An optical guidance method for the recovery of deep-sea unmanned underwater vehicles

Abstract

This invention provides an optical guidance method for the recovery of deep-sea unmanned underwater vehicles (UUVs), belonging to the field of underwater vehicle recovery technology. The invention involves installing four blue-green LED guidance light sources in a non-uniform ring distribution on the front face of the recovery cage and pre-calibrating their three-dimensional coordinates. A binocular camera is used to acquire images of the guidance light sources in real time. Adaptive exposure control technology is employed to adjust image quality. Statistical information calculations are performed on the identified light source contours, and effective guidance light source targets are determined through convex constraint configuration judgment and ellipse fitting. The four guidance light sources are constructed as a traveling salesman problem, and a nearest neighbor heuristic algorithm is used to complete numbering matching. Based on pre-measured coordinates and real-time calculated coordinates, the rotation matrix and translation vector are solved using singular value decomposition, while simultaneously performing collaborative optimization. This method solves the technical problems of insufficient optical guidance positioning accuracy and poor real-time performance during the recovery of UUVs in deep-sea environments.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of underwater vehicle recovery, and in particular, relates to an optical guiding method for deep-sea unmanned underwater vehicle recovery. BACKGROUND

[0002] Deep-sea unmanned underwater vehicle recovery is a key technical link in marine engineering. Traditional recovery guiding methods mainly use acoustic guiding and magnetic guiding technologies. By setting acoustic beacons or magnetic markers on the recovery cage, the rough positioning and guiding of the vehicle are achieved by using sound propagation or magnetic field changes. These methods have been widely used in deep-sea complex environments and provide a basic technical guarantee for the safe recovery of unmanned underwater vehicles. However, the traditional acoustic guiding method has obvious limitations in the precise positioning stage. The propagation of acoustic signals in seawater is easily affected by factors such as sea current disturbance, temperature layer change, and seabed reflection, resulting in a decrease in positioning accuracy. At the same time, the response speed of the acoustic system is relatively slow, making it difficult to meet the real-time control requirements in the dynamic recovery process. The magnetic guiding method has certain anti-interference ability, but its action distance is limited and it is easily affected by the seabed geomagnetic environment. In the existing technology, when the unmanned underwater vehicle approaches the recovery cage for final docking, due to the lack of high-precision close-range guiding means, the vehicle cannot accurately identify the precise position and attitude information of the recovery cage, resulting in a low recovery success rate and prolonged recovery time. At the same time, the robustness of the traditional guiding method in complex marine environments is insufficient and is easily affected by factors such as water turbidity, light changes, and sea current disturbance, affecting the reliability and efficiency of the entire recovery system. That is, the existing technology has the technical problem of insufficient optical guiding positioning accuracy and poor real-time performance in the recovery process of unmanned underwater vehicles in deep-sea environments. SUMMARY

[0003] Therefore, the present application provides an optical guiding method for deep-sea unmanned underwater vehicle recovery, which can solve the technical problem of insufficient optical guiding positioning accuracy and poor real-time performance in the recovery process of unmanned underwater vehicles in deep-sea environments in the existing technology.

[0004] The application is implemented in the following manner: the application provides an optical guiding method for deep-sea unmanned underwater vehicle recovery, four blue-green LED guiding light sources are installed on the front end of the recovery cage, three-dimensional spatial coordinate information of the guiding light sources is measured and stored by an industrial photogrammetry system, binocular cameras on the bow of the unmanned underwater vehicle are internally and externally parameterized, and upper and lower game models are established; the unmanned underwater vehicle is guided to a distance range of 15 m from the recovery cage through acoustic guidance, the binocular camera image acquisition system is started and the guiding light source target is detected; an adaptive exposure control mathematical model is established, images are acquired and the guiding light source target is segmented to calculate the pixel size; the guiding light source images are synchronously acquired by the binocular camera, the Rolling Ball method is used to remove background light interference, and a maximum value mask image and a light spot contour mask are generated through contrast automatic enhancement processing; statistical information of the recognized light source contour is calculated, four effective guiding light source targets are determined through convex constraint configuration judgment and long-short axis ratio screening of elliptical fitting, the four guiding light source targets are constructed as traveling salesman problem nodes, and the nearest neighbor heuristic algorithm is used to complete the guiding light source number matching; the center of the guiding light source is accurately positioned by using the barycenter method, the homonym matching of the guiding light source center point in the binocular image is completed through the principle of epipolar constraint, and the three-dimensional coordinates of the guiding light source center are calculated based on the binocular vision measurement principle; the relative pose relationship between the unmanned underwater vehicle and the recovery cage is calculated by solving the rotation matrix and the translation vector through the singular value decomposition method according to the pre-measured three-dimensional coordinate information of the guiding light source and the calculated three-dimensional coordinates of the guiding light source in the binocular coordinate system.

[0005] Among them, the four blue-green LED guiding light sources are installed in a non-uniform ring shape, the upper game model aims to maximize the recovery success rate, the lower game model aims to minimize the system energy consumption, and the lower game model is started when the guiding light source is detected.

[0006] Among them, the adaptive exposure control mathematical model specifically initializes the exposure value of the binocular camera to 25 ms, adjusts the exposure value to the original value multiplied by 1.20 if the pixel size is less than 25 pixels, and adjusts the exposure value to the original value multiplied by 0.75 if the pixel size is greater than 80 pixels.

[0007] Among them, the industrial photogrammetry system is a high-precision three-dimensional coordinate measurement device based on the principle of stereo vision, which photographs the target through multiple calibration cameras and calculates the three-dimensional coordinates of the space points by using the principle of triangulation, with a measurement accuracy of sub-millimeter level.

[0008] Among them, the Rolling Ball method is an image background removal technique based on morphological operation, which effectively separates the foreground target and the background light uneven area by simulating the rolling process of the ball on the image surface.

[0009] Specifically, the automatic contrast enhancement process involves calculating the minimum grayscale value of the image. and the maximum gray value of the image Calculate the grayscale value of the enhanced image pixel by pixel. Where (x, y) are the pixel coordinates. This represents the grayscale of the original image.

[0010] Specifically, the steps for calculating the statistical information include obtaining the center coordinates, contour area, average gray level of the contour, length of the major axis of the ellipse fitting, and length of the minor axis of the ellipse fitting.

[0011] The step of calculating the three-dimensional coordinates of the guiding light source in the binocular coordinate system further includes defining the centroids of two sets of three-dimensional point sets and calculating the centroid-free coordinates. Specifically, the centroid-free coordinates are the relative coordinates obtained by subtracting the centroid coordinates of the corresponding point set from the three-dimensional point coordinates, calculated using the following formula: and ,in and These are the centroids of the two point sets, respectively.

[0012] Specifically, the singular value decomposition method is a mathematical method in linear algebra that decomposes a matrix into a product of three matrices, by constructing a matrix... Singular value decomposition of matrix W yields Where U and V are orthogonal matrices, For diagonal matrices, rotation matrices .

[0013] Specifically, the translation vector is a three-dimensional vector describing the translation relationship between two coordinate systems, and its calculation formula is as follows: .

[0014] Specifically, the maximum value mask image is initialized as an all-zero matrix with the same size as the original image. For each pixel, if its grayscale value is not less than the grayscale value of pixels within an 8-connected region, the pixel is considered a maximum value and marked as 1 at the corresponding position. The Traveling Salesman Problem is a classic combinatorial optimization problem that optimizes the matching order of guide light source numbers by finding the shortest path to all nodes. The nearest neighbor heuristic algorithm is a greedy algorithm for solving the Traveling Salesman Problem, starting from any node and selecting the nearest unvisited node as the next node to be visited.

[0015] The gravity center method is specifically a method for determining the target center position by calculating the weighted average of all pixel coordinates in the target region. The epipolar constraint principle is specifically a basic geometric constraint relationship in binocular stereo vision, which uses epipolar geometry theory to constrain the search range of corresponding points in left and right images. The binocular vision measurement principle is specifically to use two cameras to observe the same target from different angles, and calculate the three-dimensional coordinates of the target by triangulation principle.

[0016] The application solves the technical problems of insufficient positioning accuracy and poor real-time performance of traditional guiding methods by establishing a binocular vision measurement system based on blue-green LED guide light sources and using a double-layer game model to optimize the recovery strategy. The application uses a non-uniform ring distribution design of four blue-green LED guide light sources, combines with the pre-calibrated three-dimensional spatial coordinate information of the industrial photogrammetry system, and uses the RollingBall background removal technology and contrast automatic enhancement processing to improve the accuracy and robustness of light source identification. At the same time, through adaptive exposure control and gravity center positioning technology, the measurement accuracy of the center coordinates of the guide light source is significantly improved. Based on the epipolar constraint principle, binocular stereo matching and triangulation principle are used to realize high-precision calculation of the three-dimensional coordinates of the guide light source. The relative pose relationship between the vehicle and the recovery cage is solved by the singular value decomposition method, providing reliable spatial positioning information for accurate recovery. In summary, the application solves the technical problems of insufficient optical guiding positioning accuracy and poor real-time performance in the process of recovering the unmanned underwater vehicle in the deep sea environment mentioned in the background art. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The flowchart of the method of the application.

[0018] Figure 2 The overall scheme of the optical guiding system in Example 2.

[0019] Figure 3 The camera installation structure diagram in Example 2, including three sub-diagrams: sub-diagram A is a left view, sub-diagram B is a perspective view, and sub-diagram C is a top view.

[0020] Figure 4 The flowchart of the optical guiding method in Example 2.

[0021] Figure 5 The light source identification flowchart in Example 2. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the embodiments of the application clearer, the technical scheme in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application.

[0023] As Figure 1 shown is a flow chart of an optical guidance method for deep-sea unmanned underwater vehicle recovery provided by the application, the method comprises the following steps:

[0024] An optical guidance method for deep-sea unmanned underwater vehicle recovery, comprising the following steps:

[0025] S01, four blue-green LED guide light sources are installed on the front end surface of the recovery cage in a non-uniform ring shape, the three-dimensional spatial coordinate information of the guide light sources in the recovery cage body is measured by an industrial photogrammetry system and stored in the software processing system configuration file, and the binocular camera installed on the bow of the unmanned underwater vehicle is calibrated for internal and external parameters, and an upper game model with the goal of maximizing recovery success rate is established;

[0026] S02, the unmanned underwater vehicle is guided to the recovery cage within a distance range of 15m through acoustic guidance, the binocular camera image acquisition system is started and the guide light source target is detected, if the guide light source is detected, a lower game model with the goal of minimizing system energy consumption is started;

[0027] S03, the exposure value of the binocular camera is initialized to 25ms, the image is acquired and the guide light source target is segmented to calculate the pixel size, if the pixel size is less than 25 pixels, the exposure value is adjusted to the original value multiplied by 1.20, if the pixel size is greater than 80 pixels, the exposure value is adjusted to the original value multiplied by 0.75, and an adaptive exposure control mathematical model is established;

[0028] S04, the binocular camera is used to synchronously acquire the guide light source image, the Rolling Ball method is used to remove background light interference, the image statistical information including the minimum gray value of the image and the maximum gray value of the image is calculated through contrast automatic enhancement processing, the maximum value mask image and the light spot contour mask are generated;

[0029] S05, the statistical information of the recognized light source contour is calculated to obtain the center coordinates, the contour area, the contour average gray value, the length of the long axis of the ellipse fitting and the length of the short axis of the ellipse fitting, the four effective guide light source targets are determined through convex constraint configuration judgment and ellipse fitting long and short axis ratio screening, the four guide light source targets are constructed as traveling salesman problem nodes and the nearest neighbor heuristic algorithm is used to solve the shortest loop to complete the guide light source number matching;

[0030] S06, the barycentric method is used to realize the accurate positioning of the guide light source center and record the center coordinates, the homonym matching of the guide light source center points in the binocular image is completed through the principle of polar constraint, and the three-dimensional coordinates of the guide light source center are calculated based on the binocular vision measurement principle and the pre-calibrated internal and external parameters of the binocular camera;

[0031] S07. According to the pre-measured guide light source three-dimensional coordinate information and the calculated guide light source three-dimensional coordinates in the binocular coordinate system, the centroids of the two sets of three-dimensional point sets are defined and the centroid coordinates are calculated, the rotation matrix and the translation vector are solved by the singular value decomposition method, and the relative pose relationship between the unmanned underwater vehicle and the recovery cage is calculated.

[0032] The industrial photogrammetry system is a high-precision three-dimensional coordinate measurement device based on the principle of stereo vision. A plurality of calibration cameras are used to shoot the target, and the three-dimensional coordinates of the space points are calculated by using the principle of triangulation. The measurement accuracy reaches the sub-millimeter level.

[0033] The Rolling Ball method is an image background removal technique based on morphological operation, which effectively separates the foreground target and the uneven background light area by simulating the rolling process of the rolling ball on the image surface.

[0034] The contrast automatic enhancement processing is to calculate the minimum gray value of the image and the maximum gray value of the image , and calculate the gray value of the enhanced image pixel by pixel , wherein is the pixel point coordinate, is the original image gray value.

[0035] The maximum value mask image is initialized as a full zero matrix, and the image size is consistent with the original image. For each pixel point, if its gray value is not less than the gray value of the pixels in the eight-connected region, the pixel point is considered as a maximum value point and marked as 1 at the corresponding position.

[0036] The light spot contour mask is initialized to be consistent with the maximum value mask, and the connected region of the maximum value point in the maximum value mask image is searched based on the region growing method. Any maximum value pixel point is taken as the initial seed point to start region growing. If the pixel gray value and the seed point gray value differ by no more than a set threshold, it is added to the growing region.

[0037] The convex constraint configuration judgment refers to the geometric constraint condition for checking whether the four guide light source targets form a convex quadrilateral, to ensure that the distribution of the guide light source targets meets the geometric requirements of spatial positioning.

[0038] The long-short axis ratio of the ellipse fitting refers to the ratio of the length of the long axis of the ellipse fitting to the length of the short axis of the ellipse fitting after the contour is fitted with an ellipse.

[0039] The traveling salesman problem is a classic combinatorial optimization problem, which optimizes the matching order of the guide light source number by finding the shortest path to visit all nodes.

[0040] The nearest neighbor heuristic algorithm is a greedy algorithm for solving the traveling salesman problem, which starts from an arbitrary node and selects the nearest unvisited node to the current node as the next visited node each time.

[0041] The epipolar constraint principle is a basic geometric constraint relationship in binocular stereo vision, which uses epipolar geometry theory to constrain the search range of corresponding points in left and right images.

[0042] The barycenter method is a method for determining the target center position by calculating the weighted average of all pixel coordinates in the target region.

[0043] The binocular vision measurement principle is to observe the same target from different angles by using two cameras, and calculate the three-dimensional coordinates of the target by the principle of triangulation.

[0044] The decentered coordinates refer to the relative coordinates obtained by subtracting the corresponding point set centroid coordinates from the three-dimensional point coordinates, and the calculation formula is and , where and are the centroids of the two sets of points, respectively.

[0045] The singular value decomposition method is a mathematical method in linear algebra that decomposes a matrix into the product of three matrices, by constructing a matrix , and performing singular value decomposition on the matrix W to obtain , where U and V are orthogonal matrices, is a diagonal matrix, and the rotation matrix .

[0046] The rotation matrix is a 3x3 orthogonal matrix that describes the rotation relationship between two coordinate systems.

[0047] The translation vector is a three-dimensional vector that describes the translation relationship between two coordinate systems, and the calculation formula is .

[0048] The upper game model objective function is .

[0049] The lower game model objective function is .

[0050] The positioning accuracy parameter is derived from the comprehensive evaluation of the guide light source recognition accuracy, binocular vision measurement error and pose solution accuracy, and is used for upper game model objective function calculation and recovery success rate evaluation.

[0051] The recovery time parameter is derived from the accumulation of image processing time, pose calculation time, spacecraft adjustment time and communication delay time, and is used for upper game model objective function calculation and recovery efficiency optimization.

[0052] The energy consumption parameter is derived from the sum of LED guide light source power consumption, binocular camera power consumption, computing processing power consumption, and communication system power consumption, and is used for upper game model and lower game model objective function calculation.

[0053] The maximum allowed time parameter is derived from task time constraints and safety margin design, and is used for time efficiency evaluation of upper game model objective function.

[0054] The robustness parameter is derived from comprehensive evaluation of ocean current disturbance intensity, illumination variation degree, water transparency, and target occlusion situation, and is used for system stability evaluation of upper game model objective function.

[0055] The optical system power consumption parameter is derived from the sum of LED guide light source power consumption and binocular camera power consumption, and is used for optical equipment energy consumption optimization of lower game model objective function.

[0056] The processing time parameter is derived from the accumulation of image processing algorithm execution time and pose solution calculation time, and is used for calculation efficiency optimization of lower game model objective function.

[0057] The iteration number parameter is derived from the iteration calculation number of guide light source identification algorithm and pose optimization algorithm, and is used for algorithm complexity control of lower game model objective function.

[0058] The computing power consumption parameter is derived from microcomputer processor power consumption and software processing system running power consumption, and is used for computing device energy consumption optimization of lower game model objective function.

[0059] The image quality parameter is derived from comprehensive evaluation of image contrast, definition, and signal-to-noise ratio, and is used for imaging quality guarantee and algorithm robustness improvement of lower game model objective function.

[0060] The coupling parameter represents the mutual influence relationship between the upper game model and the lower game model, and is derived from the trade-off analysis between recovery accuracy requirement and energy consumption limitation, and is used for coordinating the optimization objectives of the two game models.

[0061] The specific implementation of the above steps is described in detail below.

[0062] The specific implementation of step S01 is to complete the system initialization and modeling process through multiple sub-steps. First, four blue-green LED guide light sources are installed on the front end of the recovery cage in a non-uniform ring-shaped distribution manner, with wavelengths in the range of 470-520 nm selected for the blue-green band to achieve optimal transmission characteristics in seawater. The installation positions of the four light sources adopt a scalene quadrilateral layout to ensure unique identification, and the distance between adjacent light sources is controlled within the range of 0.5-1.5 m. Next, an industrial photogrammetry system is used to measure the three-dimensional coordinates of the guide light sources. This system is based on the stereo vision principle and uses multiple calibrated cameras to capture the light source positions from different angles. The three-dimensional coordinates of the space points are calculated using the triangulation principle, with a measurement accuracy of sub-millimeter level. The measurement results are stored in the software processing system configuration file in the form of Cartesian coordinate system. Then, the binocular camera installed on the bow of the unmanned underwater vehicle is calibrated for internal and external parameters. The internal parameter calibration includes the determination of focal length, principal point coordinates, and distortion coefficients, while the external parameter calibration includes the determination of the relative pose relationship between the two cameras. The calibration process uses Zhang Zhengyou's calibration method to complete the image capture of the calibration board at multiple angles. Finally, an upper game model is established to maximize the recovery success rate. This model considers multiple factors such as positioning accuracy, recovery time, energy consumption control, system robustness, and coupling constraints. Mathematical optimization methods are used to find the optimal system parameter configuration. In the model, the positioning accuracy weight coefficient α is set to 0.4, the time efficiency weight coefficient β is set to 0.3, the energy consumption weight coefficient γ is set to 0.2, and the robustness weight coefficient δ is set to 0.1.

[0063] The specific implementation of step S02 is to achieve target capture through the cooperation of acoustic guidance and optical detection. First, the unmanned underwater vehicle is guided to the range of 15 m from the recovery cage by the acoustic guidance system. The acoustic guidance uses ultra-short baseline positioning technology combined with an inertial navigation system to achieve coarse positioning, with a positioning accuracy controlled within 1-3 m. During the guidance process, real-time position information and control instructions are transmitted through the acoustic communication link. When the vehicle enters the optical guidance working range, the binocular camera image acquisition system is started. The binocular camera working frame rate is set to 120 fps to meet the real-time requirement, the image resolution is 1280x1024 pixels, and the initial exposure time is set to 25 ms. During image acquisition, the guiding light source detection algorithm is started synchronously. This algorithm uses a target detection method based on color features and shape features to determine whether the guiding light source is detected by analyzing the distribution characteristics of the blue-green light spot in the image. If the guiding light source is successfully detected, the lower-level game model with the goal of minimizing system energy consumption is started. This model mainly optimizes parameters such as optical system power consumption, processing time, calculation power consumption, and image quality, and realizes energy consumption minimization by dynamically adjusting the working state of each subsystem. The optical system power consumption weight coefficient μ is set to 0.35, the processing time weight coefficient ν is set to 0.25, the calculation power consumption weight coefficient ω is set to 0.25, and the image quality weight coefficient η is set to 0.15.

[0064] The specific implementation of step S03 is to achieve the best imaging effect through adaptive exposure control. First, the binocular camera exposure value is initialized to 25 ms, which is based on the experience setting of underwater lighting conditions and LED light brightness. Then, the image under the current exposure parameter is acquired and the guiding light source target is segmented. The image segmentation uses a method based on color space transformation and threshold segmentation to convert the RGB image to the HSV color space and extract the target using the blue-green channel information. Next, the pixel size of the guiding light source target is calculated by counting the number of pixels in the segmented light source region, considering the integrity of the light source shape and the clarity of the boundary. The exposure value is adjusted according to the pixel size. If the pixel size is less than 25 pixels, it means that the light source is too dark and the exposure time needs to be increased. In this case, the exposure value is adjusted to the original value multiplied by 1.20 times. If the pixel size is greater than 80 pixels, it means that the light source is too bright and may produce saturation. In this case, the exposure value is adjusted to the original value multiplied by 0.75 times. Through multiple adjustment processes, an adaptive exposure control mathematical model is established, which describes the nonlinear relationship between the exposure value and the light source pixel size. The model parameters are obtained by least squares fitting, with a fitting accuracy requirement of more than 95%. Finally, the best exposure parameter is automatically set under different distances and lighting conditions.

[0065] The specific implementation of step S04 is to realize the accurate extraction of the light source target through image preprocessing and mask generation. First, the guide light source image is synchronously collected by the binocular camera, and the synchronization error is controlled within 10 ns to ensure the accuracy of the stereo matching. The hardware trigger method is used to ensure the timing consistency of the two cameras during image acquisition. Then, the Rolling Ball method is used to remove the background light interference. This method is based on the principle of morphological operation to separate the foreground target and the background area by simulating the rolling process of the ball on the image surface. The rolling ball radius parameter is set to 1.5-2.0 times the expected size of the light source, usually in the range of 30-50 pixels. Next, the contrast automatic enhancement processing is performed. This processing realizes the global contrast optimization by calculating the minimum and maximum gray values of the image. The enhanced image gray value is mapped to the range of 0-255 through linear transformation, and the relative gray relationship of the image is kept unchanged during the transformation process. Then, the maximum value mask image is generated. This mask is searched for local maximum points, and for each pixel point, it is checked whether it is the maximum value in the eight-connected region. If the condition is met, it is marked as 1 in the mask, otherwise it is marked as 0. Finally, the light spot contour mask is generated. This mask is based on the region growing method and takes the maximum value point as the seed point to start growing. If the gray difference between the neighborhood pixel and the seed point is less than the set threshold, the neighborhood pixel is added to the growing region. The threshold is set to 15% of the seed point gray value, and the growing process continues until there are no more pixels that meet the conditions.

[0066] The specific implementation of step S05 is to realize accurate recognition of the guide light source through contour analysis and optimization matching. First, statistical information of the recognized light source contour is calculated, including contour center coordinates, contour area, contour average gray, ellipse fitting long axis length and ellipse fitting short axis length, etc. The contour center coordinates are calculated by geometric moments, the contour area is obtained by pixel counting, and the average gray is calculated by the arithmetic mean of the pixel gray values in the contour. Then, the effective guide light source target is screened, and the screening conditions include the contour area being greater than 30 pixels, the contour average gray being greater than 200, and the ellipse fitting long-short axis ratio being less than 3. These conditions are determined based on the physical characteristics and imaging characteristics of the LED light source. Next, it is judged whether the four light sources constitute an effective spatial configuration through convex constraint configuration, which requires that the four light source points form a convex quadrilateral. The judgment is realized by calculating the convex hull of the four points and verifying its consistency with the original point set. Then, the effectiveness of the light source target is further verified through ellipse fitting long-short axis ratio screening, and the long-short axis ratio threshold is set to 3.0. The target is excluded if the long-short axis ratio exceeds the threshold. Finally, the four guide light source targets are constructed as nodes of the traveling salesman problem, and the nearest neighbor heuristic algorithm is used to solve the shortest loop. The algorithm starts from any node and selects the nearest unvisited node as the next visited node each time. The uniqueness of the guide light source number is realized by constructing the shortest loop, and the matching process considers the geometric relationship and spatial distribution characteristics of the light sources.

[0067] The specific implementation of step S06 is to realize three-dimensional coordinate measurement through accurate positioning and stereo matching. First, the center of the guide light source is accurately positioned using the barycenter method, which determines the center position by calculating the weighted average of all pixel coordinates in the light source area. The weight uses the pixel gray value, and the calculation process considers the brightness distribution characteristics of the light source to improve the positioning accuracy, which can reach the sub-pixel level. Then, the center coordinates of each guide light source are recorded and a coordinate index table is established, with the coordinate recording precision maintained to three decimal places to meet the subsequent calculation requirements. Next, the epipolar constraint principle is used to complete the homonymic matching of the guide light source center points in the binocular images. The epipolar constraint is based on the epipolar geometry theory, which constrains the points in the left image to the corresponding epipolar line in the right image, greatly reducing the search range and improving the matching accuracy. The matching process uses a method based on feature correlation, with a correlation coefficient threshold set to 0.8 or above. Then, the three-dimensional coordinates of the guide light source center are calculated based on the binocular vision measurement principle. This principle uses two cameras to observe the same target from different angles, and the three-dimensional position of the space point is solved by the triangulation method. The calculation process needs to use the pre-calibrated internal and external parameters of the binocular camera, including focal length, principal point coordinates, distortion coefficients, and the relative pose relationship between the cameras. The three-dimensional coordinate calculation accuracy is influenced by the comprehensive effects of camera calibration accuracy and image measurement accuracy, and the measurement accuracy can usually reach the millimeter level.

[0068] The specific implementation of step S07 is to realize the accurate calculation of the relative pose relationship through coordinate system transformation and pose optimization. First, according to the pre-measured three-dimensional coordinate information of the guide light source and the calculated three-dimensional coordinates of the guide light source in the binocular coordinate system, two corresponding three-dimensional point sets are constructed, the number of point sets is 4 corresponding light source positions, and each point contains X, Y and Z three coordinate components. Then the centroids of the two three-dimensional point sets are defined and the centroid coordinates are calculated, the centroid is calculated by the arithmetic average of all point coordinates, and the centroid coordinates are obtained by subtracting the corresponding centroid coordinates from the original coordinates. The purpose of this step is to eliminate the influence of the translation component and facilitate the solution of the rotation matrix. Next, a covariance matrix for pose solution is constructed, which is formed by the outer product accumulation of the two sets of centroid coordinates, the matrix dimension is 3x3, and it contains the rotation information between the two coordinate systems. Then the covariance matrix is decomposed by singular value decomposition method, the singular value decomposition represents the matrix as the product of three matrices, which contains two orthogonal matrices and a diagonal matrix, the decomposition process is realized by numerical calculation method, and the calculation accuracy is required to reach the level of double-precision floating-point number. Finally, the rotation matrix and the translation vector are solved according to the singular value decomposition result, the rotation matrix is obtained by the product of the two orthogonal matrices, and the translation vector is calculated by the linear relationship between the centroid coordinates and the rotation matrix. The calculation result represents the relative pose relationship between the unmanned underwater vehicle and the recovery cage, including three rotation angles and three translation components, the pose calculation accuracy is affected by the point coordinate measurement accuracy and the numerical stability of the algorithm, and the accuracy level of angle error less than 0.1 degree and position error less than 5mm can be usually achieved.

[0069] The key technical ideas of the present application mainly reflect in the following aspects. First, the optimization strategy of the double-layer game model, by constructing the upper game model to maximize the recovery success rate as the target, the lower game model to minimize the system energy consumption as the target, the coordination optimization of recovery performance and energy consumption control is realized, compared with the traditional single target optimization method, the double-layer game model can significantly reduce the system energy consumption under the premise of ensuring the recovery success rate, and improves the adaptability and robustness of the system. Secondly, the fusion technology of adaptive exposure control and traveling salesman problem solving, by establishing the mathematical relationship model of exposure value and light source pixel size to realize adaptive exposure control, combined with the shortest loop solving of traveling salesman problem to realize the uniqueness of the number matching of the guide light source, compared with the traditional fixed exposure and simple matching method, this technology can maintain stable recognition performance under different distances and lighting conditions, greatly improve the accuracy and reliability of light source recognition. Third is the cooperative processing technology of Rolling Ball background removal and region growing method, by using the Rolling Ball method to effectively separate the foreground target and the background uneven illumination area, combined with the region growing method to realize the accurate extraction of the light spot contour, compared with the traditional threshold segmentation method, this cooperative processing technology can better adapt to the complex underwater lighting environment, significantly improve the robustness and precision of light source target detection. The synergistic effect of these key technical ideas forms a complete optical guidance system solution, through the organic combination of multi-level optimization, adaptive control and intelligent processing, the high precision, high reliability and low energy consumption of deep sea unmanned underwater vehicle recovery are realized, compared with the existing technology, it has significant comprehensive performance advantage and practical value.

[0070] It should be noted that the present application also solves the technical problems of unstable identification accuracy of the guide light source in the deep sea complex environment. The traditional optical guidance method often has large fluctuations in the identification accuracy of the guide target when facing the complex conditions such as turbid water, uneven light, and sea current disturbance in the deep sea environment, which affects the reliability of the whole recovery system. The present application effectively separates the foreground target and the uneven background light area by using the Rolling Ball background removal technology, improves the visual quality of the image by combining the contrast automatic enhancement processing technology, at the same time, establishes the maximum mask image and the light spot contour mask to realize the accurate target segmentation, ensures the geometric effectiveness of the identified target through the convex constraint configuration judgment and the long-short axis ratio screening mechanism of ellipse fitting, uses the traveling salesman problem modeling and the nearest neighbor heuristic algorithm to solve the combination optimization problem of guide light source number matching, thereby significantly improves the accuracy and stability of the identification of guide light source in the complex marine environment. In addition, the present application also solves the technical problem that the energy consumption control and the positioning accuracy of the unmanned underwater vehicle recovery system are difficult to balance. In actual deep sea recovery operation, improving the positioning accuracy usually needs to increase the light source power, prolong the exposure time, and improve the calculation complexity, which will significantly increase the system energy consumption, and under the constraint of limited battery capacity, too high energy consumption will affect the task execution time and system reliability. The present application builds a double-layer game model architecture, the upper model optimizes the macro indicators such as positioning accuracy, recovery time, system robustness, and the lower model focuses on the micro energy consumption control of the optical system power consumption, processing time, and calculation power consumption, the two models realize dynamic coordination through coupling parameters, maximize the system energy efficiency on the premise of ensuring the recovery accuracy, at the same time, the adaptive exposure control technology dynamically adjusts the camera parameters according to the pixel size of the guide light source, avoids the energy waste caused by fixed parameter setting, and realizes the optimal balance between positioning accuracy and energy consumption control.

[0071] Specifically, the principle of the present application is that the present application can solve the problem of insufficient positioning accuracy and poor real-time performance of optical guidance in deep sea environment. The root cause of the problem lies in the adoption of a multi-level collaborative optimization technical architecture and an accurate visual measurement method. First, the selection of the blue-green LED guidance light source is based on the optical characteristic that the blue-green light wavelength has good water penetration capability in deep sea environment. The non-uniform ring distribution design of the four guidance light sources ensures that sufficient spatial constraint information can be provided under different observation angles, avoiding the ambiguous solution problem caused by the traditional single-point or symmetrically distributed guidance method in certain attitudes. The design logic of the double-layer game model is to decompose the recovery problem into two levels of macro-strategy optimization and micro-execution optimization. The upper game model aims to maximize the recovery success rate, considering multiple constraint conditions such as positioning accuracy, recovery time, energy consumption control, and system robustness. The lower game model aims to minimize the system energy consumption, focusing on the optimization of the execution level such as optical system power consumption, processing time, and computational complexity. The two models are coupled through parameters to achieve coordination and ensure the optimal balance between accuracy and efficiency of the overall system. The adaptive exposure control technology solves the influence of light condition changes on image quality in deep sea environment by monitoring the pixel size of the guidance light source in real time and dynamically adjusting the camera exposure parameters. The RollingBall background removal and contrast automatic enhancement processing technology effectively separates the guidance light source from the complex seabed background, improving the accuracy of target recognition. The barycentric method positioning combined with the epipolar constraint binocular stereo matching technology realizes sub-pixel level positioning accuracy of the light source center. Based on the pre-calibrated three-dimensional coordinate information of the guidance light source and the real-time calculated binocular coordinate system coordinates, the rigid transformation matrix solved by the singular value decomposition method can accurately describe the relative pose relationship between the vehicle and the recovery cage, providing high-precision spatial positioning information for the real-time control system, thereby realizing high-precision and real-time optical guidance recovery.

[0072] A specific embodiment 1 of the present application is provided below, and the specific implementation of each step in embodiment 1 is described in detail as follows.

[0073] In this embodiment, the specific implementation of steps S01-S02 is the same as described above, and will not be described in detail here.

[0074] The specific implementation of step S03 is to achieve the best imaging effect through the adaptive exposure control algorithm, and the exposure value adjustment rule is represented as follows:

[0075] When the target pixel size of the light source is , the exposure value adjustment formula is:

[0076] ;

[0077] When the target pixel size of the light source is , the exposure value adjustment formula is:

[0078] ;

[0079] wherein, is the adjusted exposure value, unit: ms; is the exposure value before adjustment, unit: ms; is the pixel size of the light source target, unit: pixel number.

[0080] The parameter acquisition method is as follows: The initial value is set to 25 ms, which is obtained by camera exposure time control; The pixel number of the light source region is obtained by image segmentation algorithm, including the following steps: 1. Color space transformation is performed on the collected image, and the RGB image is converted to HSV color space; 2. Threshold segmentation is performed using blue-green channel information to extract the light source target region; 3. The pixel number of the light source region after segmentation is counted to obtain the value of .

[0081] The specific implementation of step S04 is to realize accurate extraction of the light source target through image enhancement algorithm and background removal algorithm. The mathematical expression of the contrast automatic enhancement processing is as follows:

[0082] ;

[0083] wherein, is the image gray value after contrast automatic enhancement; is the image gray value after Rolling Ball method processing; is the minimum gray value of the image; is the maximum gray value of the image; is the pixel point coordinate, represents the horizontal coordinate, represents the vertical coordinate.

[0084] The parameter acquisition method is as follows: It is obtained by Rolling Ball background removal algorithm. The algorithm adopts the principle of morphological operation, and the rolling ball radius is set to 1.5 times the expected size of the light source; and is obtained by traversing all pixel points of the image, , . .

[0085] The specific implementation of step S05 is to realize accurate recognition and matching of the guide light source through contour analysis and combination optimization. The mathematical expression of the candidate light spot combination generation is as follows:

[0086] ;

[0087] wherein, is the number of generated candidate spot combinations; is the number of candidate spots; is the number of combinations of selecting 4 from .

[0088] The contour center coordinate calculation formula is:

[0089] , ;

[0090] The pixel gray scale and calculation formula of the final spot combination selection are:

[0091] ;

[0092] The guide light source distance calculation formula is:

[0093] ;

[0094] wherein, and are the horizontal and vertical coordinates of the contour center, respectively, in units of pixels; is the contour area, i.e., the number of pixel points within the contour; and are the coordinates of the th pixel point within the contour; is the total pixel gray scale sum of the spot combination; is the contour area of the th spot; is the contour average gray scale of the th spot; is the Euclidean distance between the center of the th light source and the center of the th light source, in units of pixels; , , , are the coordinates of the center of the th and th light source, respectively.

[0095] wherein, the parameter acquisition method is: obtained through the aforementioned screening step, and the screening conditions include a contour area greater than 30 pixels, a contour average gray scale greater than 200, and an ellipse fitting major-to-minor axis ratio less than 3; obtained by counting the number of pixel points marked as 1 in the contour mask; obtained by traversing the contour mask to obtain all pixel point coordinates within the contour; and It is obtained by calculating through contour statistics, where .

[0096] The specific implementation of step S06 is to achieve three-dimensional coordinate measurement of the guiding light source through the principle of binocular stereo vision. The mathematical expression of the center positioning using the centroid method is as follows:

[0097] , ;

[0098] In the formula, and These are the precise coordinates of the center of the light source, in pixels. This represents the total number of pixels within the light source area; and The first in the light source area The coordinates of each pixel; For pixels The grayscale value at that location is used as a weighting coefficient.

[0099] The parameter acquisition method is as follows: This is obtained by counting the number of pixels in the mask of the light source area; Read the grayscale value of the corresponding pixel directly from the enhanced image; The coordinates of all pixels within the light source region are obtained by traversing the mask of the light source region.

[0100] The specific implementation of step S07 is to calculate the relative pose relationship between the unmanned underwater vehicle and the recovery cage using a 3D-3D pose estimation algorithm. The mathematical model is established as follows:

[0101] Let the matched 3D point set in the two coordinate systems be:

[0102] , ;

[0103] The pose transformation relationship is as follows:

[0104] ;

[0105] In the formula, The set of three-dimensional coordinate points of the guiding light source in the coordinate system of the recovery cage; This is the set of three-dimensional coordinate points of the guiding light source in the camera coordinate system; For the first time in the coordinate system of the recovery cage The three-dimensional coordinate vector of a guiding light source; In the camera coordinate system, the first The three-dimensional coordinate vector of a guiding light source; It is a 3×3 rotation matrix; It is a three-dimensional translation vector; To guide the number of light sources, in the present application .

[0106] Define the first The error term for the points is:

[0107] ;

[0108] In the formula, is the first The pose transformation error vector of the matching points.

[0109] The least squares optimization problem is expressed as:

[0110] ;

[0111] In the formula, is the objective function; The 2-norm of the vector is denoted.

[0112] Define the centroid of the two sets of points as:

[0113] , ;

[0114] In the formula, is the centroid of the point set in the recovery cage coordinate system; is the centroid of the point set in the camera coordinate system.

[0115] Calculate the de-centroid coordinates:

[0116] , ;

[0117] In the formula, is the de-centroid coordinate of the first point in the recovery cage coordinate system; is the de-centroid coordinate of the first point in the camera coordinate system.

[0118] Construct the covariance matrix:

[0119] ;

[0120] In the formula, is a 3x3 covariance matrix containing rotation information between the two coordinate systems.

[0121] Perform singular value decomposition on the matrix :

[0122] ;

[0123] In the formula, and is a 3x3 orthogonal matrix; is a 3x3 diagonal matrix containing singular values.

[0124] When is full rank, the rotation matrix is:

[0125] ;

[0126] The translation vector is:

[0127] .

[0128] where the parameter acquisition method is: The three-dimensional spatial coordinates of the guide light source in the recycling cage body are obtained by measuring the industrial photogrammetry system; It is calculated by the binocular stereo vision measurement principle, which requires the use of pre-calibrated binocular camera internal and external parameters; singular value decomposition is realized by numerical calculation method, which requires the calculation accuracy to reach the double-precision floating point level.

[0129] The principles and effects of the above formulas are as follows: Exposure value adaptive adjustment formula and By establishing the mathematical relationship between the exposure parameter and the pixel size of the light source, the best imaging effect under different distances and lighting conditions is realized, which significantly improves the image quality and the stability of light source recognition compared with the traditional fixed exposure method. This formula uses a piecewise linear adjustment strategy to effectively avoid overexposure and underexposure, improving the adaptability of the system in complex marine environments. Contrast automatic enhancement formula Through linear transformation, the image gray value is mapped to the full dynamic range, maximizing the contrast and detail information of the image. Compared with the traditional global threshold method, it can better handle the problem of uneven lighting. This formula keeps the relative gray relationship of the image unchanged, ensuring the stability and accuracy of subsequent image processing algorithms. Combination number calculation formula It provides a mathematical basis for the unique identification of guide light sources. By enumerating all possible four-light source combinations and combining geometric constraints, accurate light source matching in complex environments is achieved. Compared with traditional heuristic matching methods, it has higher reliability and robustness. Pixel gray and formula By considering the product of the spot area and the average gray value, the quality of the spot combination is evaluated. Compared with single feature screening methods, it can more accurately identify the real guide light source combination. The area term in this formula reflects the geometric characteristics of the light source, and the average gray term reflects the brightness characteristics of the light source. The combination of the two improves the accuracy and anti-interference ability of light source recognition. Distance calculation formula The Euclidean distance is used to measure the spatial relationship between the light sources, which provides a geometric constraint condition for the light source number matching, and the formula realizes the unique number by finding the shortest distance between the adjacent light sources, which has better numerical stability and calculation efficiency compared with the traditional angle or topological matching method. and The centroid calculation method with gray weight is used, which fully utilizes the brightness distribution information of the light source, realizes the positioning accuracy of sub-pixel level, and significantly improves the positioning accuracy compared with the traditional geometric center calculation method, and the formula effectively suppresses the influence of noise and background interference through the weight mechanism. The 3D-3D pose estimation formula group is optimized by least square and singular value decomposition Method, which converts the complex pose estimation problem into a linear algebra problem for solving, realizes high-precision relative pose calculation, and has better numerical stability and calculation efficiency compared with the traditional iterative optimization method, and the mathematical model through the center of mass operation and and covariance matrix construction Effectively separate the rotation and translation components, simplify the solving process and improve the robustness of the algorithm, and the singular value decomposition method ensures the orthogonality constraint of the rotation matrix , ensures the physical meaning and mathematical correctness of the pose estimation result, and the calculation of the translation vector uses the linear property of the center of mass transformation to avoid complex nonlinear optimization process.

[0130] In order to better understand and implement the present application, the following provides an embodiment 2 of a specific application scenario of the present application: a certain ocean technology team needs to accurately recover an autonomous underwater vehicle with a working depth of 1800m in the process of deep-sea resource exploration. The traditional acoustic guidance recovery method has the problems of insufficient positioning accuracy and signal interference in complex seabed topography environment, and the recovery success rate is only 67%, and the recovery time is relatively long, which takes an average of 42 minutes. In order to solve this technical problem, the team decided to use the optical guidance method of the present application to improve the recovery system. As Figure 2 shown, the overall scheme of the optical guidance system includes the overall layout and connection relationship of the core components such as the guidance light source, binocular camera, electric control system and software processing system.

[0131] Specifically, the guide light source is installed on the recovery cage, and the binocular camera, the electric control system and the software processing system are located in the sealed control cabin of the AUV. During operation, the guide light source should be in the common field of view of the binocular camera. The electric control system performs hardware triggering, so that the binocular camera collects image data of the guide light source target, and transmits the image data to the microcomputer through the data transmission line. Then, the three-dimensional measurement result and the relative positioning information of the light source target are calculated through the computer processing software, and finally uploaded to the control system of the AUV.

[0132] Four circular LED guide light sources 1 are installed on the AUV recovery cage, and a non-uniform ring-shaped distribution is arranged at the front end of the recovery cage 2. The guide light source is composed of 4 light sources 1 (including No. 1 light source, No. 2 light source, No. 3 light source and No. 4 light source) and light source control and driving circuit 3. Through pressure-resistant packaging, the material of the body is TC4, the material of the window glass is sapphire, the structure sealing form adopts O-ring sealing, and the pressure depth design is 2000m.

[0133] The binocular camera 4 is installed in the bow dry cabin of the AUV. The camera bottom installation interface is provided with a positioning groove and a positioning. The positioning connection hole is realized through the adapter plate and the AUV. Figure 2 A 520nm single bandpass filter is installed in front of the lens to reduce the influence of stray light on the imaging quality of the light source. The frame rate of the camera is 1280x1024@120fps, which can meet the requirement of attitude data≥10Hz. The binocular camera has a gigabit network interface and a hard synchronization trigger mode. The synchronization trigger pulse error of the two cameras is extremely small, far less than the exposure time (10ns).

[0134] The function of the electric control system is to realize the driving and control of the cooperative light source, the synchronization trigger control of the camera, the collection and processing of image data. The microcomputer 5, the power system 6 and the trigger system 7 are integrally sealed in the metal box, which is connected with the outside through a 5-core power cable 8. Among them, 3 cores are used for RS485 serial communication to ensure that data can be transmitted to the AUV in time to realize corresponding attitude adjustment, and 2 cores are used for power supply.

[0135] The software processing system includes cooperative light source target detection, positioning and matching, three-dimensional reconstruction of target light source, relative pose calculation and other functions of the two images obtained by the binocular camera. Data output is mainly realized through serial port, including opening serial port, serial port configuration, serial port reading and writing, and closing serial port.

[0136] The implementation process first carries out system deployment and initialization configuration. The technical team installs four blue-green LED guide light sources in a non-uniform annular distribution manner in front of the recovery cage, with a wavelength of 485 nm selected to obtain the best seawater transmission characteristics. The four light sources are installed according to an isosceles quadrilateral layout, with No. 1 light source located at the upper left of the center of the front end of the recovery cage, with coordinates (-0.6, 0.8, 0), No. 2 light source located at the upper right, with coordinates (0.7, 0.9, 0), No. 3 light source located at the lower right, with coordinates (0.8, -0.7, 0), and No. 4 light source located at the lower left, with coordinates (-0.5, -0.8, 0), with units in meters. After the light sources are installed, an industrial photogrammetry system is used to accurately measure the three-dimensional coordinates of the guide light sources, with a measurement accuracy of 0.3 mm, and the measurement results are stored in the software processing system configuration file. Subsequently, the binocular camera installed on the bow of the unmanned underwater vehicle is calibrated, as shown in the camera installation structure Figure 3 The adapter plate is connected to the vehicle for precise positioning, the baseline distance of the binocular camera is 120 mm, the focal length is 8 mm, the image resolution is 1280x1024 pixels, and the working frame rate is 120 fps. The Zhang Zhengyou calibration method is used in the calibration process, and the parameters are determined through 15 different angle calibration board images. The calibration accuracy verification result shows that the re-projection error is less than 0.4 pixels. When establishing the upper game model, the positioning accuracy weight coefficient a is set to 0.4, the time efficiency weight coefficient β is set to 0.3, the energy consumption weight coefficient γ is set to 0.2, and the robustness weight coefficient δ is set to 0.1, and the model convergence accuracy requirement is .

[0137] The acoustic guidance stage uses ultra-short baseline positioning technology to guide the unmanned underwater vehicle to the recovery cage within a range of 15 m. When the vehicle is about 18 m away from the recovery cage, the acoustic guidance system has a positioning accuracy of 2.3 m, meeting the starting conditions of the optical guidance. When the vehicle enters the optical guidance working range, the binocular camera image acquisition system is started immediately, and the guide light source detection algorithm is activated. The initial image acquisition detects 3 suspected light source targets, triggering the start of the lower game model. The optical system power consumption weight coefficient μ is set to 0.35, the processing time weight coefficient v is set to 0.25, the calculation power consumption weight coefficient ω is set to 0.25, and the image quality weight coefficient η is set to 0.15. The detection algorithm analysis shows that 2 of the targets meet the geometric and brightness feature requirements, and 1 target is excluded because the ellipse fitting major and minor axis ratio exceeds 3.2.

[0138] The adaptive exposure control stage first initializes the binocular camera exposure value ms. The image under the current parameters is collected and guided light source target segmentation is performed, the RGB image is converted to HSV color space by color space transformation, the light source target area is extracted by threshold segmentation using blue-green channel information, and the pixel size is obtained by counting the pixel number of the segmented light source area 18, 22, 156 and 178 pixels, respectively. According to the pixel size judgment criterion, when , the first two light source targets need to increase the exposure time, and the exposure value is adjusted to ms according to the formula When , the last two light source targets need to reduce the exposure time to avoid oversaturation, and the exposure value is adjusted to ms according to the formula After 3 iterations of adjustment, the pixel size of the four light source targets is stabilized in the range of 32, 38, 67 and 71 pixels, which meets the best imaging requirements. The fitting accuracy of the established adaptive exposure control mathematical model reaches 97.3%, and the best exposure parameter automatic setting under different distance conditions is realized.

[0139] In the image preprocessing and target extraction stage, the binocular camera synchronously collects the guided light source image, and the synchronization error is controlled within 8 ns. The Rolling Ball method is used to remove the background light interference, and the rolling ball radius parameter is set to 35 pixels, effectively separating the foreground light source target and the sea water background scattered light. The contrast automatic enhancement processing calculates the minimum gray value , the maximum gray value , and the global contrast is optimized by the linear transformation formula , wherein is the pixel point coordinate, is the image gray value after the Rolling Ball method processing, is the image gray value after the contrast automatic enhancement. In the process of generating the maximum value mask image, 67 local maximum points are identified by eight-connected region search. Based on the region growing method, the light spot contour mask is generated, the maximum value point is used as the seed point, and the gray threshold is set to 15% of the seed point gray value. Finally, 8 candidate light source contours are extracted.

[0140] In the guided light source recognition and matching stage, the statistical information of the extracted 8 candidate light source contours is calculated, and the whole light source recognition process is strictly performed according to the light source recognition flow chart shown in Figure 5 . The contour statistical results shown in Table 1 show the key parameters of each candidate light source.

[0141] Table 1. Statistical information of candidate light source contours

[0142]

[0143] Through screening condition verification, the candidate light sources with a profile area greater than 30 pixels, an average gray level greater than 200, and an aspect ratio less than 3 are numbered 1, 2, 3, 4, and 7, a total of 5 targets. Number 5 is excluded because the area is insufficient, number 6 is excluded because the aspect ratio is excessive, and number 8 is excluded because both the area and the gray level do not meet the requirements. Four of the five effective candidate light sources are selected to form a combination, and the number of generated candidate combinations is . The profile center coordinates are calculated by the formula, taking light source No. 1 as an example: , , wherein is the profile area. Through convex constraint configuration judgment, all 5 combinations form effective convex quadrilaterals. The pixel gray levels of each combination are calculated, and the gray level sum of the combination {1, 2, 3, 4} is calculated according to the formula , , which is the maximum value, and is selected as the final light spot combination. The nearest neighbor heuristic algorithm is used to guide the light source number matching, and the distance between each light source is calculated, wherein the distance between light source 1 and light source 4 is calculated according to the formula , pixels, which is the minimum value, and is determined as the adjacent light source pair, and the unique number matching is completed by anticlockwise sorting.

[0144] The barycentric method is used to realize the accurate positioning of the guide light source center in the precise positioning and three-dimensional coordinate measurement stage. Taking light source No. 1 as an example, the total number of pixel points in the light source area is , and the accurate center coordinates are obtained by the gray level weighted calculation formula and , , , wherein and are the coordinates of the th pixel point in the light source area, is the gray level value at the pixel point as a weight coefficient, realizing sub-pixel level positioning accuracy. The homonymous matching of the guide light source center points in the binocular image is completed through the principle of epipolar constraint, and the matching correlation coefficients are all greater than 0.85, meeting the matching reliability requirements. The three-dimensional coordinates of each guide light source center are calculated based on the binocular vision measurement principle, and the measurement results shown in Table 2 show the light source position information in the camera coordinate system.

[0145] Table 2 Measurement results of three-dimensional coordinates of guide light sources

[0146]

[0147] In the relative pose relationship calculation stage, two sets of corresponding point sets are constructed according to the pre-measured three-dimensional coordinates of the guide light sources and the calculated light source coordinates in the camera coordinate system. Let the three-dimensional coordinate point set of the guide light sources in the recovery cage coordinate system be , the three-dimensional coordinate point set of the guide light source in the camera coordinate system is , and the pose conversion relationship is , wherein is a 3x3 rotation matrix, is a three-dimensional translation vector. Define the first The error term of the point is , and the least squares optimization problem is established . The centroids of the two groups of point sets are calculated as and , wherein is the number of guide light sources. The translation component is eliminated by the centroid operation and . After expanding and simplifying the error function, the least squares problem becomes . The pose calculation is divided into three steps: first, solve the optimization problem , expand the error term about to get . Since , the objective function is simplified to . The covariance matrix is constructed and singular value decomposition is performed . The singular value decomposition result shows that the singular values are 2847.3, 2651.8 and 2398.7, all positive and the matrix is full rank, meeting the solution conditions. When is full rank, the rotation matrix , and the translation vector . The calculated Euler angle representation of the rotation matrix is pitch angle -2.3°, yaw angle 1.7°, and roll angle -0.8°, and the translation vector is mm. The pose calculation accuracy verification result shows that the angle error is less than 0.08°, and the position error is less than 3.2 mm, meeting the high-precision recovery guidance requirements.

[0148] The performance parameters of the entire optical guidance process are shown in Table 3, which shows the technical indicators of the system in actual application.

[0149] Table 3 Performance parameters of the optical guidance system

[0150]

[0151] After 15 consecutive recovery tests, the entire optical guidance implementation process strictly follows Figure 4The optical guiding method flowchart is shown, and the recovery system adopting the optical guiding method of the application has significant technical advantages in complex marine environments. Compared with the traditional acoustic guiding recovery method, the light source recognition accuracy is improved from 82.1% to 96.3%, mainly due to the synergistic effect of adaptive exposure control and Rolling Ball background removal technology, effectively adapting to the imaging requirements under different distances and lighting conditions. The pose calculation accuracy is improved from 15.6mm to 3.2mm, with an accuracy improvement of 79.5%, which is due to the application of the gravity method sub-pixel positioning technology and the singular value decomposition pose estimation algorithm, realizing the relative pose measurement accuracy of millimeter level. The total recovery time is shortened from 42.0 minutes to 28.5 minutes, with an efficiency improvement of 32.1%, mainly because the double-layer game model optimizes the system parameter configuration, reducing the search and adjustment time. The system energy consumption is reduced from 3.4W·h to 2.8W·h, with an energy saving effect of 17.6%, through the energy consumption optimization strategy of the lower game model, the power consumption of the optical system and the computing device is significantly reduced under the premise of ensuring performance. The most critical recovery success rate index is improved from 67.0% to 89.2%, with a success rate improvement of 33.1%, which is a significant improvement in the core index, reflecting the practical value and reliability of the technical solution of the application. The traditional method is prone to positioning failure and signal interruption in complex seabed topography and strong sea current interference environment, while the optical guiding method of the application overcomes the environmental interference factors through multi-level optimization control and intelligent image processing technology, realizes stable and reliable high-precision recovery guidance, and provides important technical support for the safe recovery of deep-sea unmanned underwater vehicles.

[0152] It should be noted that the variables involved in the application are explained in detail as shown in Table 4.

[0153] Table 4 Variable explanation table

[0154]

[0155] The above is only a specific embodiment of the application, but the protection scope of the application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the application, which should be covered within the protection scope of the application.

Claims

1. An optical guidance method for deep-sea unmanned underwater vehicle recovery, characterized by, Four blue-green LED guiding light sources are installed on the front end of the recovery cage, three-dimensional spatial coordinate information of the guiding light sources is measured and stored through an industrial photogrammetry system, binocular camera of the unmanned underwater vehicle is calibrated, upper game model and lower game model are established; the unmanned underwater vehicle is guided to the range of 15m from the recovery cage through acoustic guidance, the binocular camera image acquisition system is started and the guiding light source target is detected; an adaptive exposure control mathematical model is established, images are acquired and the guiding light source target is segmented to calculate pixel size; The binocular camera synchronously acquires guiding light source images, the Rolling Ball method is used to remove background light interference, contrast automatic enhancement processing is used to generate maximum mask image and light spot profile mask; statistical information of the recognized light source profile is calculated, four effective guiding light source targets are determined through convex constraint configuration judgment and long-short axis ratio screening of ellipse fitting, the four guiding light source targets are constructed as traveling salesman problem nodes and nearest neighbor heuristic algorithm is used to complete guiding light source number matching; the barycentric method is used to realize accurate positioning of the center of the guiding light source, the same name matching of the center of the guiding light source in the binocular image is completed through the principle of epipolar constraint, three-dimensional coordinates of the center of the guiding light source are calculated based on binocular vision measurement principle; according to the pre-measured three-dimensional coordinate information of the guiding light source and the calculated three-dimensional coordinates of the guiding light source in the binocular coordinate system, the rotation matrix and the translation vector are solved through singular value decomposition method, the relative pose relationship between the unmanned underwater vehicle and the recovery cage is calculated.

2. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 1, characterized in that, The four blue-green LED guiding light sources are installed in non-uniform ring distribution, the upper game model takes maximizing recovery success rate as the target, the lower game model takes minimizing system energy consumption as the target, and the lower game model is started when the guiding light source is detected.

3. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 2, characterized in that, The adaptive exposure control mathematical model is initialized with the exposure value of the binocular camera being 25ms, if the pixel size is less than 25 pixels, the exposure value is adjusted to the original value multiplied by 1.20, and if the pixel size is greater than 80 pixels, the exposure value is adjusted to the original value multiplied by 0.

75.

4. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 3, characterized in that, The industrial photogrammetry system is a high-precision three-dimensional coordinate measurement device based on stereo vision principle, a plurality of calibration cameras are used to shoot the target, and the three-dimensional coordinates of the space points are calculated by using the principle of triangulation, and the measurement accuracy reaches sub-millimeter level.

5. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 4, characterized in that, The Rolling Ball method is an image background removal technology based on morphological operation, which effectively separates the foreground target and the background light uneven area by simulating the rolling process of the ball on the image surface.

6. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 5, characterized in that, The contrast automatic enhancement processing is calculated by calculating the minimum gray value and the maximum gray value of the image, and the gray value of the enhanced image is calculated pixel by pixel.

7. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 6, characterized in that, The statistical information calculation includes obtaining the center coordinates, the profile area, the profile average gray value, the length of the long axis of the ellipse fitting and the length of the short axis of the ellipse fitting.

8. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 7, characterized in that, In the step of calculating the three-dimensional coordinates of the guiding light source in the binocular coordinate system, the centroids of the two groups of three-dimensional point sets are defined and the de-centroid coordinates are calculated; the de-centroid coordinates are the relative coordinates obtained by subtracting the centroid coordinates of the corresponding point set from the three-dimensional point coordinates.

9. The method for optical guidance of deep-sea unmanned underwater vehicle recovery according to claim 8, characterized in that, The extreme value mask image is initialized as a full zero matrix, and the image size is consistent with the original image; for each pixel point, if the gray value is not less than the gray value of the pixel in the eight-connected region, the pixel point is considered as an extreme value point and marked as 1 at the corresponding position.

10. The method for optical guidance of deep-sea unattended underwater vehicle recovery according to claim 9, characterized in that, The translation vector is a three-dimensional vector describing the translation relationship between the two coordinate systems.

Images