Nuclear fuel assembly grabbing positioning and deviation correction method and system based on industrial vision

Abstract

The invention provides a nuclear fuel assembly grabbing positioning and deviation correction method and system based on industrial vision. The method comprises the steps of obtaining multi-view image data of a target nuclear fuel assembly and structured light deformation image data under a corresponding view; extracting an outer contour boundary of the target nuclear fuel assembly in a three-dimensional space based on the multi-view image data and the structured light deformation image data to obtain initial outer contour boundary data; performing pose positioning based on the initial outer contour boundary data in combination with original design data of the target nuclear fuel assembly to obtain a current pose parameter of the target nuclear fuel assembly in a three-dimensional space; performing deviation analysis based on the current pose parameter in combination with a preset standard pose parameter to obtain current deviation data and a corresponding deviation analysis result; and if the deviation analysis result is correctable deviation, performing instruction generation based on the current deviation data to obtain a pose correction instruction. The problem that the correction effect is poor in an existing method is solved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and system for grasping, locating and correcting deviations of nuclear fuel assemblies based on industrial vision. Background Technology

[0002] During the nuclear fuel cycle, nuclear fuel assemblies, as the core components of the reactor, are subject to strict requirements for high precision and low risk in terms of the quality of their handling and positioning operations during production, transfer, and refueling.

[0003] Currently, the grasping and positioning of nuclear fuel assemblies mainly relies on manual assistance, simple mechanical limiting devices, or laser and infrared positioning technologies. Due to the high radiation and cleanliness requirements of nuclear fuel assembly handling environments, long-term human exposure poses a serious health threat. Furthermore, manual operation is affected by factors such as experience and visual fatigue, making it difficult to control positioning accuracy and achieve precise deviation correction. Simultaneously, the high radiation and dust environments of the nuclear industry are susceptible to electromagnetic interference and medium shielding, causing positioning signal distortion in laser and infrared technologies, thus compromising stability and repeatability. In addition, laser sensors are easily affected by nuclear radiation and on-site dust interference, further amplifying positioning deviations. Moreover, the fixed compensation rules used in existing methods cannot dynamically adapt to deviation changes under different scenarios, resulting in poor correction effects and increasing the risk of safety issues such as assembly collision damage. Summary of the Invention

[0004] This invention provides a method and system for grasping, positioning and deviation correction of nuclear fuel assemblies based on industrial vision, which improves the accuracy, stability and repeatability of grasping and positioning, and solves the safety problems of poor correction effect and easy collision damage to assemblies in existing methods.

[0005] In a first aspect, the present invention provides a method for grasping, locating, and correcting deviations of nuclear fuel assemblies based on industrial vision, comprising: Acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints; Based on the multi-view image data and the structured light deformation image data, the outer contour boundary of the target nuclear fuel assembly in three-dimensional space is extracted to obtain the initial outer contour boundary data; Based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, the pose is located to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space. Based on the current pose parameters and the preset standard pose parameters, a deviation analysis is performed to obtain the current deviation data and the corresponding deviation analysis results. If the deviation analysis result is a correctable deviation, then an instruction is generated based on the current deviation data to obtain a pose correction instruction, and the pose correction instruction is transmitted to the robot grasping system.

[0006] Secondly, the present invention also provides a nuclear fuel assembly grasping, positioning, and deviation correction system based on industrial vision, applied to the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision as described in the first aspect; the nuclear fuel assembly grasping, positioning, and deviation correction system based on industrial vision includes: The data acquisition module is used to acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints; The contour extraction module is used to extract the outer contour boundary of the target nuclear fuel assembly in three-dimensional space based on the multi-view image data and the structured light deformation image data, so as to obtain the initial outer contour boundary data. The pose localization module is used to perform pose localization based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, so as to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space. The deviation analysis module is used to perform deviation analysis based on the current pose parameters and preset standard pose parameters to obtain the current deviation data and the corresponding deviation analysis results. The correction output module is used to generate a pose correction instruction based on the current deviation data if the deviation analysis result is a correctable deviation, and then transmit the pose correction instruction to the robot grasping system.

[0007] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing computer software programs; and a processor for reading and executing the computer software programs, thereby realizing the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision as described above.

[0008] Fourthly, the present invention also provides a non-transitory computer-readable storage medium storing a computer software program, which, when executed by a processor, implements the above-described method for grasping, locating, and correcting deviations of nuclear fuel assemblies based on industrial vision.

[0009] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for grasping, locating, and correcting deviations of nuclear fuel assemblies based on industrial vision.

[0010] The nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision provided in this invention provides multi-dimensional, high-precision basic data support by acquiring multi-view image data and structured light deformation image data of the target nuclear fuel assembly. This eliminates the need for manual intervention, avoiding the health risks of long-term human exposure to strong radiation. Based on the multi-dimensional image data, the initial outer contour boundary data of the target nuclear fuel assembly in three-dimensional space is extracted, accurately reflecting the actual spatial shape of the assembly. This overcomes the shortcomings of traditional laser and infrared positioning methods, which are susceptible to signal distortion due to electromagnetic interference, medium obstruction, and dust. Finally, the current pose is determined by combining the initial outer contour boundary data with the original design data of the target nuclear fuel assembly. The pose parameters enable precise perception of the spatial position and orientation of components, solving the problems of difficult control of positioning accuracy in manual operation and poor stability and repeatability of traditional positioning techniques. In addition, based on the current pose parameters and preset standard pose parameters, deviation analysis is performed to obtain the current deviation data and deviation analysis results, which can dynamically identify deviation changes in different scenarios. This breaks through the limitation that traditional fixed compensation rules cannot dynamically adapt to deviation changes. Finally, based on the current deviation data, a pose correction command is generated and transmitted to the robot grasping system, realizing accurate dynamic correction of deviation and automated control of grasping and positioning. This improves the accuracy, stability and repeatability of grasping and positioning, and effectively solves the safety problems of poor correction effect and easy collision damage to components in existing methods. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the nuclear fuel assembly grasping, positioning and deviation correction system based on industrial vision provided in an embodiment of the present invention; Figure 3 An embodiment diagram of the electronic device provided in this invention; Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] See Figure 1 , Figure 1This is a flowchart illustrating the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision provided by the present invention. In this embodiment, the execution entity of the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision is the grasping, positioning, and deviation correction system. Therefore, the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision includes: Step 10: Acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints.

[0014] Optionally, the grasping, positioning, and deviation correction system uses a pre-installed industrial vision acquisition device to simultaneously acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data from the corresponding viewpoints. The industrial vision acquisition device includes multiple sets of industrial cameras and structured light projection modules. The industrial cameras are distributed according to a preset spatial layout to ensure that the target nuclear fuel assembly can be photographed from different directions, covering the critical positioning area of ​​the assembly. The structured light projection module is configured one-to-one with each industrial camera. While the industrial cameras are capturing images, a preset pattern of structured light is projected onto the target nuclear fuel assembly. After the structured light illuminates the surface of the assembly, deformation occurs, and the industrial cameras simultaneously acquire the deformed structured light image, i.e., the structured light deformation image data. During the acquisition process, acquisition trigger commands can be sent to the industrial vision acquisition device via industrial Ethernet to control the synchronous operation of the industrial cameras and structured light projection modules, ensuring that the timestamps of the multi-view image data and the corresponding structured light deformation image data are consistent, avoiding data deviations caused by asynchronous acquisition.

[0015] In one embodiment, the industrial vision acquisition device of the grasping positioning and deviation correction system includes four acquisition units. Each acquisition unit consists of one Baslerac A2500-14gm high-resolution area array industrial camera and one Keyence LJ-V7000 series structured light projection module. The four acquisition units are respectively installed at the front, rear, left, and right of the target nuclear fuel assembly storage area. The horizontal distance between each acquisition unit and the assembly is set to 1.5m, and the shooting angle is at a 30° angle to the central axis of the assembly, ensuring complete acquisition of multi-view images of key positioning features such as the top positioning pin and contour edges of the assembly. The structured light projection module projects sinusoidal fringe structured light with a fringe spacing of 0.5mm, and the projection intensity is set to 800cd / m² via system preset parameters. 2To adapt to the lighting environment of nuclear industry sites, synchronous acquisition trigger commands are sent to four acquisition units via the PROFINET bus, with the trigger frequency set to 10Hz. The resolution of the images acquired by the industrial camera is 2592×1944 pixels, and the pixel size is 3.45μm×3.45μm. The timestamp deviation between the structured light deformation image data and the corresponding multi-view image data is controlled within ±1ms. After acquisition, the image data is transmitted through the GigEVision interface with a transmission delay of less than 10ms.

[0016] Step 20: Extract the outer contour boundary of the target nuclear fuel assembly in three-dimensional space based on multi-view image data and structured light deformation image data to obtain the initial outer contour boundary data.

[0017] Optionally, since the nuclear fuel assembly is subjected to industrial visual environments such as underwater, strong reflection, and lack of texture during the automated grasping process, these industrial visual environments can affect the contour extraction of the target nuclear fuel assembly. Therefore, after obtaining the acquired multi-view image data and structured light deformation image data, the grasping positioning and deviation correction system extracts the outer contour boundary of the target nuclear fuel assembly in three-dimensional space with high precision and robustness, and obtains the complete, closed and geometrically accurate three-dimensional outer contour boundary of the target assembly, forming the initial outer contour boundary data. This data is presented in the form of a three-dimensional coordinate set, as shown in steps 201-205.

[0018] Step 30: Based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, the pose is located to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space.

[0019] Optionally, after obtaining high-precision and robust initial outer contour boundary data, the grasping, positioning, and deviation correction system calls the pre-stored original design data of the target nuclear fuel assembly. This data includes the three-dimensional model parameters of the target nuclear fuel assembly (such as assembly length, width, height, design position coordinates of positioning pins, design three-dimensional coordinate set of outer contour boundary, etc.). The system performs pose positioning calculation on the initial outer contour boundary data determined under harsh industrial vision conditions such as no texture, high reflectivity, underwater optical distortion, and partial occlusion, thereby calculating the current pose parameters of the target nuclear fuel assembly in three-dimensional space, as described in steps 301-305.

[0020] Step 40: Perform deviation analysis based on the current pose parameters and the preset standard pose parameters to obtain the current deviation data and the corresponding deviation analysis results.

[0021] Optionally, after obtaining the current pose parameters, the capture, positioning, and deviation correction system performs deviation analysis on them against the preset standard pose parameters of the target nuclear fuel assembly to determine the current deviation data, as described in steps 401-404. Then, based on the determined current deviation data, it determines the deviation analysis result at that moment, as described in steps 405-406. The standard pose parameters refer to the pose parameters of the target nuclear fuel assembly in an ideal installation state or design reference state, including standard translation parameters and standard rotation parameters. The current deviation data includes translational deviation components and rotational deviation components. The deviation analysis results include correctable and uncorrectable deviations. When the deviation analysis result is an uncorrectable deviation, a safety alarm is directly triggered to remind personnel to make adjustments. Subsequent instruction generation continues only when the deviation analysis result is a correctable deviation.

[0022] Step 50: If the deviation analysis result is a correctable deviation, then generate an instruction based on the current deviation data to obtain a pose correction instruction, and transmit the pose correction instruction to the robot grasping system.

[0023] Optionally, when the deviation analysis result indicates a correctable deviation, the grasping, positioning, and deviation correction system calculates the required position compensation and attitude adjustment angles for the robot's end effector based on the translational and rotational deviation components in the current deviation data. This generates correction commands. During generation, the position compensation and translational deviation components are equal in magnitude but opposite in direction, as are the attitude adjustment angles and rotational deviation components. Following the robot grasping system's communication protocol, the position compensation and attitude adjustment angles are encapsulated into pose correction commands, which include command identifiers, compensation parameters, and execution priority. These commands are then transmitted to the robot grasping system via an industrial bus, ensuring real-time and reliable transmission. Simultaneously, the command transmission status is fed back to the system's central control module, completing the command issuance process.

[0024] This invention provides multi-dimensional, high-precision basic data support by acquiring multi-view image data and structured light deformation image data of the target nuclear fuel assembly. This eliminates the need for manual intervention, avoiding the health risks of long-term human exposure to strong radiation. Based on the multi-dimensional image data, the initial outer contour boundary data of the target nuclear fuel assembly in three-dimensional space is extracted, accurately reflecting the actual spatial morphology of the assembly. This overcomes the shortcomings of traditional laser and infrared positioning methods, which are susceptible to electromagnetic interference, medium obstruction, and dust-induced signal distortion. Furthermore, based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, pose positioning is performed to obtain the current pose parameters, achieving spatial positioning of the assembly. Precise perception of position and attitude solves the problems of difficult control of positioning accuracy in manual operation and poor stability and repeatability of traditional positioning technology. In addition, based on the current pose parameters and preset standard pose parameters, deviation analysis is performed to obtain the current deviation data and deviation analysis results. It can dynamically identify deviation changes in different scenarios, breaking through the limitation that traditional fixed compensation rules cannot dynamically adapt to deviation changes. Finally, based on the current deviation data, a pose correction command is generated and transmitted to the robot grasping system, realizing precise dynamic correction of deviation and automated control of grasping and positioning. This improves the accuracy, stability and repeatability of grasping and positioning, and effectively solves the safety problems of poor correction effect and easy collision damage to components in existing methods.

[0025] In one embodiment, steps 201-205 include: Step 201: Perform pixel-level geometric consistency verification based on multi-view image data and structured light deformation image data to obtain a set of multi-source pixels that have passed the verification.

[0026] Optionally, the capture positioning and deviation correction system, based on the acquired multi-view image data and the structured light deformation image data under the corresponding viewpoints, performs pixel-level geometric consistency verification on the image data under each set of viewpoints, using the system's preset world coordinate system as a reference. Specifically, it first determines the pixel correspondence between images from different viewpoints, and maps the pixels in the images from different viewpoints to the world coordinate system based on camera calibration parameters (intrinsic and extrinsic matrices). It then calculates the theoretical pixel coordinate values ​​corresponding to the same physical point in images from different viewpoints. Next, it compares the theoretical pixel coordinate values ​​with the actually acquired pixel coordinates to calculate the coordinate deviation value. Simultaneously, it combines the structured light encoding phase consistency of the corresponding pixels in the structured light deformation image data. If the phase difference exceeds a preset phase threshold, the pixel is determined to be an invalid pixel. Only when the pixel coordinate deviation value is less than or equal to the preset coordinate deviation threshold, and the structured light encoding phase difference value is less than or equal to the preset phase threshold, is the pixel determined to have passed the geometric consistency verification. Finally, all the verified pixels from all perspectives are aggregated to form a set of verified multi-source pixels. This set contains the perspective identifier, image coordinates, and structured light encoded phase information of each pixel, thereby eliminating erroneous phase correspondences caused by environmental interference during structured light decoding and removing false bright fringes caused by water surface disturbance, lens damage, or structured light defocus, ensuring the physical authenticity of subsequent 3D point coordinates.

[0027] In one embodiment, taking the acquisition of multi-view image data from four perspectives and corresponding structured light deformation image data as an example, the intrinsic parameter matrix in the camera calibration parameters is... (focal length Pixels, principal point coordinates The extrinsic parameter matrix contains the translation vectors and rotation matrices of each camera viewpoint relative to the world coordinate system. Based on the perspective projection model, pixels in the first-view image are... Points mapped to the world coordinate system Then Projecting the images onto the second, third, and fourth viewpoints yields the theoretical pixel coordinates. , , Calculate the deviation between the actual pixel coordinates and the theoretical pixel coordinates: The preset coordinate deviation threshold is 2 pixels. Simultaneously, the structured light encoded phase value of the corresponding pixel is extracted. Calculate the phase difference value The preset phase threshold is For the first-view image, the coordinates are... The theoretical pixel coordinates of the pixels mapped to the world coordinate system and projected onto the second-viewpoint are: The actual pixel coordinates are , , All are less than 2 pixels; phase value , , Less than The pixel passed the verification. Similarly, pixels from all viewpoints were verified one by one, resulting in a multi-source pixel set containing 800,000 verified pixels, each pixel accompanied by a viewpoint identifier (1-4) and image coordinates. and phase value .

[0028] Step 202: Based on the structured light encoded phase information of each pixel in the multi-source pixel set and the corresponding position in the structured light deformation image under the corresponding viewpoint, determine the three-dimensional spatial point corresponding to the pixel, and construct an initial three-dimensional point cloud dataset based on the three-dimensional spatial points of all pixels.

[0029] Optionally, the capture, localization, and deviation correction system extracts the image coordinates, corresponding camera calibration parameters, and structured light coded phase information for each pixel in the determined multi-source pixel set. Then, based on the mapping relationship between the structured light coded phase and depth, it calculates the depth value corresponding to each pixel, which is the pixel's Z-axis coordinate in the world coordinate system. Combining the camera intrinsic and extrinsic parameter matrices, it uses inverse perspective projection to convert the pixel's image coordinates (u, v) and depth value Z into 3D spatial point coordinates (X, Y, Z) in the world coordinate system. This calculation is performed on each pixel in the multi-source pixel set to obtain the corresponding 3D spatial point. Finally, all 3D spatial points are aggregated, and duplicate points are removed (points with a distance less than 0.01 mm are considered duplicates), constructing an initial 3D point cloud dataset. This dataset is stored in point cloud file format and contains the coordinate information and corresponding viewpoint identifier for each 3D spatial point.

[0030] Continuing with the above embodiments, the coordinates of the third viewpoint in the multi-source pixel set are extracted as follows: The pixel, its structured light encoded phase value The stripe spacing of the structured light projection module Projection angle For example, the mapping formula between phase and depth is: ;in, The reference depth for the structured light projection module (preset as) Substitute the data to calculate the depth value. In the camera extrinsic matrix from this perspective, the translation vector... Rotation matrix The identity matrix, combined with the intrinsic parameter matrix Through the inverse perspective projection transformation formula: Substitute Calculations yielded , That is, the coordinates of the three-dimensional point corresponding to this pixel are By calculating each of the 800,000 pixels in the multi-source pixel set and removing 10,000 duplicate points, an initial 3D point cloud dataset containing 790,000 3D spatial points is constructed, with a point cloud density of 8 points per square millimeter.

[0031] Step 203: Based on each point in the initial 3D point cloud dataset, perform neighborhood curvature-guided selection of boundary candidate points to obtain a set of boundary candidate points.

[0032] Optionally, after obtaining the initial 3D point cloud dataset, the grasping, localization, and deviation correction system performs a neighborhood search for each 3D spatial point in the dataset, using the K-Nearest Neighbors (KNN) algorithm to find the K nearest neighbors for each point (K is preset to 20). Subsequently, the normal vector of the local plane formed by each point and its neighbors is calculated, and the covariance matrix of the local point set is solved using Principal Component Analysis (PCA). The eigenvectors corresponding to the eigenvalues ​​of the covariance matrix are the normal vector and tangent direction vector of the local plane. Based on the normal vector, the neighborhood curvature of each point is calculated. The curvature calculation formula is the ratio of the eigenvalues; a larger curvature value indicates a more drastic surface change at the point's location, making it more likely to be a boundary point. Finally, a preset curvature threshold is set, and 3D spatial points with neighborhood curvature greater than or equal to the threshold are selected as boundary candidate points. All boundary candidate points are then aggregated to form a boundary candidate point set, which includes the 3D coordinates and neighborhood curvature values ​​of each candidate point.

[0033] Continuing with the above embodiments, the boundary candidate point selection module uses the KNN algorithm and sets... The coordinates of the initial 3D point cloud dataset are A neighborhood search is performed on the given point in three-dimensional space, finding 20 neighboring points. The covariance matrix is ​​then calculated for the local point set formed by the given point and its 20 neighbors. Its formula is ;in, The coordinates of the neighboring points, Let be the coordinates of the center point of the local point set. Solve for the three eigenvalues ​​of the covariance matrix. The corresponding feature vector ,in Let be the normal vector of the local plane. Then the neighborhood curvature The calculation formula is: The preset curvature threshold is 0.3, and the eigenvalues ​​of the covariance matrix of the above three-dimensional spatial points are... ,but The selection criteria are not met; the coordinates of another 3D point are... The eigenvalues ​​of the covariance matrix of the local point set formed by its neighboring points are... , This point was selected as a candidate boundary point. The process involved calculating the boundary points for each of the 790,000 points in the initial 3D point cloud dataset, ultimately resulting in a candidate boundary point set containing 50,000 3D spatial points.

[0034] Step 204: Perform directional consistency clustering based on any two points in the boundary candidate point set to obtain a set of boundary segments with consistent orientation.

[0035] Optionally, the grasping, positioning, and deviation correction system calculates the local tangent direction vector for each boundary candidate point in the determined boundary candidate point set. This local tangent direction vector is derived from the eigenvector of the covariance matrix obtained in step 203. and Determine (tangent direction in) and (Within the plane formed). Then, the density clustering algorithm (DBSCAN) is used, with points in the boundary candidate point set as the clustering objects. The cluster radius (preset 0.5mm) and minimum number of cluster points (preset 10) are set. During the clustering process, not only is it determined whether the spatial distance between two points is less than or equal to the cluster radius, but the angle between the local tangent direction vectors of the two points is also calculated. If the angle is less than or equal to a preset direction angle threshold (preset 15°), the two points are considered to have the same direction and can be classified into the same cluster. Each cluster corresponds to a continuous boundary segment. All boundary segments obtained from clustering are aggregated to form a set of boundary segments with consistent directions. Each boundary segment contains the 3D coordinates and local tangent direction vectors of all points within the segment, thereby eliminating boundary point breaks caused by occlusion or reflection loss and restoring continuous edge direction.

[0036] Continuing with the above embodiments, the coordinates of the candidate boundary points calculated by the grasping, positioning, and deviation correction system are as follows: Local tangent direction vector of candidate point (Unit vector), the coordinates of another candidate point are Its local tangent direction vector (Unit vectors), first calculate the angle between the two vectors. The formula is Substituting the data yields ,but And the spatial distance between the two points is Therefore, the two points belong to the same cluster. The DBSCAN algorithm is used to cluster 50,000 candidate boundary points, and finally 80 boundary segments with consistent orientation are obtained. Each segment contains 500-800 three-dimensional spatial points, forming a set of boundary segments with consistent orientation.

[0037] Step 205: Based on the set of boundary segments, perform outer contour refinement to obtain initial outer contour boundary data.

[0038] Optionally, the capture, positioning and deviation correction system performs outer contour refinement for each boundary segment in the determined set of boundary segments, including coplanarity verification, topological closure detection, component geometric prior matching and multi-view spatial alignment refinement process, to reconstruct and fuse complete, closed and globally consistent initial outer contour boundary data that conforms to the physical structure characteristics of nuclear fuel assemblies, as specifically in steps 2051-2054.

[0039] This invention ensures data reliability by eliminating invalid data through pixel-level geometric consistency verification; it accurately solves three-dimensional spatial points based on structured light encoded phase information, constructs a high-density initial three-dimensional point cloud, and then accurately locates boundary candidate points and forms continuous boundary segments through neighborhood curvature-guided screening and direction consistency clustering. Finally, it obtains high-fidelity initial outer contour boundary data through outer contour refinement, effectively resisting noise interference in the nuclear industry environment and making the fitting error between the extracted initial outer contour boundary data and the actual outer contour of the component smaller.

[0040] In one embodiment, the process of steps 2051-2054 includes: Step 2051: Perform coplanarity verification on each boundary segment in the boundary segment set to obtain a subset of coplanar boundary segments.

[0041] Optionally, the grasping, localization, and deviation correction system, based on a determined set of boundary segments, performs coplanarity verification on each boundary segment in the set. Each boundary segment contains multiple 3D spatial points and their corresponding local tangent direction vectors. Specifically, for a single boundary segment, the Random Sample Consensus Algorithm (RANSAC) is used to fit a candidate plane containing all 3D spatial points within that segment, yielding the plane equation. (in (using standardized normal vectors). Then, the distance from each 3D point within the boundary segment to the candidate plane is calculated using the following formula: ; ( For the first (Coordinates of 3D spatial points). A preset distance threshold (set based on the accuracy requirements of the outer contour of nuclear fuel assemblies) is used. The proportion of 3D spatial points within a boundary segment whose distance to the candidate plane is less than or equal to the distance threshold is counted out of the total number of points in that segment. If the proportion is greater than or equal to the preset proportion threshold, the boundary segment is determined to meet the coplanarity requirement and included in the coplanar boundary segment subset. This process is repeated, verifying the coplanarity of each boundary segment in the boundary segment set one by one. Finally, all boundary segments that meet the coplanarity requirement are summarized to obtain the coplanar boundary segment subset.

[0042] Continuing with the above embodiment, the boundary segment set contains 80 boundary segments with consistent orientations, each containing 500-800 three-dimensional spatial points. Verification is performed on one of these boundary segments (containing 600 three-dimensional spatial points) using the following process: The RANSAC algorithm is used to fit a candidate plane, with 1000 iterations and an interior point threshold of [value missing]. Finally, the plane equation is obtained. And satisfy Calculate the distance from each 3D point within the segment to the plane, with a preset distance threshold. The ratio threshold is Statistical analysis revealed that 585 points in three-dimensional space are less than or equal to the distance from the plane. The proportion is The boundary segment satisfies the coplanarity requirement. After verifying each of the 80 boundary segments, a subset of coplanar boundary segments containing 65 boundary segments is finally obtained, and the coplanarity of each segment meets the proportional threshold requirement.

[0043] Step 2052: Perform topological closure detection on each segment in the coplanar boundary segment subset to obtain a set of closed contour segments that are determined to be closed contours.

[0044] Optionally, after obtaining the subset of coplanar boundary segments, the grasping, positioning, and deviation correction system performs topological closure detection on each segment in the subset, that is, it first extracts the three-dimensional coordinates of the starting point of the fitted curve of that segment. and the three-dimensional coordinates of the endpoint Calculate the spatial distance between the starting point and the ending point. Then, based on the preset closure distance threshold (set according to the point density of boundary segments and component contour accuracy), if the spatial distance... If the distance is less than or equal to the closed distance threshold, then the tangent direction vector at the starting point is further calculated. tangent direction vector at the endpoint The included angle ( (All are unit vectors), and then based on the preset tangent angle threshold (to ensure a smooth transition of the contour at the closure). If the angle If the angle between the tangents is less than or equal to the threshold, the fitted curve of the boundary segment is determined to be a topologically closed contour; if the spatial distance... Greater than the closing distance threshold, or the included angle If the angle between the boundary segments exceeds the threshold of the tangent, it is determined to be a non-closed contour. Topological closure detection is performed on all boundary segments in the coplanar boundary segment subset one by one, and all boundary segments determined to be closed contours are summarized to obtain the closed contour segment set.

[0045] Continuing with the above embodiment, the coplanar boundary segment subset contains 65 boundary segments. The grasping, positioning, and deviation correction system detects one of these boundary segments (the fitted curve contains 300 discrete points, with a point density of 5 points per millimeter): first, it extracts the starting coordinates of the fitted curve. End point coordinates Calculate spatial distance The preset closing distance threshold is , Simultaneously extract the starting point tangent direction vector. , End point tangent direction vector Calculate the dot product included angle The preset threshold for the included angle of the tangent is , If the boundary segment is detected, it is determined to be a closed contour. After detecting each of the 65 boundary segments, a set of closed contour segments containing 40 closed contours is finally obtained. The spatial distance between the start and end points of each closed contour is... The included angle of the tangents is .

[0046] Step 2053: Perform geometric prior matching of the target nuclear fuel assembly based on each closed contour in the set of closed contour segments to obtain the outer contour segment that conforms to the shape constraints of the assembly.

[0047] Optionally, after obtaining the set of closed contour segments, the grasping, positioning, and deviation correction system performs geometric prior matching of the target nuclear fuel assembly for each closed contour in the set. The grasping, positioning, and deviation correction system pre-stores geometric prior data of the target nuclear fuel assembly, which includes geometric parameters of the assembly's key closed contours, such as contour type (circular, rectangular, polygonal, etc.), size parameters (diameter, side length, circumscribed circle radius, etc.), and positional constraints (relative distances and angular relationships between contours, etc.). Therefore, the fitted curve of each closed contour is called to extract its geometric feature parameters: if it is a circular contour, the coordinates of the center are calculated. and diameter If the outline is rectangular, calculate the coordinates of the four vertices and the length of the longest side. and the length of the shorter side If the outline is polygonal, the number of vertices, the length of each side, and the interior angles are calculated. The extracted geometric feature parameters are then compared with the corresponding parameters in the prior geometric data to calculate the parameter deviation rate. ;in, To detect the obtained geometric parameters, These are the geometric prior parameters. Then, based on a preset parameter deviation threshold (set according to component design tolerances), if the deviation rate of all geometric feature parameters is less than or equal to the parameter deviation threshold, the closed contour is determined to conform to the shape constraints of the target nuclear fuel assembly and is considered an outer contour segment; otherwise, it is determined to be a non-target contour and discarded. Geometric prior matching is performed on all closed contours in the closed contour segment set one by one, and all closed contours conforming to the shape constraints are summarized to obtain the outer contour segment.

[0048] Continuing with the above embodiments, in the geometric prior data of the target nuclear fuel assembly, the key closed profile includes the circular profile of four locating pins (geometric prior parameter: diameter). The offset of the center coordinates relative to the component center is ) and the rectangular outline of one component body (geometric prior parameters: long side length) Short side length (The center coincides with the component center), and the preset parameter deviation threshold is... The algorithm detects a closed contour within a set of closed contour segments and extracts its geometric feature parameters, identifying it as a circular contour with center coordinates of [coordinates missing]. (Component center coordinates are) The offset is ),diameter Calculate the deviation rate It conforms to the shape constraints. The other closed contour is extracted as a rectangular contour, with the longer side measuring [length missing]. Short side length The deviation rates are respectively , ,all The shape constraints are met. After matching the 40 closed contours one by one, 10 non-target contours (such as small closed contours formed by dust interference) are removed, resulting in 30 outer contour segments that meet the component shape constraints, covering 4 locating pin contours and 26 component body-related contours.

[0049] Step 2054: Spatial alignment and fusion are performed based on the outer contour segments to obtain the initial outer contour boundary data of the target nuclear fuel assembly.

[0050] Optionally, during the spatial alignment and fusion process after obtaining the outer contour segments, the grasping, positioning, and deviation correction system uses the reference coordinates of the component (component design center coordinates) in the geometric prior data as a reference to spatially align all outer contour segments. It then employs the Iterative Closest Point (ICP) algorithm to register the geometric center of each outer contour segment with the design center of the corresponding contour in the geometric prior data, calculates the spatial transformation matrix (translation and rotation matrices), and maps the outer contour segments to the target position in a unified world coordinate system. Next, the aligned outer contour segments are fused. For 3D coordinate points in overlapping areas, the optimal coordinate points are retained (points closer to the design contour have higher priority); for outer contour segments in non-overlapping areas, their discrete coordinate points are directly retained. Finally, according to the outer contour topology of the target nuclear fuel assembly, all fused coordinate points are arranged in an ordered manner to form a continuous and complete set of 3D coordinate points. This set of coordinate points serves as the initial outer contour boundary data, with a density of 5 points per millimeter to ensure accurate representation of the component's outer contour shape.

[0051] Continuing with the above embodiment, the outer contour segment contains 30 contours that meet the constraints. The grasping, positioning, and deviation correction system uses the component design center coordinates (150mm, 150mm, 500mm) in the geometric prior data as a reference to perform ICP registration on four of the locating pin outer contour segments. Taking a certain locating pin outer contour as an example, its current geometric center is (100mm, 100mm, 500mm), and its design center is (100mm, 100mm, 500mm). It can be directly aligned without translation or rotation. The geometric center of another component body outer contour segment is (150.2mm, 150.1mm, 500mm). The translation matrix T = (-0.2mm, -0.1mm, 0mm) and the rotation matrix is ​​the identity matrix are calculated through the ICP algorithm. After transformation, it is aligned to the design center. After alignment, for the overlapping edge area of ​​the component's main body contour and the locating pin contour, discrete points of the two contours are extracted, and the distance from each point to the design contour is calculated. Points with smaller distances are retained (e.g., a point on the design contour with a distance of 0.01mm is prioritized over a point with a distance of 0.03mm). Finally, according to the topological order of the component's outer contour from top to bottom and from the main body to the locating pin, all the fused coordinate points are arranged in an orderly manner, ultimately forming initial outer contour boundary data containing 10,000 three-dimensional coordinate points. This data covers all key outer contours of the component, and the fitting error with the actual outer contour of the component is less than 0.05mm.

[0052] This invention eliminates non-planar redundant segments through coplanarity verification, ensuring spatial consistency of the contour. Topological closure detection filters out closed contours that conform to the component's shape characteristics, reducing interference factors. Geometric prior matching ensures precise correspondence between segments and component design features, improving contour effectiveness. Spatial alignment fusion integrates all valid segments, forming complete and continuous outer contour data. Ultimately, this effectively filters noise and interference in the nuclear industrial environment, providing highly accurate data support for subsequent pose positioning and ensuring the overall accuracy and reliability of nuclear fuel assembly grasping, positioning, and deviation correction.

[0053] In one embodiment, the process of steps 301-305 includes: Step 301: Extract geometrically corresponding point pairs based on the initial outer contour boundary data and the theoretical outer contour boundary data of the original design data in the target nuclear fuel assembly to obtain the initial corresponding point pair group.

[0054] Optionally, the capture, positioning, and deviation correction system uses the obtained initial outer contour boundary data (containing 10,000 three-dimensional coordinate points) and the theoretical outer contour boundary data from the original design data of the standard nuclear fuel assembly (also containing 10,000 three-dimensional coordinate points, arranged in the same topological order as the coordinate points in the initial outer contour boundary data). Using the three-dimensional coordinate points of the theoretical outer contour boundary data as a reference, a k-nearest neighbor search algorithm (k=3) is employed to find the three closest points in the theoretical outer contour boundary data for each three-dimensional coordinate point in the initial outer contour boundary data, calculating the Euclidean distance between each candidate point and the target point. The candidate point with the smallest distance is selected as the corresponding theoretical point of the initial outer contour boundary point, forming a geometrically corresponding point pair (initial outer contour boundary point coordinates, theoretical outer contour boundary point coordinates). Furthermore, to ensure the validity of the corresponding point pairs, a preset distance threshold (set based on the component design tolerance and the initial outer contour extraction accuracy) is used. If the minimum Euclidean distance is greater than the distance threshold, the initial outer contour boundary point is determined to have no valid corresponding theoretical point and is discarded. Repeat this process to match all three-dimensional coordinate points in the initial outer contour boundary data one by one, and summarize all valid geometric corresponding point pairs to form an initial corresponding point pair group.

[0055] In one embodiment, a certain three-dimensional coordinate point in the initial outer contour boundary data is Within the candidate regions corresponding to the topological order in the theoretical outer contour boundary data, through Nearest neighbor search Three theoretical points were found: , , The calculated Euclidean distance is: ; ; Minimum distance The preset distance threshold is , ,therefore and To form valid corresponding point pairs, 10,000 points in the initial outer contour boundary data are matched one by one, and 300 noise points without valid corresponding points are removed, finally resulting in an initial corresponding point pair group containing 9,700 sets of valid geometric corresponding point pairs.

[0056] Step 302: Based on all point pairs in the initial corresponding point pair group, perform rigid body transformation parameter analysis to obtain the initial pose transformation matrix.

[0057] Optionally, the grasping, positioning, and deviation correction system uses an orthogonal iterative algorithm to solve for the rigid body transformation parameters, including the translation vector, based on all valid corresponding point pairs in the determined initial corresponding point pair group. and rotation matrix (3×3 orthogonal matrix), and the rigid body transformation satisfies the formula: ,in, For the initial corresponding point pair group, the first The initial outer contour boundary point coordinates of the group These are the coordinates of the corresponding theoretical outer contour boundary points. This is the error vector. That is, first calculate the center point of the initial outer contour boundary points in the initial corresponding point pair group. The center point of the theoretical outer contour boundary point Decentralize all corresponding point pairs to obtain a decentralized point set. , Construct the covariance matrix ( (the number of point pairs in the initial corresponding point pair group), for the covariance matrix Perform singular value decomposition (SVD) to obtain ,in and It is an orthogonal matrix. It is a diagonal matrix. The rotation matrix is ​​calculated based on the singular value decomposition results. ,like (If the determinant is -1, it indicates that a mirror transformation exists, which does not meet the requirements of a rigid body transformation), then adjust... The last column of symbols makes Translation vector Rotation matrix Translation vector Combined, we obtain the initial pose transformation matrix. (4×4 homogeneous transformation matrix).

[0058] Continuing with the above embodiments, the initial corresponding point pair group contains 9700 valid point pairs, and the center point of the initial outer contour boundary point is calculated. The center point of the theoretical outer contour boundary. After decentralizing all point pairs, the covariance matrix is ​​constructed. ,right Singular value decomposition yields , , Calculate the rotation matrix. , This meets the requirements of rigid body transformation. Translation vector The final initial pose transformation matrix is: .

[0059] Step 303: Based on the initial pose transformation matrix, perform spatial mapping on all feature control points in the original design data to obtain the mapped feature control point set.

[0060] Optionally, after obtaining the initial pose transformation matrix, the grasping, positioning, and deviation correction system first determines all feature control points in the original design data. These feature control points are the design coordinate points of key positioning features of the target nuclear fuel assembly, including key positions such as the center of the positioning pin at the top of the assembly, the inflection points of the contour edges, and the center of the assembly. Each feature control point contains three-dimensional design coordinates. Then, the initial pose transformation matrix is ​​used. For each feature control point, a spatial mapping is performed, and the mapping formula is a homogeneous coordinate transformation: ,in The coordinates of the mapped feature control points are shown below. The spatial mapping operation is performed on each feature control point in the original design data, and the coordinates of all mapped feature control points are then summarized to obtain the mapped feature control point set.

[0061] Continuing with the above embodiment, the original design data contains 24 feature control points, including 4 locating pin centers and 20 contour edge inflection points, where the design coordinates of one of the locating pin centers are... The design coordinates of the other contour edge inflection point are: The obtained initial pose transformation matrix is ​​used. Perform spatial mapping: For The homogeneous transformation yields: ; ; That is, the mapped coordinates are After mapping each of the 24 feature control points, a feature control point set containing the 24 mapped feature control points is obtained. Each point is the actual spatial position coordinate of the original design feature control point after the initial pose transformation.

[0062] Step 304: Perform local geometric structure matching based on the feature control point set and the initial outer contour boundary data to obtain a verification point set with consistent structure.

[0063] Optionally, after obtaining the feature control point set, the grasping, positioning, and deviation correction system searches for a local neighborhood point set (using a spherical neighborhood search with a preset neighborhood radius of 5mm) in the initial outer contour boundary data for each mapped feature control point in the feature control point set, obtaining the local outer contour point group corresponding to each feature control point. Then, it calculates the geometric structure parameters of each local outer contour point group, including the coordinates of the center point, covariance matrix, principal direction vector, etc.; simultaneously, it calculates the local design geometric structure parameters of the corresponding mapped feature control point in the original design data (based on the 3D model extraction from the original design data). The structural similarity is calculated by comparing the geometric structure parameters of the local outer contour point group with the local design geometric structure parameters, using the following formula: ;in, The angle between the main direction vector of the local outer contour point group and the main direction vector of the local design geometry; The Euclidean distance between the center point of the local outer contour point group and the mapped feature control point; The maximum allowable distance is preset (consistent with the neighborhood radius). A judgment is made based on a preset structural similarity threshold; if the structural similarity... If the values ​​are greater than or equal to the threshold, the local outer contour point group is determined to have the same local geometric structure as the feature control point, and all points in the local outer contour point group are included in the verification point set; otherwise, the local outer contour point group is removed. The local outer contour point groups corresponding to all mapped feature control points are verified one by one, and all locally outer contour point groups with consistent structures are summarized, removing duplicate points (those with a distance less than a certain threshold). (The points are determined to be duplicate points), thus obtaining a set of verification points with consistent structure.

[0064] Continuing with the above embodiments, the center of the positioning pin after the feature control points are mapped is... A spherical neighborhood search (radius) is used. In the initial outer contour boundary data, 80 local neighborhood points were found, forming a local outer contour point group. The coordinates of the center point of this point group were calculated as follows: The principal direction vector of the covariance matrix is The principal direction vector of the local design geometry of the locating pin in the original design data is: , , . , Structural similarity The preset structural similarity threshold is 0.9. The local outer contour point group was included in the verification point set. The local outer contour point groups corresponding to the 24 mapped feature control points were verified one by one. Two point groups with inconsistent structures were removed (due to local noise in the initial outer contour boundary data). After summarizing all valid point groups and removing 50 duplicate points, a verification point set with consistent structure containing 8500 three-dimensional coordinate points was obtained.

[0065] Step 305: Based on the verification point set, the initial pose transformation matrix, the original design data, and the initial outer contour boundary data, pose transformation and verification are performed to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space.

[0066] Optionally, after obtaining the verification point set, the grasping, positioning and deviation correction system combines it with the initial pose transformation matrix, the original design data and the initial outer contour boundary data to perform pose transformation and verification, and finally obtains the current pose parameters of the target nuclear fuel assembly in three-dimensional space, as in steps 3051-3054.

[0067] This invention extracts and filters out effective matching points through geometric correspondence point pairs, laying the data foundation for pose calculation. The initial pose transformation matrix is ​​obtained through rigid body transformation parameter analysis, and the correlation between design data and actual data is initially established. Then, the data effectiveness is further optimized based on feature control point spatial mapping and local geometric structure matching, and noise interference is eliminated. Finally, through pose transformation optimization and verification, high-precision current pose parameters are obtained.

[0068] In one embodiment, the process of steps 3051-3054 includes: Step 3051: Based on each verification point in the verification point set, a local coordinate system is constructed using the principal direction of the normal vector in its neighborhood. Alignment constraint analysis is then performed between the local coordinate system and the corresponding design points in the original design data to obtain the enhanced pose constraint conditions.

[0069] Optionally, the grasping, positioning, and deviation correction system, based on a determined set of verification points, applies the following to each verification point in that set: Nearest Neighbor Algorithm Find 15 neighboring verification points of the verification point to form a local point set. Then, calculate the covariance matrix of the local point set using Principal Component Analysis (PCA). Solve for the three eigenvalues ​​and corresponding eigenvectors of the covariance matrix. The eigenvector corresponding to the smallest eigenvalue is the principal direction of the normal vector in the neighborhood of the verification point. (Unit vector). Then, using verification points... With the origin as the origin, the principal direction of the normal vector is... For local coordinate system axis( (axis), within the tangent plane of the verification point (perpendicular to) (The plane), the local coordinate system is determined based on the principal tangent direction of the local point set. axis( (axis), then through cross product get axis( (axis), complete the local coordinate system for each verification point. The construction of.

[0070] Furthermore, the capture positioning and deviation correction system retrieves the design point corresponding to the verification point from the original design data. Extract design points Principal direction of the design normal vector in the original 3D design model (Unit vector), and construct design points. Local design coordinate system (The construction rules and the local coordinate system of the verification points are consistent.) The axis is Direction). Enhanced pose constraints are constructed by calculating the attitude and position deviations of the two local coordinate systems: the attitude constraint is the angle constraint between the direction vectors of each axis of the local coordinate system ( All must meet the preset angle constraint threshold); the position constraint is the verification point. With design points Euclidean distance constraints in the world coordinate system (must meet a preset position constraint threshold). Summarize the attitude and position constraints of all verification points to form an enhanced pose constraint set.

[0071] Continuing with the above embodiment, the verification point set contains 8500 verification points, and one of the verification points is selected. ,pass Nearest Neighbor Algorithm Find 15 neighborhood validation points, calculate the covariance matrix of the local point set, and obtain the eigenvalues. ( (Minimum), corresponding eigenvector (Principal direction of the normal vector). With With the origin as the point, for The Z-axis determines the direction vector of the principal tangent line within the tangent plane. (X-axis), via cross product (Y-axis) Construct a local coordinate system The corresponding design points in the original design data. Its design normal vector principal direction Construct a local design coordinate system Calculate the included angle of the attitude constraint: , , The preset angle constraint threshold is Position constraint distance The preset position constraint threshold is The constraints at this verification point are valid and are included in the enhanced pose constraint set. Each of the 8500 verification points is processed individually, ultimately forming an enhanced pose constraint set containing 8500 sets of constraints.

[0072] Step 3052: Perform iterative projection optimization based on the enhanced pose constraints and the initial pose transformation matrix to obtain the optimized target pose transformation matrix.

[0073] Optionally, the grasping, positioning, and deviation correction system first uses the initial pose transformation matrix. ( It is a 3×3 rotation matrix. Using a 3×1 translation vector as the initial value, a constrained nonlinear least squares optimization model is constructed. The objective function is to minimize the sum of the transformation errors between the verification point and the corresponding design point, as well as the deviations from the enhanced pose constraints. Therefore, the objective function of this optimization model is: ;in, The number of verification points; For the first Coordinates of the verification points; These are the coordinates of the corresponding design points; Let be the rotation matrix and translation vector to be optimized; For the first Local coordinate system axis vectors of each verification point; For the local design coordinate system axis vectors corresponding to the design points; Constraint weights are set to 10 (equivalent to the weight of the position error term). The Levenberg-Marquardt algorithm is then used to solve the optimization model. During the iteration process, the convergence characteristics of the Gauss-Newton method and the steepest descent method are balanced by adjusting the damping factor. When the iteration error is less than a preset convergence threshold... Alternatively, iteration can stop when the maximum number of iterations reaches a threshold (50 times), and the optimized rotation matrix is ​​output. Translation vector The target pose transformation matrix is ​​obtained by combining the results. .

[0074] Continuing with the above embodiments, the initial pose transformation matrix middle, , Using 8500 sets of enhanced pose constraints as constraints, a nonlinear least squares optimization model is constructed, and the following settings are made: The convergence threshold is The maximum number of iterations is 50. During the iteration process, the error converges to [value] on the 10th iteration. The iteration stops when the convergence condition is met. The optimized rotation matrix is... Translation vector The final target pose transformation matrix is: .

[0075] Step 3053: Based on the target pose transformation matrix, perform global reprojection on the complete outer contour in the original design data to obtain the reprojected outer contour data.

[0076] Optionally, the capture positioning and deviation correction system first retrieves the complete outer contour data from the original design data, which contains 20,000 three-dimensional design coordinate points of the complete outer contour of the target nuclear fuel assembly. This covers all external contour features, including the main body of the component, locating pins, and edge corners. Then, after obtaining the target pose transformation matrix, the target pose transformation matrix is ​​used... For each complete outer contour design point, a homogeneous coordinate transformation is performed, and its reprojection formula is: ;in, For the first The three-dimensional coordinates of each design point after reprojection. The reprojection transformation is performed on each of the 20,000 complete outer contour design points in the original design data. All the reprojected three-dimensional coordinate points are summarized and arranged in the original outer contour topological order to form the reprojected outer contour data. The coordinate point density of this data is consistent with that of the initial outer contour boundary data (5 points per millimeter) to ensure the consistency of contour comparison.

[0077] Continuing with the above embodiments, a design point of the complete outer contour in the original design data. Using the target pose transformation matrix Reprojection calculations yielded the following results: ; ; That is, the coordinates of the reprojected design point are... The reprojection transformation was performed on each of the 20,000 complete outer contour design points. After arranging them in the original topological order, a total of 20,000 three-dimensional coordinate points were obtained.

[0078] Step 3054: Perform contour closure deviation integral analysis based on the reprojected outer contour data and the initial outer contour boundary data to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space.

[0079] Optionally, the grasping, positioning, and deviation correction system first aligns the reprojected outer contour data with the initial outer contour boundary data. Using the center of the reprojected contour of the positioning pin at the top of the component as a reference, linear interpolation is used to adjust the coordinate point order of the initial outer contour boundary data, ensuring a one-to-one correspondence between the contour sampling points of the two sets of data (20,000 corresponding point pairs for each set). For each corresponding point pair, the Euclidean distance in three-dimensional space is calculated. ( For the initial outer contour boundary data, the first One point, For the reprojection of the outer contour data, the first (points), construct the deviation sequence The maximum Euclidean distance between all verification points and their corresponding design points is 0.0173 mm, and the average deviation is 0.0021 mm (based on 8 locating pin feature points). Based on the contour closure characteristics, the deviation integral value is calculated using the curve integral method, and its integral formula is as follows: ;in, The arc length of the closed contour; Contour arc length parameter The corresponding deviation function; The arc length between adjacent sampling points (calculated based on reprojected outer contour data, as in this embodiment). ). Then, if the integral value of the deviation If the value is less than or equal to a preset integration threshold (set based on component grasping accuracy requirements), the target pose transformation matrix is ​​deemed valid, and the current pose parameters are analyzed based on this matrix: translation parameters. , , ( The target translation vector (The three components); then the target rotation matrix is ​​obtained using the Rodriguez formula. Convert to wrap Rotation angle of the shaft That is, attitude parameters. If the integral value of the deviation... If the result exceeds the preset integration threshold, return to step 301 and start again until the requirement is met.

[0080] Continuing with the above embodiments, after aligning the reprojected outer contour data with the initial outer contour boundary data, 20,000 corresponding point pairs are formed. The calculation of a given corresponding point pair... and The Euclidean distance is: Calculate the deviation values ​​for all corresponding point pairs using this method, construct a deviation sequence, and set... Calculate the integral value of the deviation. According to the preset integration threshold ,like If the target pose transformation matrix is ​​valid, then it is determined that the target pose transformation matrix is ​​valid. And the Rodriguez formula is used to... Convert to rotation angle: ; ; The final pose parameters of the target nuclear fuel assembly are: translation parameters. , , Attitude parameters .

[0081] This invention introduces more refined geometric attitude constraints through local coordinate system construction and alignment constraint analysis, enhancing the stability of pose calculation. The nonlinear optimization model of the constraints effectively integrates position and attitude errors, improving the accuracy of the pose transformation matrix. Furthermore, the effectiveness of the pose parameters is verified from the overall contour level based on global reprojection and contour closure deviation integral analysis. Finally, the translation accuracy and rotation angle accuracy of the current pose parameters are improved.

[0082] In one embodiment, the process of steps 401-404 includes: Step 401: Construct a rigid body space alignment reference based on the current pose parameters and the standard pose parameters to obtain the reference alignment transformation matrix.

[0083] Optionally, the grasping, positioning, and deviation correction system uses the determined current pose parameters and standard pose parameters, where the current pose parameters include the current translation parameters. and current rotation parameters (3×3 rotation matrix, as obtained through step 3054) (obtained through Rodriguez formula conversion); these standard pose parameters include standard translation parameters. and standard rotation parameters (3×3 unit rotation matrix, corresponding to standard pose) The core of constructing a rigid body space alignment reference is to build a reference alignment transformation matrix from the current pose coordinate system to the standard pose coordinate system, and to ensure that this matrix cancels out the rigid body motion differences between the current pose and the standard pose. Reference Alignment Transformation Matrix for The homogeneous transformation matrix is ​​derived from the alignment rotation matrix. and alignment translation vector Composition, its expression is: Among them, the alignment rotation matrix ( for The transpose of the matrix (since the rotation matrix is ​​an orthogonal matrix, the transpose is equal to the inverse) is used to offset the rotational deviation of the current pose; aligning the translation vector. This is used to offset the translational deviation of the current pose, so as to ensure that the current pose can be spatially aligned with the standard pose after the transformation by this matrix.

[0084] In one embodiment, taking the current pose parameters obtained in step 3054 as an example, its translation parameters... Attitude parameters , , The attitude parameters are converted into the current rotation matrix using the Rodriguez formula. Standard pose parameters: , (Identity matrix). Calculate the alignment and rotation matrix. (because (Approximately symmetric, the transpose of which differs very little from the original matrix). Calculate the alignment translation vector. .

[0085] The final reference alignment transformation matrix is: .

[0086] Step 402: Based on the reference alignment transformation matrix, the coordinate system of the positioning feature point set of the target nuclear fuel assembly is transferred to obtain the aligned actual feature point set.

[0087] Optionally, after obtaining the reference alignment transformation matrix, the grasping, positioning, and deviation correction system calls the positioning feature point set of the target nuclear fuel assembly. This positioning feature point set is the actual three-dimensional coordinate set of the key positioning feature points identified in step 20 (such as the center of the positioning pin at the top of the assembly and the inflection point of the contour edge), denoted as... ( To determine the number of feature points, in this embodiment... Then, a reference alignment transformation matrix is ​​used. A homogeneous coordinate transformation is performed on each localization feature point to achieve the migration from the current pose coordinate system to the standard pose coordinate system. The migration formula is as follows: ;in, The coordinates of the localization feature points before migration. These are the coordinates of the migrated feature points. For the localized feature point set... Perform a migration operation on each point one by one, and summarize the coordinates of all migrated feature points to form the aligned set of actual feature points. .

[0088] Continuing with the above embodiments, the feature point set is located. It contains 24 feature points, and the center of one of them is selected as the positioning pin. (Based on the current pose coordinate system). A reference alignment transformation matrix is ​​used. The coordinate system migration calculation yielded the following results: ; ; That is, the coordinates of the feature points after migration are The migration operation was performed on each of the 24 localization feature points one by one, and the final aligned set of actual feature points was obtained. The coordinates of each feature point are approximately consistent with the coordinates of the designed feature points in the standard pose, and the deviation is within... Within.

[0089] Step 403: Perform point-by-point deviation vector analysis based on the actual feature point set and the actual observed feature point set determined in the current pose parameters to obtain the original deviation vector set.

[0090] Optionally, when the grasping, positioning, and deviation correction system performs point-by-point deviation vector analysis based on the obtained actual feature point set combined with the actual observed feature point set determined in the current pose parameters, the actual feature point set is... The actual set of observed feature points determined in the current pose parameters is: This set consists of the theoretical observation coordinates (i.e., the coordinates of the design features under the standard pose, transformed by the current pose) of the corresponding positioning feature points extracted from the original design data based on the current pose parameters. A point-by-point deviation vector calculation is performed on the corresponding points in the two feature point sets (corresponding one-to-one according to the feature point number). Deviation vector at each corresponding point The formula for subtracting the actual observed feature point coordinates from the aligned actual feature point coordinates is: ;in, The three components correspond to The deviation component of direction. For all Calculate the deviation vector for each corresponding point, and summarize them to obtain the original deviation vector set. .

[0091] Continuing with the above embodiments, the actual observed feature point set To design the coordinates of feature points in the standard pose after transformation to the current pose, taking the first feature point as an example, the coordinates of the designed feature point are: After the current pose transformation matrix After transformation, we get Aligned actual feature points The deviation vector is calculated as follows: The original deviation vector set was obtained by calculating each of the 24 corresponding points. , of which Deviation vector (It should be noted that due to slight differences in the positioning errors of each feature point, the deviation vector components exist.) (fluctuations).

[0092] Step 404: Decouple the rigid body motion mode based on the original deviation vector set to obtain the translational deviation component and the rotational deviation component.

[0093] Optionally, the grasping positioning and deviation correction system uses the obtained original deviation vector set... Based on the spatial distribution of positioning feature points, the deviation vector is decomposed into translational deviation components and rotational deviation components. Specifically, the translational deviation components are calculated first. The translational deviation component is the mean of all original deviation vectors, and its formula is: ; in, The first Deviation vector exist Component of direction; This represents the number of feature points. Next, the rotational deviation components are calculated. Based on the rotational characteristics of a rigid body, the deviation vector generated by the rotational deviation at the characteristic point... satisfy (cross product), where This is the angular displacement vector corresponding to the rotational deviation component. This represents the position vector of the actual observed feature point. Combined with the original deviation vector... Construct a system of linear equations: The cross product is expanded into a matrix form, and the angular displacement vector is solved using the least squares method. This is the rotational deviation component. .

[0094] Continuing with the above embodiment, the original deviation vector set contains 24 deviation vectors, then the translational deviation components are: ; ; That is, the translational deviation component. Four non-coplanar actual observation feature points (centers of the positioning pins) were selected, with coordinates as follows: , , ,

[0095] calculate (Since translational deviations have been removed, the remaining deviations consist of rotational deviations and measurement noise.) Therefore, a system of linear equations is constructed and solved. Finally, the rotational deviation component is obtained: , , That is, the rotational deviation component. In summary, the current deviation data is: translational deviation component. Rotational deviation component .

[0096] This invention provides a unified coordinate system for deviation analysis by constructing a spatial alignment reference, ensuring the consistency of deviation calculation. Then, based on coordinate system migration, it realizes the mapping of actual feature points to the standard coordinate system, eliminating the interference of pose differences. Furthermore, point-by-point deviation vector analysis captures the deviation details of each key feature point. Finally, through rigid body motion mode decoupling, it accurately separates the translational and rotational deviation components, providing clear targets for subsequent deviation correction, thereby improving the accuracy of the current deviation data.

[0097] In one embodiment, the process of steps 405-406 includes: Step 405: Perform three-dimensional vector amplitude quantization processing on the translational deviation component and the rotational deviation component respectively to obtain the total position deviation and the total attitude deviation.

[0098] Optionally, after obtaining the translational and rotational deviation components, the grasping positioning and deviation correction system converts the translational deviation component into a three-dimensional vector. Characterizing the target nuclear fuel assembly in The degree of deviation from the standard pose in three straight lines; the rotational deviation component is a three-dimensional vector. Characterizing the component around The deviation angles of the three coordinate axes from the standard pose are given. Therefore, the translational deviation components are quantized using three-dimensional vector amplitude quantization, and the total position deviation is calculated using the Euclidean distance formula. Its formula is: That is, by integrating the translational deviations in three directions, a single quantized value of the total position deviation is obtained, which can intuitively reflect the degree of positional deviation of the entire component. Simultaneously, the rotational deviation component is quantized using three-dimensional vector amplitude, and the total attitude deviation is calculated using the Euclidean norm (L2 norm). Its formula is: That is, by integrating the deviation angles of the three rotation directions, the overall attitude deviation is quantified by the norm measurement, which can accurately reflect the comprehensive degree of component attitude deviation.

[0099] Step 406: Based on the total position deviation and the total attitude deviation, a comparative analysis is performed using preset position deviation thresholds and attitude deviation thresholds. If the total position deviation is less than or equal to the position deviation threshold and the total attitude deviation is less than or equal to the attitude deviation threshold, the deviation analysis result is determined to be a correctable deviation. If the total position deviation is greater than the position deviation threshold or the total attitude deviation is greater than the attitude deviation threshold, the deviation analysis result is determined to be an uncorrectable deviation, and a safety alarm is triggered.

[0100] Optionally, the grasping positioning and deviation correction system obtains the total position deviation. and total attitude deviation Then, the preset position deviation threshold is invoked. and attitude deviation threshold Among them, the position deviation threshold This is the maximum permissible positional deviation set based on the mechanical precision of nuclear fuel assembly grasping, the robot's adjustment capability, and the safe clearance of the assemblies; attitude deviation threshold. This is the maximum allowable posture deviation set based on avoiding collisions and ensuring fitting accuracy during component grasping. The total positional deviation... With position deviation threshold For comparison, the total attitude deviation With attitude deviation threshold By comparison, we obtain: like and If the deviation is within the adjustment capability of the robot grasping system, the deviation analysis result is a correctable deviation, and the system outputs a correctable signal to the subsequent correction output module. like or If the current deviation exceeds the robot's adjustment capability, continuing to grasp may lead to component collision, damage, or grasping failure. The deviation analysis result is an uncorrectable deviation, and the system triggers a safety alarm. The safety alarm includes an audible and visual alarm (implemented through the audible and visual alarm of the safety monitoring and emergency unit) and a pre-triggered shutdown command from the system's main control module. At the same time, the alarm information is uploaded to the host computer's interactive interface.

[0101] This invention integrates multi-directional deviations by quantifying translational deviations using Euclidean distance and rotational deviations using Euclidean norm. This ensures computational efficiency while comprehensively characterizing the degree of position and attitude deviation of the components. Furthermore, based on comparative analysis using preset thresholds, it clarifies whether the deviations are within the adjustable range, providing clear guidance for subsequent operations. This ensures the accuracy and reliability of the deviation analysis results and effectively avoids component damage or grasping failure caused by improper deviation judgment, thus guaranteeing the safety and stability of nuclear fuel assembly grasping operations.

[0102] Furthermore, the nuclear fuel assembly grasping, positioning and deviation correction system based on industrial vision provided by the present invention will be described below. The nuclear fuel assembly grasping, positioning and deviation correction system based on industrial vision described below can be referred to in correspondence with the nuclear fuel assembly grasping, positioning and deviation correction method based on industrial vision described above.

[0103] Optional, refer to Figure 2 , Figure 2 This is a schematic diagram of the structure of the nuclear fuel assembly grasping, positioning, and deviation correction system based on industrial vision provided by the present invention. The nuclear fuel assembly grasping, positioning, and deviation correction system based on industrial vision includes: Data acquisition module 210 is used to acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints; The contour extraction module 220 is used to extract the outer contour boundary of the target nuclear fuel assembly in three-dimensional space based on multi-view image data and structured light deformation image data, so as to obtain the initial outer contour boundary data. The pose localization module 230 is used to perform pose localization based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, so as to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space. The deviation analysis module 240 is used to perform deviation analysis based on the current pose parameters and the preset standard pose parameters to obtain the current deviation data and the corresponding deviation analysis results. The calibration output module 250 is used to generate a pose correction instruction based on the current deviation data if the deviation analysis result is a correctable deviation, and then transmit the pose correction instruction to the robot grasping system.

[0104] The embodiments of the present invention realize precise dynamic correction of deviation and automated control of grasping and positioning, improve the accuracy, stability and repeatability of grasping and positioning, and effectively solve the safety problems of poor correction effect and easy to cause component collision damage in the existing methods.

[0105] Please see Figure 3 , Figure 3 An embodiment diagram of an electronic device provided in accordance with the present invention. For example... Figure 3 As shown, the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, it implements steps 10-50.

[0106] Please see Figure 4 , Figure 4 An embodiment diagram of a computer-readable storage medium provided in accordance with an embodiment of the present invention is shown. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, it implements steps 10-50.

[0107] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the nuclear fuel assembly grasping, positioning and deviation correction method based on industrial vision provided by the above methods, which includes steps 10-50.

Claims

1. A method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision, characterized in that, include: Acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints; Based on the multi-view image data and the structured light deformation image data, the outer contour boundary of the target nuclear fuel assembly in three-dimensional space is extracted to obtain the initial outer contour boundary data; Based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, the pose is located to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space. Based on the current pose parameters and the preset standard pose parameters, a deviation analysis is performed to obtain the current deviation data and the corresponding deviation analysis results. If the deviation analysis result is a correctable deviation, then an instruction is generated based on the current deviation data to obtain a pose correction instruction, and the pose correction instruction is transmitted to the robot grasping system.

2. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 1, characterized in that, The extraction of the outer contour boundary of the target nuclear fuel assembly in three-dimensional space based on the multi-view image data and the structured light deformation image data yields initial outer contour boundary data, including: Pixel-level geometric consistency verification is performed based on the multi-view image data and the structured light deformation image data to obtain a set of multi-source pixels that pass the verification. Based on the structured light encoded phase information of each pixel in the multi-source pixel set and the corresponding position in the structured light deformation image under the corresponding view, the three-dimensional spatial point corresponding to the pixel is determined, and an initial three-dimensional point cloud dataset is constructed based on the three-dimensional spatial points of all pixels. Based on each point in the initial 3D point cloud dataset, a neighborhood curvature-guided selection of boundary candidate points is performed to obtain a set of boundary candidate points; Based on any two points in the set of candidate boundary points, perform directional consistency clustering to obtain a set of boundary segments with consistent orientation. Based on the set of boundary segments, the outer contour is refined to obtain the initial outer contour boundary data.

3. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 2, characterized in that, The process of refining the outer contour based on the set of boundary segments to obtain the initial outer contour boundary data includes: Coplanarity verification is performed on each boundary segment in the set of boundary segments to obtain a subset of coplanar boundary segments. Topological closure detection is performed on each segment in the subset of coplanar boundary segments to obtain a set of closed contour segments that are determined to be closed contours; Based on each closed contour in the set of closed contour segments, geometric prior matching of the target nuclear fuel assembly is performed to obtain an outer contour segment that conforms to the shape constraints of the assembly. Spatial alignment and fusion are performed based on the outer contour segments to obtain the initial outer contour boundary data of the target nuclear fuel assembly.

4. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 1, characterized in that, The pose localization based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space, includes: Based on the initial outer contour boundary data and the theoretical outer contour boundary data of the original design data in the target nuclear fuel assembly, geometric corresponding point pairs are extracted to obtain the initial corresponding point pair group; Based on all point pairs in the initial corresponding point pair group, rigid body transformation parameters are analyzed to obtain the initial pose transformation matrix; Based on the initial pose transformation matrix, all feature control points in the original design data are spatially mapped to obtain the mapped feature control point set. Based on the feature control point set and the initial outer contour boundary data, a local geometric structure matching is performed to obtain a verification point set with consistent structure. Based on the verification point set, the initial pose transformation matrix, the original design data, and the initial outer contour boundary data, pose transformation and verification are performed to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space.

5. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 4, characterized in that, The process of performing pose transformation and verification based on the verification point set, the initial pose transformation matrix, the original design data, and the initial outer contour boundary data to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space includes: Based on each verification point in the verification point set, a local coordinate system is constructed with the principal direction of the normal vector in its neighborhood, and an alignment constraint analysis is performed between the local coordinate system and the design points in the corresponding original design data to obtain the enhanced pose constraint conditions. Based on the enhanced pose constraints and the initial pose transformation matrix, iterative projection optimization is performed to obtain the optimized target pose transformation matrix. Based on the target pose transformation matrix, the complete outer contour in the original design data is globally reprojected to obtain the reprojected outer contour data. Based on the reprojected outer contour data and the initial outer contour boundary data, an integral analysis of the contour closure deviation is performed to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space.

6. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 1, characterized in that, The current deviation data includes translational deviation components and rotational deviation components; the deviation analysis based on the current pose parameters combined with preset standard pose parameters to obtain the current deviation data includes: Based on the current pose parameters and the standard pose parameters, a rigid body space alignment reference is constructed to obtain the reference alignment transformation matrix; Based on the reference alignment transformation matrix, the coordinate system of the positioning feature point set of the target nuclear fuel assembly is shifted to obtain the aligned actual feature point set; Based on the actual feature point set and the actual observed feature point set determined in the current pose parameters, a point-by-point deviation vector analysis is performed to obtain the original deviation vector set; Based on the original deviation vector set, the rigid body motion mode is decoupled to obtain the translational deviation component and the rotational deviation component.

7. The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision according to claim 6, characterized in that, The deviation analysis results include correctable and uncorrectable deviations; The steps to obtain the deviation analysis results include: Based on the translational deviation component and the rotational deviation component, three-dimensional vector amplitude quantization is performed to obtain the total position deviation and the total attitude deviation. Based on the total position deviation and total attitude deviation, a comparative analysis is performed using preset position deviation thresholds and attitude deviation thresholds. If the total position deviation is less than or equal to the position deviation threshold and the total attitude deviation is less than or equal to the attitude deviation threshold, the deviation analysis result is determined to be a correctable deviation. If the total position deviation is greater than the position deviation threshold or the total attitude deviation is greater than the attitude deviation threshold, the deviation analysis result is determined to be an uncorrectable deviation, and a safety alarm is triggered.

8. A nuclear fuel assembly grasping, positioning, and deviation correction system based on industrial vision, characterized in that, The method for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision, as described in any one of claims 1 to 7, is applied; the system for grasping, positioning, and correcting deviations of nuclear fuel assemblies based on industrial vision includes: The data acquisition module is used to acquire multi-view image data of the target nuclear fuel assembly and structured light deformation image data at the corresponding viewpoints; The contour extraction module is used to extract the outer contour boundary of the target nuclear fuel assembly in three-dimensional space based on the multi-view image data and the structured light deformation image data, so as to obtain the initial outer contour boundary data. The pose localization module is used to perform pose localization based on the initial outer contour boundary data and the original design data of the target nuclear fuel assembly, so as to obtain the current pose parameters of the target nuclear fuel assembly in three-dimensional space. The deviation analysis module is used to perform deviation analysis based on the current pose parameters and preset standard pose parameters to obtain the current deviation data and the corresponding deviation analysis results. The correction output module is used to generate a pose correction instruction based on the current deviation data if the deviation analysis result is a correctable deviation, and then transmit the pose correction instruction to the robot grasping system.

9. An electronic device, characterized in that, include: Memory, used to store computer software programs; A processor is configured to read and execute the computer software program, wherein when the processor executes the computer software program, it implements the nuclear fuel assembly grasping, positioning and deviation correction method based on industrial vision as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that, The storage medium stores a computer software program, which, when executed by a processor, implements the nuclear fuel assembly grasping, positioning, and deviation correction method based on industrial vision as described in any one of claims 1 to 7.

Images