the features. The discontinuity of gray intensity is the most prominent feature in an image and is also the first feature to be discovered. Subsequently, various filters were introduced, such as Roberts, Sobel, Laplacian, Log, and Canny operators. Roberts and Sobel belong to first-order differential operators. Roberts obtains edges through local differences. However, Roberts is sensitive to noise and easily loses edges. Sobel achieves better edge detection by using the weighted difference of gray values of adjacent points, and compared with Roberts, Sobel is less sensitive to noise. Laplacian is an isotropic detector, sensitive to noise, and may produce double responses. Log first uses a Gaussian filter to smooth the picture and then processes it with a Laplacian sharpening filter. Log overcomes the influence of noise to a certain extent, but its positioning accuracy for curved edges is not ideal. Canny is a multi-criterion edge detector that includes finding the most edges by minimizing the error rate, marking edges as close as possible to the actual edge, and only marking an edge once when a single edge exists to obtain the minimum response. Compared with other operators, Canny has excellent performance in almost all aspects. Therefore, until now, it has still been widely used in various fields.

[0004] Although the Canny algorithm has excellent performance, when faced with the task of satellite component contour extraction in space, it still faces the following challenges: Due to factors such as illumination changes and texture ambiguity, it is difficult for the Canny algorithm to strike a balance between removing non-contour edges and retaining contour edges, as Figure 3 shown, that is, it cannot effectively complete the task of satellite component contour extraction.

[0005] With the rise of artificial intelligence, deep learning-based methods have become a research hotspot. DeepContour is a patch-based method that first divides an image into multiple patches and then feeds these patches into a CNN to detect whether the patches have contours. HED is an end-to-end method that directly inputs an image and outputs an edge map. It can generate high-quality results by using weighted cross-entropy loss and skip layer structure. RCF modifies the skip layer architecture of HED and achieves better results. CASENet uses an end-to-end network to assign each edge pixel to one or more semantic labels. LPCB proposes a method that allows a CNN to predict clear boundaries. Compared with traditional methods, the above learning-based methods have better performance, but they are still difficult to effectively distinguish internal non-contour edges and contour edges. For this reason, Zhu Xiaoxi et al. from the National University of Defense Technology tried to extract the contours of each part of the human body by using the color information of the semantic segmentation results, which effectively removed the internal edges. However, applying contour extraction methods based on semantic segmentation in space will bring another tricky problem, that is, it is difficult for them to get rid of the dependence on the Graphics Processing Unit (GPU) and achieve a real-time working rate, which is not friendly to lightweight space mobile platforms with limited computing resources, size, and energy storage.

[0006] In summary, for a lightweight space mobile working platform, traditional methods cannot effectively complete the contour extraction task, and deep learning-based methods are difficult to be deployed on mobile platforms with limited computing resources. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems in the prior art and provide a satellite component contour extraction method and system for a space lightweight platform.

[0008] To achieve the above purpose, the present invention adopts the following technical solutions:

[0009] In the first aspect, the present invention provides a satellite component contour extraction method for a space lightweight platform, including the following steps:

[0010] Extract the discontinuous edges of the satellite component depth image to obtain a depth contour edge binary map;

[0011] Extract the surface normal edges of the satellite component depth image to obtain a normal contour edge binary map;

[0012] Perform morphological processing on the depth contour edge binary map and the normal contour edge binary map to obtain the satellite component contour.

[0013] In the second aspect, the present invention provides a satellite component contour extraction system for a space lightweight platform, including:

[0014] A depth contour edge extraction module, configured to extract discontinuous edges of the depth image of the satellite component to obtain a depth contour edge binary map;

[0015] A normal contour edge extraction module, configured to extract surface normal edges of the depth image of the satellite component to obtain a normal contour edge binary map;

[0016] A morphological processing module, configured to perform morphological processing on the depth contour edge binary map and the normal contour edge binary map to obtain the satellite component contour.

[0017] In a third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0018] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] The present invention uses the depth map as the data source to avoid non-contour edges introduced by the texture of the satellite component and the complex spatial illumination environment; at the same time, the edges extracted by the improved Canny algorithm and the surface normal edges are combined as two clues for contour extraction, so that the obtained satellite component contour has good integrity, solving the problem that the current traditional methods cannot effectively complete the task of satellite component contour extraction; it has contour extraction performance comparable to that of learning-based methods, and at the same time has real-time working efficiency on a consumer-grade CPU platform, laying a solid foundation for the deployment of a space lightweight platform. In addition, in complex scenarios such as close-range occlusion of similar satellite components, the present invention has stronger robustness compared with learning-based methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a flowchart of the method of the present invention.

[0023] Figure 2 It is a schematic diagram of the system of the present invention.

[0024] Figure 3This is the overall flowchart of the satellite component contour extraction method of the present invention.

[0025] Figure 4 Schematic diagram for solving neighborhood points and normal vectors of target points.

[0026] Figure 5 Contour extraction results of Scheme 1 under different high and low thresholds.

[0027] Figure 6 Contour extraction results of Scheme 2 under different high and low thresholds.

[0028] Figure 7 Contour extraction results of the algorithm of Scheme 3.

[0029] Figure 8 Merged edge binary image.

[0030] Figure 9 Final contour map of the method of the present invention.

[0031] Figure 10 Comparison chart of contour extraction effects of three schemes.

[0032] Figure 11 Comparison chart of contour extraction effects of three schemes in complex scenes. Detailed implementation manners

[0033] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations.

[0034] Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0035] It should be noted that: similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0036] In the description of the embodiments of the present invention, it should be noted that if terms such as "upper", "lower", "horizontal", "inner", etc. are used to indicate the orientation or positional relationship, it is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the invention product is usually placed during use. This is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. In addition, terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0037] In addition, if the term "horizontal" appears, it does not mean that the component is required to be absolutely horizontal, but it can be slightly inclined. For example, "horizontal" only means that its direction is more horizontal relative to "vertical", and does not mean that the structure must be completely horizontal, but it can be slightly inclined.

[0038] In the description of the embodiments of the present invention, it should also be noted that unless otherwise clearly specified and limited, if terms such as "set", "installed", "connected", "connected" are used, they should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0039] The following further describes the present invention in detail with reference to the drawings:

[0040] See Figure 1 , the embodiments of the present invention disclose a method for extracting the contour of a satellite component for a space lightweight platform, including the following steps:

[0041] S1 Extract the discontinuous edges of the depth image of the satellite component to obtain a binary depth contour edge map; specifically as follows:

[0042] S1-1, Input the depth image D of the satellite component as the data source into the bilateral filter to remove the depth image noise while retaining the edge details;

[0043] S1-2, Calculate the gradient magnitude and direction of the depth image D of the satellite component to obtain the gradient edge;

[0044] S1-3, Use the non-maximum suppression method to refine the gradient edge;

[0045] S1-4, Select the high and low thresholds (15, 40), retain the pixel points with a gradient value greater than 40, discard the pixel points with a gradient value lower than 15, and determine the pixel points between the high and low thresholds through an 8-connected domain; finally generate a binary depth contour edge map Ecanny 。

[0046] S2 extracts the surface normal edges of the satellite component depth image to obtain a binary image of the normal contour edges; specifically as follows:

[0047] S2-1, transforms the satellite component depth image D into a vertex map V through formula (1):

[0048]

[0049] where K is the camera internal parameter matrix, u is the image pixel unit (x, y), is its homogeneous expression form;

[0050] S2-2, solves the normal map to obtain the normal vector Specifically as follows:

[0051] The normal of each target point P is obtained by taking the arithmetic mean of the normal values of the two planes spanned by its eight adjacent vertices N, S, W, E, NE, SW, NW, SS, and the value of each vertex is the average of its three adjacent vertices to reduce the influence of noise; the calculation method of the vertex N above the target point P is as follows:

[0052]

[0053] Similarly, solve the other seven vertices, and the normal of the target point P The calculation method is as follows:

[0054]

[0055]

[0056]

[0057] where, and are the normals of the two planes spanned by the four vertices respectively, is the average value of the two, is the vector form of the vertex N above the target point, is the vector form of the vertex S below the target point, is the vector form of the vertex W on the left of the target point, is the vector form of the vertex E on the right of the target point, is the vector form of the vertex in the upper right corner of the target point, is the vector form of the vertex in the lower left corner of the target point, is the vector form of the vertex in the upper left corner of the target point, is the vector form of the vertex in the lower right corner of the target point.

[0058] S2-3. Normalize the normal vector to obtain the unit normal vector

[0059] S2-4. Solve the eight-neighborhood normal factors of each pixel as follows:

[0060]

[0061] where \(i = 1, 2, 3, \ldots, 8\), \(v(u)\) and \(n(u)\) are the vertex and normal corresponding to the target pixel, and \(v(u i ) and \(n(u i ) are the eight-neighborhood vertices and eight-neighborhood normals of \(v(u)\) and \(n(u)\) respectively.

[0062] S2-5. Select the minimum normal factor within the eight-neighborhood range of each pixel as the final normal factor of the target point thus generating a normal factor map;

[0063]

[0064] S2-6. Perform a binarization operation on the normal factor map to obtain the surface normal edge, and obtain the final binary map \(E\) of the normal contour edge normal .

[0065] S3. Perform morphological processing on the depth contour edge binary map and the normal contour edge binary map to obtain the satellite component contour, specifically as follows:

[0066] S3-1. Perform an OR operation on the depth contour edge binary map \(E canny and the normal contour edge binary map \(E normal pixel by pixel to obtain a merged edge map \(E combined ;

[0067] S3-2. Perform an erosion morphological operation on the merged edge map \(E combined ;

[0068] S3-3. Perform a dilation morphological operation on the edge map \(E combined after the erosion morphological operation, and use this result as the final satellite component contour extraction result and output it.

[0069] As Figure 2 shown, an embodiment of the present invention discloses a satellite component contour extraction system for a space lightweight platform, including:

[0070] A depth contour edge extraction module for extracting the discontinuous edges of the satellite component depth image to obtain a depth contour edge binary map;

[0071] The normal contour edge extraction module is used to extract the surface normal edge of the depth image of the satellite component to obtain a binary normal contour edge map;

[0072] The morphological processing module is used to perform morphological processing on the depth contour edge binary map and the normal contour edge binary map to obtain the satellite component contour.

[0073] Embodiment

[0074] Figure 3 is the overall flowchart of the satellite component contour extraction method of the present invention. The input data type of the present invention is a depth image rather than a color image, and the final contour is composed of two parts of edge clues. One of the edge clues comes from the edge extraction result of the improved Canny algorithm, and the other edge clue comes from the surface normal edge extraction result, and the two edge clue extraction processes are parallel. When the two edge clues are merged, the final satellite component contour can be obtained through simple morphological operations. The overall scheme can be divided into three parts, and the specific process is as follows:

[0075] Step 1. Extract the edge of depth discontinuity:

[0076] Step 1-1, Take the depth image D as the data source and input it to the bilateral filter to remove the depth image noise while retaining more edge details.

[0077] Step 1-2, Calculate the gradient magnitude and direction of the depth map to obtain the gradient edge.

[0078] Step 1-3, Use the non-maximum suppression method to refine the gradient edge.

[0079] Step 1-4, Select high and low thresholds (15, 40). Pixel points with a gradient value greater than 40 are retained, and pixel points with a gradient value lower than 15 are discarded. For pixel points between the high and low thresholds, they are determined by the 8-connected domain. Finally, a depth contour edge binary map E canny (white background) is obtained, as shown in Figure 6 shown.

[0080] Step 2. Extract the surface normal edge (as shown in Figure 4 )

[0081] Step 2-1, Transform the depth map D into a vertex map V through formula (1). Where K is the camera internal parameter matrix, u is the image pixel unit (x, y), and

[0082]

[0083] Step 2-2: Solve the normal map. The normal of each target point P is obtained by averaging the normals of two planes spanned by its eight adjacent vertices N, S, W, E, NE, SW, NW, and SE. Moreover, each vertex takes the average value of its three adjacent vertices to reduce the influence of noise, as shown in Figure 5 . The calculation formula for the vertex N above the target point P is shown in Formula (2). Similarly, the other seven vertices are solved. The calculation formula for the normal of the target point P is completed jointly by Formulas (3), (4), and (5), where and and are the normals of the planes spanned by four vertices respectively, and takes the average value of the two, which can make the normal of the target point smoother and reduce the interference of noise.

[0084]

[0085]

[0086]

[0087]

[0088] Step 2-3: Normalize the normal vector to obtain the unit normal vector

[0089] Step 2-4: Solve the normal factor of the eight-neighborhood of each pixel The normal factor is designed as shown in Formula (6), where v(u) and n(u) are the vertex and normal corresponding to the target pixel, and v(u i ) and n(u i ) are the eight-neighborhood vertices and eight-neighborhood normals of v(u) and n(u) respectively.

[0090]

[0091] The normal factor defined above can not only extract edge pixels but also reduce the number of non-contour edge pixels.

[0092] Step 2-5: Use Formula (7) to select the smallest normal factor within the eight-neighborhood of each pixel as the final normal factor of the target point thus generating the normal factor map.

[0093]

[0094] Step 2-6: Binarize the normal factor map obtained in Step 5 to obtain the surface normal edge. The threshold a = 0.7. Pixel values higher than the threshold are set to 0, and pixel values lower than the threshold are set to 255. The final binary map E of the normal contour edge normal (white background), as Figure 7 shown.

[0095] Step 3. Merge the above two edges and perform morphological operations

[0096] Step 3-1: Perform an OR operation on the obtained E canny and E normal pixel by pixel to obtain the merged edge map E comdined , as Figure 8 shown.

[0097] Step 3-2: Perform an erosion morphological operation on the merged edge map E combined to ensure the closure of the satellite component contour.

[0098] Step 3-3: Perform a dilation morphological operation on the result of Step 2 to refine the edge and improve the edge extraction accuracy. Output this result as the final satellite component contour extraction result, as Figure 9 shown.

[0099] The input data in this embodiment is a depth image, rather than a color image. The contour extraction uses two cues, depth discontinuity and normal discontinuity, to form the target contour. Finally, the two edge extraction processes are parallel.

[0100] The present invention uses a depth image as the original data for contour extraction, solving the problem of a large number of non-contour edge interferences caused by using a color image as the original image; and uses depth discontinuity edges and surface normal edges as two cues for contour extraction, solving the defect of incomplete depth discontinuity edges alone and improving the robustness of contour extraction.

[0101] Example:

[0102] Use the publicly available dataset AFDL-SCD as the data for testing the present invention, and compare it with the contour extraction methods based on MaskR-CNN and SOLOv2. At the same time, conduct a robustness comparison test under the complex scenario of close-range occlusion of the same satellite components. The AFDL-SCD dataset contains 1340 images, including 670 color images and 670 corresponding depth images. The resolution of the dataset images is 640×480. The dataset is randomly divided into a training set and a test set in a ratio of 4:1. The training set is used for the model training of MaskR-CNN and SOLOv2, and the test set is used for the contour extraction experimental comparison of the three schemes. The satellite component contour extraction metrics include four commonly used quantitative evaluation metrics in addition to qualitative analysis:

[0103]

[0104]

[0105]

[0106]

[0107] Among them, P is the precision of contour detection, R is the recall rate, F1 is the F1 score, and Q is the working frame rate in frames per second. TP represents the number of pixels correctly identified as edges, FN represents the number of pixels misidentified as the background, and FP represents the number of pixels misidentified as edge pixels.

[0108] 1. Comparative experiments in general scenarios

[0109] To verify the effectiveness of this method, six groups of experimental schemes are set in this paper according to different input data and adopted algorithms, as shown in Table 1.

[0110] Table 1

[0111] solution input data algorithm one color map improved Canny algorithm two depth map improved Canny algorithm three depth map surface normal edge detection algorithm four depth map the present invention five color map deep learning - MASKR - CNN six color map deep learning - SOLOv2

[0112] Randomly select an image in the test set as an example, among which Figure 5 are the contour extraction results of Scheme 1 under different high and low thresholds. The results show that when the color map is used as the data source, due to the texture, color, and illumination changes of satellite components, the edge detection results contain a large number of non-contour detailed edges. Even by adjusting the threshold, it is still very difficult to balance between removing detailed edges and retaining contour edges. Scheme 2 uses the depth image as the original data, and the contour extraction results are as Figure 6 shown. According to the results, it can be seen that using the depth image for edge extraction can easily avoid the detailed edges caused by factors such as component texture, color, or illumination, and is more suitable for the satellite component contour extraction task. However, at the same time, it can be seen that with the change of the threshold, the satellite component contours also appear incomplete to varying degrees. Figure 7 is the contour detection result of Scheme 3, that is, the surface normal edge. Generally speaking, this scheme can outline all the contours of satellite components, but the obtained contours are all discontinuous. The present invention combines the methods of Scheme 2 and Scheme 3, and the merged contour edges are as Figure 8 shown. In this way, the integrity of the contour edges can be ensured through two aspects of edge clues, and at the same time, the final satellite component contour result can be obtained only after one erosion and dilation operation, as Figure 9 shown.

[0113] To facilitate comparison with two deep learning-based methods, we used the connected component labeling method to color-label the satellite component contours obtained by the method of the present invention. At the same time, we used the color information of the MaskR-CNN and SOLOv2 instance segmentation results to extract the contours of the satellite components as the contour extraction results, as Figure 10 shown. To obtain more persuasive conclusions, contour extraction experiments were completed for all 134 images in the test set on a consumer-grade CPU (i7 4910MQ) platform, and the performance index data are statistically shown in Table 2. Overall, the performance of the methods based on MaskR-CNN and SOLOv2 is very close. Compared with the above two deep learning methods, the precision rate of the method of the present invention is slightly lower, the recall rate is relatively higher, the F1 score is slightly higher, and the speed is much higher than that of the deep learning-based scheme.

[0114] Table 2

[0115]

[0116]

[0117] Reason analysis: The reason for the slightly lower precision rate of the present invention is that the contours in the present invention are derived from two edge clues and are subjected to morphological operations, so the final contour edges are relatively thick, resulting in some non-real contour pixels in the final contour, so the precision is slightly lower. However, also because of this, the thicker contours contain more real contour pixels, which results in a higher recall rate. Therefore, according to formula (8), the F1 score is slightly higher. Since the two deep learning-based methods contain a large number of parameters and complex network structures, their operating efficiency is very low and they can hardly work in real time on the CPU platform. In contrast, the method in the present invention is simple and effective and can achieve an operating efficiency of 36.5 FPS on the CPU platform.

[0118] 2. Comparison of contour extraction effects in complex scenarios

[0119] For the same type of different individual components (taking the solar panel as an example in the present invention), the distance with a ratio of the vertical distance between two solar panels to the image acquisition distance less than 0.01 is defined as a short distance. Three scenarios with different occlusion ratios at short distances were set for experimental comparison, and the results are as Figure 11As shown. The first column is the input image, and the other three columns are the target partial enlarged images so that readers can observe clear details. First, for the first two input pictures, since the colors and textures of the two solar panels are very close, the deep learning-based method cannot distinguish the two solar panels individually. Therefore, the extracted contour combines the contours of the two solar panels. For the third image with the smallest occlusion ratio, although MaskR-CNN distinguishes the two solar panels (for easy distinction, one of the solar panels is marked in magenta), the extracted contour has a large error. SOLOv2 still cannot distinguish the two solar panels and shows the phenomenon of incomplete contour extraction. In contrast, the method of the present invention can distinguish the two solar panels individually in all three cases of close-range occlusion, which benefits from the introduction of the surface normal edge, as shown in Figure 11 the last column, making up for the deficiency that the improved Canny algorithm cannot effectively extract the contours of targets occluding each other at close range. Therefore, in complex scenarios such as close-range occlusion of the same or similar targets, the method of the present invention also has certain advantages.

[0120] A computer device provided by an embodiment of the present invention. The computer device of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of each module / unit in the above-mentioned device embodiments are implemented.

[0121] The computer program can be divided into one or more modules / units. The one or more modules / units are stored in the memory and executed by the processor to complete the present invention.

[0122] The computer device can be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device may include, but is not limited to, a processor and a memory.

[0123] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0124] The memory can be used to store the computer program and / or module. By running or executing the computer program and / or module stored in the memory, and invoking the data stored in the memory, the processor implements various functions of the computer device.

[0125] If the modules / units integrated in the computer device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0126] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for extracting the contour of satellite components for a space lightweight platform, characterized in that, Including the following steps: Extract the discontinuous edges of the depth image of the satellite component to obtain a binary image of the depth contour edges. The specific method is as follows: Step 1, input the depth image of the satellite component as the data source into the bilateral filter to remove the depth image noise while retaining the edge details; Step 2, calculate the depth image of the satellite component to obtain the gradient magnitude and direction, and get the gradient edge; Step 3: Refine the gradient edges using the non-maximum suppression method; Step 4: Select the high and low thresholds (15, 40), retain the pixels with gradient values greater than 40, discard the pixels with gradient values lower than 15, and determine the pixels between the high and low thresholds through 8-connected regions; finally, generate a binary depth contour edge map ; Extract the surface normal edges of the depth image of the satellite component to obtain a binary image of the normal contour edges. The specific method is as follows: Step 1, transform the depth image of the satellite component into a vertex map through formula (1) : Among them, is the camera internal parameter matrix, is the image pixel unit (x, y), is its homogeneous representation form; Step 2, solve the normal map to obtain the normal vector ; Step 3, for the normal vector perform normalization to obtain the unit normal vector ; Step 4, solve the octal neighborhood normal factor of each pixel ; Step 5: Select the minimum normal factor within the eight-neighborhood range of each pixel as the final normal factor of the target point , thereby generating a normal factor map; Step 6, perform a binarization operation on the normal factor graph to obtain the surface normal edge, and obtain the final binary image of the normal contour edge ; Perform morphological processing on the binary image of the depth contour edges and the binary image of the normal contour edges to obtain the satellite component contour.

2. The method for extracting the satellite component profile for a space lightweight platform according to claim 1, wherein Solving the normal map to obtain a normal vector , including: Each target point has a normal vector obtained by taking the arithmetic mean of the normal vectors of the two planes spanned by its eight adjacent vertices , and the value of each vertex is the average of its three adjacent vertices to reduce the influence of noise; the upper vertex of the target point is calculated as follows: Similarly, solve for the other seven vertices, the target point The normal vector of is calculated as follows: Among them, and are the normal vectors of the planes spanned by the four vertices respectively, is the average value of the two, is the vector form of the vertex above the target point N ; is the vector form of the vertex below the target point S ; is the vector form of the vertex to the left of the target point W ; is the vector form of the vertex to the right of the target point E ; is the vector form of the vertex at the upper right corner of the target point, is the vector form of the vertex at the lower left corner of the target point, is the vector form of the vertex at the upper left corner of the target point, is the vector form of the vertex at the lower right corner of the target point.

3. The method for extracting the satellite component profile for a space lightweight platform according to claim 1 or 2, characterized in that The normal factor , is as follows: Among them, , and are the vertex and normal corresponding to the target pixel, and are respectively and the octahedral neighborhood vertices and octahedral neighborhood normals of 4. The method for extracting the satellite component profile for a space lightweight platform according to claim 1, wherein The performing morphological processing on the binary image of the depth contour edges and the binary image of the normal contour edges to obtain the satellite component contour includes: Step 1, perform an OR operation on the depth contour edge binary image and the normal contour edge binary image pixel by pixel to obtain a combined edge image ; Step 2, perform an erosion morphological operation on the merged edge map ; Step 3, perform a dilation morphological operation on the edge map that has completed the erosion morphological operation and use this result as the final satellite component contour extraction result and output it.

5. A satellite component contour extraction system for a space lightweight platform for implementing the method described in claim 1, characterized in that, Including: A depth contour edge extraction module, configured to extract the discontinuous edges of the depth image of the satellite component to obtain a binary image of the depth contour edges; A normal contour edge extraction module, configured to extract the surface normal edges of the depth image of the satellite component to obtain a binary image of the normal contour edges; A morphological processing module, configured to perform morphological processing on the binary image of the depth contour edges and the binary image of the normal contour edges to obtain the satellite component contour.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1-4 are implemented.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1-4 are implemented.

Images