A visual positioning system and method for a large array splicing mechanism

Abstract

This invention discloses a visual positioning system and method for a large array splicing mechanism, belonging to the field of intelligent manufacturing technology. The system of this invention is used to splice arrays with a docking frame according to a set pose. It includes an array splicing mechanism, a visual target, an image acquisition module, and a control module. The array is stably connected to the array splicing mechanism. The visual target is installed on the end face of the docking frame facing the array. The image acquisition module acquires image information from the visual target. The control module obtains the real-time relative pose of the array and the docking frame from the image information of the visual target, calculates the real-time pose deviation between the array and the docking frame based on the real-time relative pose and a set initial teaching reference pose, and sends instructions to the array splicing mechanism to adjust the relative pose of the array and the docking frame, so that the array is spliced ​​with the docking frame according to the set pose. This invention can automatically complete the splicing action, improving the efficiency and repeatability of array splicing.

Description

Technical Field

[0001] This invention belongs to the field of intelligent manufacturing technology, specifically relating to a visual positioning system and method for a large array splicing mechanism. Background Technology

[0002] In some large electronic devices, large arrays need to be spliced ​​together before they can be put into normal use. The splicing efficiency of large arrays has a great impact on the timeliness of equipment setup, and the splicing accuracy of arrays has an important impact on equipment performance. Therefore, positioning large arrays is a very important technical means when using splicing mechanisms.

[0003] Currently, the main method for locating large arrays in the market is manual inspection: Surveyors climb onto the docking frame and the array to be assembled, using levels and laser angle measuring instruments to measure the attitude deviations between the array and the docking frame in the pitch, roll, and yaw directions. They then use laser rangefinders, gauge blocks, and feeler gauges to measure the positional deviations between the array and the docking frame in the horizontal, vertical, and depth directions. Based on the measurement data, they calculate the attitude and positional adjustments required for the array to be assembled onto the docking frame. Installation personnel then adjust the positioning studs on the fixture to achieve this adjustment. This attitude measurement and adjustment process needs to be repeated until the measurement results meet the accuracy requirements for array assembly. This method is inefficient, requires on-site personnel, suffers from poor measurement consistency, and increases labor costs.

[0004] Current automatic array stitching and positioning technology mainly uses visual measurement positioning technology. When using it, the visual measurement system needs to be set up manually, resulting in low positioning and stitching efficiency and reducing the degree of automation. The visual measurement system needs to adjust the angle to ensure that the field of view covers the entire array, which places high demands on the technical skills of the personnel setting up the visual measurement system and increases the uncertainty of system operation. The target of the visual measurement system is installed on the array, and large arrays are prone to large elastic deformation due to the size and material, which affects the positioning accuracy and has poor repeatability. Summary of the Invention

[0005] The purpose of this invention is to provide a visual positioning system and method for a large array splicing mechanism, which can automatically complete the splicing action and improve the efficiency and repeatability of array splicing.

[0006] Specifically, on the one hand, the present invention provides a visual positioning system for a large array splicing mechanism, used to splice arrays with a docking frame according to a set pose, including an array splicing mechanism, a visual target, an image acquisition module, and a control module;

[0007] The array splicing mechanism includes a static platform that is stationary relative to the workspace and a moving platform that moves relative to the workspace. The array is stably connected to the moving platform. The array splicing mechanism receives instructions from the control module to adjust the relative position of the array and the docking frame.

[0008] The visual target is mounted on the end face of the docking frame facing the array surface;

[0009] An image acquisition module is installed on the array splicing mechanism to acquire image information of the visual target;

[0010] The control module receives image information of the visual target acquired by the image acquisition module, performs image processing, obtains the real-time relative pose of the array surface and the docking frame from the image information of the visual target, calculates the real-time pose deviation between the array surface and the docking frame based on the real-time relative pose, and sends a command to the array surface stitching mechanism based on the real-time pose deviation. The array surface stitching mechanism adjusts the relative pose of the array surface and the docking frame so that the real-time pose deviation is within a set allowable range, thereby enabling the array surface to be stitched with the docking frame according to the set pose.

[0011] Furthermore, the system includes multiple visual targets; the number of image acquisition modules is the same as the number of visual targets, and each image acquisition module acquires the image information of the corresponding visual target; each image acquisition module is equipped with a control module, the image information of each visual target is processed in the corresponding control module, the visual target data is exchanged between the control modules, and one of the control modules sends a command to the stitching mechanism based on the real-time pose deviation.

[0012] On the other hand, the present invention also provides a visual positioning method for a large array splicing mechanism, which is implemented using the aforementioned visual positioning system for a large array splicing mechanism, and includes the following steps:

[0013] The steps for calculating the real-time relative pose are as follows: The image acquisition module acquires image information of the visual target; the control module receives the image information of the visual target, performs image processing, and calculates the real-time relative pose between the array surface and the docking frame based on the image information of the visual target. ;

[0014] The steps for calculating the real-time pose deviation are as follows: based on the real-time relative pose... Calculate the real-time pose deviation between the array surface and the docking frame. ;

[0015] The steps for determining whether the real-time pose deviation is within the set allowable range are as follows: If the real-time pose deviation is within the set allowable range, the process ends; if the real-time pose deviation is not within the allowable range, a command is sent to the array splicing mechanism based on the real-time pose deviation. The array splicing mechanism adjusts the relative pose of the array and the docking frame according to the translational and rotational motion amounts in the command. The steps of calculating the real-time relative pose, calculating the real-time pose deviation, and determining whether the real-time pose deviation is within the set allowable range are repeated until the process ends.

[0016] Furthermore, the real-time relative pose of the array surface and the docking frame is calculated from the image information of the visual target. Specifically, it includes:

[0017] Obtain the coordinates of the center points of the first, second, third, and fourth feature images on the visual target, respectively. , , , All of these include the x, y, and z spatial information of the center point of the graphic in the static platform reference frame; the control module calculates the real-time relative pose of the array surface and the docking frame according to the following formula. ,

[0018] (4)

[0019] (5)

[0020] (6)

[0021] in, , represents the real-time pose change relative to the static platform coordinate system, and is a 3x3 matrix. express point to spatial vectors, Indicates to Modulo operation on vectors This represents the cross product of vectors. For the real-time pose relative to the static platform coordinate system, where ,Right now , where x, y, and z represent the translation amounts, respectively.

[0022] Furthermore, based on the real-time relative pose, the real-time pose deviation between the array surface and the docking frame is calculated. Specifically, it includes:

[0023] (7)

[0024] Among them, real-time pose deviation Represented in the form of a homogeneous transformation matrix, it is derived from the real-time relative pose. Compared with the reference pose The inverse matrix is ​​obtained by multiplying the inverse matrix.

[0025] Furthermore, the calculation methods for the translational and rotational motion are as follows:

[0026] (8)

[0027] (9)

[0028] (10)

[0029] in, For real-time pose deviation The rotational motion relative to the reference coordinate system includes ( ), Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. For real-time pose deviation The translational motion relative to the reference coordinate system includes ( ), Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis.

[0030] Furthermore, the step of calculating the real-time relative pose also includes an initial teaching step, specifically:

[0031] Step 2.1: Use instruments to detect the pose deviation of the array surface relative to the docking frame. The pose deviation should include translational deviations (x, y, z) and rotational deviations (Rx, Ry, Rz) required for the array surface to reach the completed assembly state. The translational deviations (x, y, z) are the lengths of the projections of the lines connecting the current position and the completed assembly position on the mounting frame, measured in the x, y, and z directions with the static platform coordinate system as the reference direction. The rotational deviation is the deviation obtained by subtracting the current rotational values ​​(Rx, Ry, Rz) of the array surface in the Rx, Ry, and Rz directions from the corresponding values ​​at the completed assembly position on the mounting frame, measured in the static platform coordinate system as the reference direction.

[0032] Step 2.2: Determine the pose deviation. If the pose deviation is outside the set allowable range, it is determined that the splicing of the array surface and the docking frame is incomplete. The control module outputs the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the pose deviation to the splicing mechanism moving platform. The splicing mechanism moving platform moves according to the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the pose deviation to adjust the attitude of the array surface. Repeat steps 2.1 and 2.2. If the pose deviation is within the set allowable range, it is determined that the splicing of the array surface and the docking frame is complete, and proceed to step 2.3.

[0033] Step 2.3: Record the pose deviation between the array surface and the docking frame at this time as the initial teaching reference pose to complete the initial teaching. In the step of calculating the real-time relative pose, the initial teaching reference pose is used as the initial value of the real-time relative pose.

[0034] Furthermore, when the stated , , , When a single feature point is missing, use Substitute into equation (5) Calculation, use Substitute into equation (5) Calculation; Use or Substitute into equation (6) calculate.

[0035] Furthermore, when the center points of all feature graphics have been identified, use and Substituting the mean into equation (5) Calculation, use and Substituting the mean into equation (5) calculate.

[0036] Furthermore, the visual positioning method for the large array splicing mechanism also includes a combined preprocessing method during the image acquisition module's acquisition of image information of the visual target. The combined preprocessing steps are as follows:

[0037] S102, Gaussian blur processing is performed on the original image to filter high-frequency information in the original image;

[0038] S103 utilizes the gradient information of the grayscale image to separate the parts with large grayscale gradients and detect local feature edges.

[0039] S104, perform morphological transformation;

[0040] S105 performs contour detection to obtain the contours of feature points.

[0041] The beneficial effects of the large array splicing mechanism visual positioning system and method of the present invention are as follows:

[0042] By placing the visual targets on a high-rigidity docking frame, the elastic deformation of the docking frame is much smaller than that of the array surface, avoiding the decrease in recognition accuracy that may be caused by using visual targets on the array surface, while also reducing the material and structural costs of the array surface; by using small visual targets and image acquisition modules (such as wide-angle camera groups) installed on the end face of the docking frame, it is ensured that the targets can always appear within the field of view of the image acquisition module, reducing the requirements for the relative position of the targets and the image acquisition module; by fixing the image acquisition module to the stitching mechanism, the separate installation step of the visual positioning system is eliminated, reducing the requirements for manual operation.

[0043] By combining an image acquisition module and a visual target group, visual positioning can be achieved by recognizing only a portion of the targets, thus improving the fault tolerance and anti-interference of the visual positioning system; the recognition results of multiple targets can be mutually verified, thereby improving the recognition accuracy.

[0044] Visual positioning methods can acquire array pose in real time, which can obtain the position and orientation of the array more quickly than manual measurement, thus improving the efficiency of array splicing.

[0045] To improve the reliability of visual target image recognition and reduce the impact of complex lighting conditions and imaging environments in engineering scenarios on image recognition, this invention enhances robustness from both hardware and software dimensions. On the hardware side, two or more supplementary lighting sources are respectively set on the upper and lower sides of the visual target to ensure uniform illumination intensity at the edge and center of the target. On the software side, the visual acquisition module is configured to acquire the visual target image using a set fixed exposure time, ensuring that under single supplementary lighting conditions, the white portion of the visual target reaches a first threshold (e.g., 90%) of the maximum exposure value, and the black portion reaches a second threshold (e.g., 30%) of the maximum exposure value. The hardware design of the supplementary lighting source and the fixed exposure design ensure accurate exposure of the visual target image. The high-power supplementary lighting source and uniform illumination design can significantly reduce the impact of ambient light on target pattern capture. In the process of visual target group image recognition, the large-format stitching visual positioning method of this invention adopts a combined preprocessing method to reduce the impact of ambient light on recognition. It can still obtain the contour of feature points with a high success rate under the influence of ambient light, and finally obtain the coordinates of all or part of the graphic center point. Compared with the scheme using self-illuminating targets, the supplementary lighting source scheme used in this invention can effectively improve the illumination uniformity of the target, while avoiding glare and ghosting caused by self-illuminating targets on the lens, reducing the impact on target imaging quality.

[0046] The large-scale array splicing mechanism visual positioning system and method of the present invention uses a camera group installed on the array splicing mechanism to capture visual target groups on the docking frame, and uses computer vision recognition technology to obtain the pose deviation between the array and the docking frame. This guides the splicing mechanism to adjust the relative pose of the array and the docking frame in real time, automatically completing the splicing action. No dedicated personnel are required to operate the equipment for array splicing, reducing labor costs and eliminating the dangers of manual operation. The positioning work is considered complete only after the array splicing accuracy reaches a set allowable range, improving the repeatability and positioning accuracy of array splicing. Compared to solutions for ideal lighting conditions, the large-scale array splicing mechanism visual positioning system and method of the present invention, in terms of both hardware and software, addresses the complex lighting conditions and imaging environment of engineering scenarios. Through target light source, image acquisition module imaging parameter configuration, and combined preprocessing procedures, it reduces environmental interference on imaging and feature image contour recognition. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the system in Embodiment 1 of the present invention.

[0048] Figure 2 This is a diagram showing the physical and data relationships of visual positioning in Embodiment 1 of the present invention.

[0049] Figure 3 This is a flowchart of the visual positioning process of the present invention.

[0050] Figure 4 This is a flowchart of the visual initial teaching process of the present invention.

[0051] Figure 5 This is a schematic diagram of the target pattern of the present invention.

[0052] Figure 6 This is a schematic diagram of the target illumination of the present invention.

[0053] Figure 7 This is a flowchart of the visual target group image recognition process of the present invention.

[0054] Figure 8 This is a schematic diagram of the intermediate process of target image recognition preprocessing under the influence of ambient light in this invention. (1) is the original image before combined preprocessing, (2) is the result after Gaussian blur processing of (1), (3) is the result after edge detection of (2), (4) is the result after morphological transformation of (3), and (5) is the result after contour detection of (4).

[0055] Figure 9 This is a system schematic diagram of Embodiment 2 of the present invention.

[0056] Figure 10This is a diagram showing the physical and data relationships of visual positioning in Embodiment 2 of the present invention.

[0057] Reference numerals: 1-Array splicing mechanism, 101-Moving platform of splicing mechanism, 102-Static platform of splicing mechanism, 2-Docking frame, 3-Visual target, 301-First visual target, 302-Second visual target, 4-Array, 5-Image acquisition module, 501-First image acquisition module, 502-Second image acquisition module, 6-Processing module, 7-First feature image, 8-Second feature image, 9-Third feature image, 10-Fourth feature image, 11-Upper feature point, 12-Lower feature point. Detailed Implementation

[0058] The present invention will now be described in further detail with reference to the embodiments and the accompanying drawings.

[0059] Example 1:

[0060] One embodiment of the present invention is a visual positioning system for a large array splicing mechanism, used to splice arrays with a docking frame according to a set pose. The array is the object of the array splicing; the docking frame is the target of the array splicing, a mechanical structural component used for splicing with the array. The hardware composition of one embodiment of the present invention is as follows: Figure 1 As shown. Mainly includes:

[0061] The array splicing mechanism 1, used to adjust the relative pose of the array 4 and the docking frame 2, includes a moving platform 101 and a stationary platform 102. The stationary platform 102 is stationary relative to the workspace; the moving platform 101 moves relative to the workspace. The array 4 is securely connected to the moving platform 101. The array splicing mechanism 1 receives instructions from the control module 6 to adjust the relative pose of the array 4 and the docking frame 2.

[0062] A visual target 3 is installed on the end face of the docking frame 2 facing the array surface 4. The image acquisition module 5 acquires image information from the visual target 3 to identify the spatial position of the docking frame 2. The docking frame 2 adopts a high-rigidity design relative to the array surface 4, which reduces elastic deformation, improves measurement accuracy, and also reduces the rigidity requirements of the array surface 4, saving material costs. At the same time, only one set of visual targets needs to be installed on the docking frame 2. Compared with the scheme of installing visual targets on each array surface, this ensures the consistency of the visual target position when multiple array surfaces are spliced, and also eliminates the step of unifying the relative position of the target and the array surface on different array surfaces.

[0063] An image acquisition module 5, such as a wide-angle camera, is mounted on the array stitching mechanism 1. Preferably, the image acquisition module 5 is mounted on the moving platform 101 of the stitching mechanism near the docking frame 2. The image acquisition module 5 is used to acquire image information of the visual target 3. The length direction of the array 4 is parallel to the image acquisition module 5, and the width direction is parallel to the optical axis of the image acquisition module 5.

[0064] The control module 6 receives image information of the visual target 3 acquired by the image acquisition module 5, performs image processing, and obtains the real-time relative pose of the array surface and the docking frame from the image information of the visual target using the large array surface splicing mechanism visual positioning method of this embodiment. Based on the real-time relative pose and the set initial teaching reference pose, the real-time pose deviation between the array surface 4 and the docking frame 2 is calculated. Based on the real-time pose deviation, the control module 6 sends a command to the array surface splicing mechanism 1. The array surface splicing mechanism 1 adjusts the relative pose of the array surface 4 and the docking frame 2 so that the real-time pose deviation is within the set allowable range, thereby enabling the array surface to be spliced ​​with the docking frame according to the set pose.

[0065] The control module 6 automatically obtains the pose deviation between the array surface 4 and the docking frame 2, performs visual positioning for large array surface splicing, and guides the splicing mechanism (the moving platform 101 and the stationary platform 102 of the splicing mechanism) to complete the splicing action. This eliminates the need for dedicated personnel to perform the array surface splicing action, reducing labor costs. Compared to existing technologies that use independent mechanical structures to separately set up the visual measurement system, the large array surface splicing mechanism visual positioning system of this invention integrates the image acquisition module with the splicing mechanism, allowing it to move synchronously with the splicing mechanism, reducing setup steps and improving setup efficiency.

[0066] The relationship between the physical and data layers of the large array stitching visual positioning system in this embodiment is as follows: Figure 2 As shown.

[0067] Physical level:

[0068] The array 4 is mechanically fixed to the splicing mechanism moving platform 101; the splicing mechanism moving platform 101 and the splicing mechanism stationary platform 102 can move controllably to adjust the position and attitude of the splicing mechanism moving platform 101; the splicing mechanism stationary platform 102 is relatively stationary with respect to the workspace; the image acquisition module 5 is mechanically fixed to the splicing mechanism moving platform 101; the visual target 3 is mechanically fixed to the docking frame 2.

[0069] Data level:

[0070] The visual target 3 is imaged in the image acquisition module 5; the image acquisition module 5 transmits the array pose information to the stitching mechanism moving platform 101 to guide the movement of the stitching mechanism moving platform 101 and realize the visual positioning of the array stitching.

[0071] The process of the large array splicing visual positioning method is as follows: Figure 3 As shown.

[0072] (Optional) Steps to determine whether the system has undergone initial teaching:

[0073] Control module 6 checks the system's operating status and determines whether the system has undergone initial teaching. If no initial teaching has been performed, it proceeds to the initial teaching step; if an initial teaching has been performed, it proceeds to the step of calculating the real-time relative pose.

[0074] The determination of whether the system has undergone initial teaching is accomplished by checking the system's operational status. If the system is in positioning mode and the reference pose data for initial teaching is not empty, then the system is determined to have undergone initial teaching. The reference pose data for initial teaching consists of x, y, z, Rx, Ry, and Rz, which represent the amount of translation (in millimeters), translation (in millimeters), translation (in millimeters), rotation (in radians), rotation (in radians), and rotation (in radians) of the target pose relative to the current pose in the static platform reference frame, respectively. The static platform reference frame is defined as follows: the origin is the geometric center of the upper surface of the static platform 102 of the stitching mechanism; the x-axis is parallel to the initial width direction of the array surface 4, with its positive direction pointing from the origin to the mounting side of the image acquisition module 5; the y-axis is parallel to the initial length direction of the array surface 4, with its positive direction pointing from the right side of the target being captured by the image acquisition module to the left side; the z-axis is parallel to the initial height direction of the array surface 4, with its direction perpendicular to the array surface pointing away from the static platform 102 of the stitching mechanism. The x, y, and z axes conform to the Cartesian coordinate system. The initial teaching reference pose data can be represented by an equivalent homogeneous transformation matrix (...). The symbol () is used to describe the position and attitude of a reference in the static platform coordinate system. The attitude changes of the reference in the static platform coordinate system are represented by... This indicates that the position change of the datum in the static platform coordinate system is represented by... This indicates the reference pose data used in the initial teaching demonstration. and , The corresponding conversion relationship is as follows:

[0075] (1)

[0076] (2)

[0077] (3)

[0078] in, Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis.

[0079] (Optional) Steps for initial teaching:

[0080] The baseline pose for initial teaching It can be obtained through manual input into the visual positioning system, or by performing the initial teaching steps.

[0081] Preferably, in another embodiment, an initial teaching step is included before the step of calculating the real-time relative pose, such as... Figure 4 As shown, the initial teaching includes the following steps:

[0082] Step 2.1: Use instruments to detect the pose deviation of the array surface 4 relative to the docking frame 2.

[0083] The pose deviations should include the translational deviations (x, y, z) and rotational deviations (Rx, Ry, Rz) required for array 4 to reach the completed splicing state. Translational deviations are achieved using a measuring tape, gauge blocks, and vernier calipers. With the static platform coordinate system as the reference direction, the lengths of the lines connecting the current position and the completed splicing position at the mounting frame are measured in the x, y, and z directions, respectively. Rotational deviations are achieved using a level and gyroscope. With the static platform coordinate system as the reference direction, the current rotational values ​​(Rx, Ry, Rz) of array 4 are measured in the Rx, Ry, and Rz directions, and the corresponding values ​​are measured at the completed splicing position at the mounting frame. The deviations are then subtracted from each other to obtain the total deviation.

[0084] Step 2.2: Determine the pose deviation. If the pose deviation is outside the allowable range, it is determined that the splicing of the array surface 4 and the docking frame 2 is not completed, and steps 2.1 and 2.2 are repeated. If the pose deviation is within the allowable range, it is determined that the splicing of the array surface 4 and the docking frame 2 is completed, and proceed to step 2.3.

[0085] The allowable range is given according to the physical structure of the array surface and the docking frame, that is, it does not exceed the tolerance range of the mechanical structure at the docking point between the array surface 4 and the docking frame 2, so as to meet the splicing accuracy requirements. During the attitude adjustment process of the array surface 4, the control module 6 outputs the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the attitude deviation to the splicing mechanism moving platform 101. The splicing mechanism moving platform 101 moves according to the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the attitude deviation to adjust the attitude of the array surface 4.

[0086] Step 2.3: Record the pose deviation between the array surface 4 and the docking frame 2 at this time, and use it as the reference pose for the initial teaching to complete the initial teaching.

[0087] The initial teaching reference pose data should include the relative translation and rotation information of the array surface 4 and the docking frame 2 in the completed array assembly state. The initial teaching reference pose data is stored in the control module 6, and is used as the initial value of the real-time relative pose for subsequent acquisition of the real-time relative pose of the array surface 4 and the docking frame 2. (See the steps for calculating real-time relative pose) After that, calculate the real-time pose deviation (see the steps for calculating real-time pose deviation).

[0088] The steps to calculate real-time relative pose are:

[0089] Image acquisition module 5 acquires image information of visual target 3, and control module 6 receives the image information of visual target, performs image processing, and calculates the real-time relative pose of array surface 4 and docking frame 2 based on the image information of visual target. .

[0090] The image acquisition module 5 captures images of the visual target 3, and obtains the real-time relative pose of the array surface 4 and the docking frame 2 from the visual target images. The real-time relative pose obtained This should include the translation and rotation information of the array surface 4 relative to the docking frame 2. The real-time relative pose of the array surface 4 and the docking frame 2 is calculated from the image information of the visual target. The method is as follows.

[0091] like Figure 5 As shown, the coordinates of the center points of the first, second, third, and fourth feature graphics on the visual target are obtained through image recognition technology (for example, the coordinates of the center points of feature graphics 9, 10, 11, and 12 on visual target 3). , , , Each of these includes the x, y, and z spatial information of the center point of the graphic in the static platform reference frame. After the control module 6 calculates the coordinate data of the center points of the three feature graphics of the target, it calculates the real-time relative pose of the array surface 4 and the docking frame 2. .

[0092] The coordinates of the feature points obtained by image acquisition module 5 are now used. (Right now Taking (e.g.) as an example, this illustrates how to calculate the real-time relative pose between the array surface 4 and the docking frame 2 based on feature point coordinate data. The specific calculation formulas are Equations (4), (5) and (6).

[0093] (4)

[0094] (5)

[0095] (6)

[0096] in, , represents the real-time pose change relative to the static platform coordinate system, and is a 3x3 matrix. express point to spatial vectors, Indicates to Modulo operation on vectors This represents the cross product of vectors. For the real-time pose relative to the static platform coordinate system, where ,Right now , where x, y, and z represent the translation amounts, respectively.

[0097] It is understandable that the coordinate data of any feature point can be used as... When using When substituting into the above formula, the right side The subscript i should correspond to the subscript q, indicating the displacement data of the corresponding feature point.

[0098] Preferably, in another embodiment, when , , , When a single feature point is missing, use Substitute into equation (5) Calculation, use Substitute into equation (5) Calculation; Use or Substitute into equation (6) Computation can improve the reliability when a single feature point is missing.

[0099] Preferably, in another embodiment, when the center points of all feature patterns are identified, the large-scale array stitching visual positioning method of the present invention utilizes the redundancy of the number of feature points to improve calculation accuracy. and Substituting the mean into equation (5) Calculation, use and Substituting the mean into equation (5) The calculation reduced random errors.

[0100] The large-scale array stitching visual positioning method of this invention differs from existing solutions in that: the vectors selected for real-time relative pose calculation fully utilize the redundancy of the feature graphics on the visual target. When some feature graphics are occluded, equivalent vectors are used as substitutes, improving the robustness of pose recognition. Furthermore, in the case of full recognition, the redundancy is utilized to substitute the mean of the equivalent vectors, improving the accuracy of pose recognition and thus enhancing the reliability of the algorithm.

[0101] Large-scale image array splicing is used in practical engineering scenarios, typically outdoors, where lighting conditions are complex, which can severely impact image recognition. To improve the reliability of image recognition and reduce the influence of lighting conditions, preferably, in another embodiment, this invention enhances robustness from both hardware and software dimensions. On the hardware side, such as... Figure 6 As shown, two or more supplementary light sources are set on the upper and lower sides of the visual target. These supplementary light sources are linear, flicker-free, and diffused light sources with a total power of no less than 100 W. Hardware such as softboxes is used to ensure uniform light output. The light source's output direction is towards the center of the visual target, and its distance from the target is more than 50% of the longest side of the target to ensure uniform illumination intensity at the edge and center of the target. On the software side, the image acquisition module is set to acquire the visual target image using a fixed exposure time. Preferably, the fixed exposure time is selected to ensure that, under single supplementary light source illumination, the white portion of the visual target reaches more than 90% of its maximum exposure value, and the black portion reaches less than 30% of its maximum exposure value. The supplementary light source hardware design and fixed exposure design ensure accurate exposure of the visual target image, while the high-power supplementary light source and uniform illumination design significantly reduce the impact of ambient light on the capture of the visual target pattern.

[0102] Preferably, in another embodiment, in the step of acquiring real-time relative pose by capturing images of corresponding visual targets using the image acquisition module, the large-scale array stitching visual positioning method of the present invention further includes a combined preprocessing method during the visual target group image recognition process to reduce the influence of ambient light on recognition, such as... Figure 7 As shown.

[0103] Before combined preprocessing, the visual target image is unevenly illuminated due to the influence of ambient light, and some parts of the original image are significantly brighter, such as... Figure 8 As shown in (1). Intermediate results of combined pretreatment under the influence of ambient light can be referred to Figure 8 The image content is shown in the image. The preprocessing steps in the visual target group image recognition process are as follows:

[0104] S102, for the original image, such as the visual target image 8(1) affected by ambient lighting, Gaussian blur processing is performed, and the result is as follows: Figure 8(2) As shown. Gaussian blurring can filter high-frequency information in the original image, effectively reducing visual target image interference caused by noise, minor defects of the target, wind, sand, rain and snow.

[0105] S103 performs edge detection, and the result is as follows: Figure 8 As shown in (3), by utilizing the gradient information of the grayscale image, the parts with large grayscale gradients are separated, and the local feature edges can still be detected under uneven lighting conditions.

[0106] S104 underwent morphological transformation, and the result is as follows: Figure 8 As shown in (4), when the visual target image is not ideal, the edges of interrupted feature points can be connected to improve the recognition success rate.

[0107] S105, perform contour detection, the result is as follows: Figure 8 As shown in (5), the contour of the feature points is obtained.

[0108] After the above combined preprocessing steps, the contours of feature points can still be obtained with a high success rate even under conditions affected by ambient light. Figure 7 The remaining steps all utilize existing mature technologies, ultimately yielding the coordinates of all or part of the graphic's center point.

[0109] Compared to the solution using a self-luminous target, the supplementary light source solution used in this invention can effectively improve the illumination uniformity of the target, while avoiding glare and ghosting caused by the self-luminous target on the lens, thus reducing the impact on the target imaging quality.

[0110] Compared to solutions designed for ideal lighting conditions, this invention addresses the complex lighting conditions and imaging environment of engineering scenarios through a combination of target light source, image acquisition module imaging parameter configuration, and preprocessing procedures. This reduces the interference of the environment on imaging and feature image contour recognition.

[0111] The steps to calculate real-time pose deviation are as follows:

[0112] Based on the real-time relative pose, calculate the real-time pose deviation between the array surface and the docking frame. .

[0113] Real-time pose deviation should include information on the translational and rotational movements required for the array to reach the stitching completion state. This invention proposes a real-time pose deviation... The calculation method is as follows:

[0114] (7)

[0115] Among them, real-time pose deviation Represented in the form of a homogeneous transformation matrix, it is derived from the real-time relative pose. Compared with the reference pose The inverse matrix is ​​obtained by multiplying the inverse matrix.

[0116] Steps to determine whether the real-time pose deviation is within the set allowable range:

[0117] Determine real-time pose deviation Is it within the allowable range? If there is a real-time pose deviation? If the position is within acceptable limits, the positioning process is complete; if the real-time pose deviation is within acceptable limits... If the position deviation is outside the allowable range, the real-time pose deviation is sent as a guiding command to the stitching mechanism. The stitching mechanism then performs stitching according to the translational and rotational motion amounts given by the pose deviation, repeating the steps of obtaining the real-time relative pose, calculating the real-time pose deviation, and determining whether the real-time pose deviation is within the set allowable range, until the process ends. The calculation methods for the translational and rotational motion amounts are as follows:

[0118] (8)

[0119] (9)

[0120] (10)

[0121] in, For real-time pose deviation The rotational motion relative to the reference coordinate system includes ( ), Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. For real-time pose deviation The translational motion relative to the reference coordinate system includes ( ), Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis. The allowable range is determined based on the physical structure of the array and the docking frame, meaning it does not exceed the tolerance range of the mechanical structure at the docking point between the array and the docking frame, in order to meet the splicing accuracy requirements, and is the same as the allowable range used in the initial teaching.

[0122] Example 2:

[0123] Another embodiment of the present invention is a visual positioning system for a large array splicing mechanism, the hardware components of which are as follows: Figure 9As shown. The main difference from Embodiment 1 is that this embodiment includes two image acquisition modules and two visual targets, specifically including:

[0124] The array splicing mechanism 1 is used to adjust the relative pose of the array and the docking frame, and includes a moving platform 101 and a stationary platform 102. The stationary platform 102 is stationary relative to the workspace; the moving platform 101 moves relative to the workspace.

[0125] The docking frame 2 is the target of array splicing and is a mechanical structural component used to splice with the array.

[0126] Visual targets 3, including a first visual target 301 and a second visual target 302, are respectively installed on the end faces of the docking frame 2 facing the array surface 4. The visual targets are used to identify the spatial position and geometric features of the docking frame 2. The relatively large distance between the two visual targets ensures that they accurately reflect the spatial position of different parts of the end face of the docking frame 2, reducing the impact of differences in the flatness of the end face of the docking frame 2.

[0127] Array surface 4 is securely connected to the splicing mechanism moving platform 101. The array surface is the object to be spliced, used to splice with the docking frame 2 according to a certain positioning and orientation requirements.

[0128] The image acquisition module 5 includes a first image acquisition module 501 and a second image acquisition module 502, respectively mounted on the array splicing mechanism 1. Preferably, the first image acquisition module 501 and the second image acquisition module 502 are mounted on the moving platform 101 of the splicing mechanism near the docking frame 2, with the distance between the two image acquisition modules matching the distance between the two visual targets. The image acquisition module 5 is used to capture image information of the visual targets. The length direction of the array 4 is parallel to the arrangement direction of the first image acquisition module 501 and the second image acquisition module 502, and the width direction is parallel to the optical axis of the first image acquisition module 501 and the second image acquisition module 502.

[0129] When the large array splicing mechanism visual positioning system uses multiple visual targets 3, the number of image acquisition modules 5 is the same as the number of visual targets. Each image acquisition module acquires the image information of the corresponding visual target, and the spacing between the image acquisition modules is consistent with the spacing between the visual targets.

[0130] The control module 6 receives the visual target image captured by the image acquisition module 5 and performs image processing and analysis. Using the large-scale array stitching mechanism visual positioning method of this embodiment, it automatically obtains the pose deviation between the array 4 and the docking frame 2, performs large-scale array stitching visual positioning, and guides the stitching mechanism (stitching mechanism moving platform 101 and stitching mechanism static platform 102) to perform the stitching action. This eliminates the need for dedicated personnel to complete the array stitching action, reducing labor costs. Both the first image acquisition module 501 and the second image acquisition module 502 are equipped with the control module 6. The image information processing of the visual target is performed within the control module 6 of the image acquisition module, which reduces the communication pressure between image acquisition modules and the computational pressure of centralized processing by a single image acquisition module. The two image acquisition modules are interchangeable, which is part of the maintainability design; in case of failure, spare parts can be quickly used for replacement. Compared to the existing technology that uses independent mechanical structures to separately set up the visual measurement system, the large-scale array stitching mechanism visual positioning system of this invention integrates the image acquisition module with the stitching mechanism, allowing it to move synchronously with the stitching mechanism, reducing setup steps and improving setup efficiency.

[0131] The relationship between the physical and data layers of the large array stitching visual positioning system in this embodiment is as follows: Figure 10 As shown.

[0132] Physical level:

[0133] The array 4 is mechanically fixed to the splicing mechanism moving platform 101; the splicing mechanism moving platform 101 and the splicing mechanism stationary platform 102 can move controllably to adjust the position and attitude of the splicing mechanism moving platform 101; the splicing mechanism stationary platform 102 is relatively stationary with respect to the workspace; the first image acquisition module 501 and the second image acquisition module 502 are mechanically fixed to the splicing mechanism moving platform 101 respectively; the first visual target 301 and the second visual target 302 are mechanically fixed to the docking frame 2 respectively.

[0134] Data level:

[0135] The first visual target 301 is imaged in the first image acquisition module 501; the second visual target 302 is imaged in the second image acquisition module 502; the first image acquisition module 501 and the second image acquisition module 502 exchange visual target recognition data (e.g., the second image acquisition module 502 transmits the coordinates of the center point of the target feature graphic obtained by the second image acquisition module 502 to the first image acquisition module 501); the first image acquisition module 501 transmits the real-time pose deviation to the splicing mechanism moving platform 101 to guide the movement of the splicing mechanism moving platform 101 and realize the visual positioning of the splicing array.

[0136] The process of the large array splicing visual positioning method is as follows: Figure 3 As shown.

[0137] (Optional) Steps to determine whether the system has undergone initial teaching:

[0138] Control module 6 checks the system's operating status and determines whether the system has undergone initial teaching. If no initial teaching has been performed, it proceeds to the initial teaching step; if an initial teaching has been performed, it proceeds to the step of calculating the real-time relative pose.

[0139] The determination of whether the system has undergone initial teaching is accomplished by checking the system's operational status. If the system is in positioning mode and the reference pose data in the system is not empty, then the system is determined to have undergone initial teaching. The reference pose data are x, y, z, Rx, Ry, and Rz, which represent the amount of translation (in millimeters) in the x-direction, translation (in millimeters) in the y-direction, translation (in millimeters) in the z-direction, rotation (in radians) in the x-direction, rotation (in radians) in the y-direction, and rotation (in radians) in the z-direction of the target pose relative to the current pose in the static platform reference frame, respectively. The static platform reference frame is defined as follows: the origin is the geometric center of the upper surface of the static platform 102 of the stitching mechanism; the x-axis is parallel to the initial width direction of the array surface 4, with its positive direction pointing from the origin to the mounting side of the image acquisition module 5; the y-axis is parallel to the initial length direction of the array surface 4, with its positive direction consistent with the direction from the second image acquisition module 502 to the first image acquisition module 501; the z-axis is parallel to the initial height direction of the array surface 4, with its direction perpendicular to the array surface and pointing away from the static platform 102 of the stitching mechanism. The x, y, and z axes conform to the Cartesian coordinate arrangement rules. The reference pose data for the initial teaching can be represented by an equivalent homogeneous transformation matrix (…). The symbol () is used to describe the position and attitude of a reference in the static platform coordinate system. The attitude changes of the reference in the static platform coordinate system are represented by... This indicates that the position change of the datum in the static platform coordinate system is represented by... This indicates that the corresponding conversion relationship is:

[0140] (1)

[0141] (2)

[0142] (3)

[0143] in, Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis.

[0144] (Optional) Steps for initial teaching:

[0145] The baseline pose for initial teaching It can be obtained through manual input into the visual positioning system, or by performing the initial teaching steps.

[0146] Preferably, in another embodiment, an initial teaching step is included before the step of calculating the real-time relative pose, such as... Figure 4 As shown, the initial teaching includes the following steps:

[0147] Step 2.1: Use instruments to detect the pose deviation of the array surface 4 relative to the docking frame 2.

[0148] The pose deviations should include the translational deviations (x, y, z) and rotational deviations (Rx, Ry, Rz) required for array 4 to reach the completed splicing state. Translational deviations are achieved using a measuring tape, gauge blocks, and vernier calipers. With the static platform coordinate system as the reference direction, the lengths of the lines connecting the current position and the completed splicing position at the mounting frame are measured in the x, y, and z directions, respectively. Rotational deviations are achieved using a level and gyroscope. With the static platform coordinate system as the reference direction, the current rotational values ​​(Rx, Ry, Rz) of array 4 are measured in the Rx, Ry, and Rz directions, and the corresponding values ​​are measured at the completed splicing position at the mounting frame. The deviations are then subtracted from each other to obtain the total deviation.

[0149] Step 2.2: Determine the pose deviation. If the pose deviation is outside the allowable range, it is determined that the splicing of the array surface 4 and the docking frame 2 is not completed. Adjust the attitude of the array surface 4 through the splicing mechanism 1 and repeat steps 2.1 and 2.2. If the pose deviation is within the allowable range, it is determined that the splicing of the array surface 4 and the docking frame 2 is completed, and proceed to step 2.3.

[0150] The allowable range is given according to the physical structure of the array surface and the docking frame, that is, it does not exceed the tolerance range of the mechanical structure at the docking point between the array surface 4 and the docking frame 2, so as to meet the splicing accuracy requirements. During the attitude adjustment process of the array surface 4, the control module 6 outputs the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the attitude deviation to the splicing mechanism moving platform 101. The splicing mechanism moving platform 101 moves according to the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the attitude deviation to adjust the attitude of the array surface 4.

[0151] Step 2.3: Record the pose deviation between the array surface 4 and the docking frame 2 at this time, and use it as the reference pose for the initial teaching to complete the initial teaching.

[0152] The initial teaching reference pose data should include the relative translation and rotation information of the array surface 4 and the docking frame 2 in the completed array assembly state. This information is stored in the control module 6. The initial teaching reference pose is used as the initial value of the real-time relative pose for subsequent acquisition of the real-time relative pose of the array surface 4 and the docking frame 2. (See the steps for calculating real-time relative pose) After that, calculate the real-time pose deviation (see the steps for calculating real-time pose deviation).

[0153] The steps to calculate real-time relative pose are:

[0154] Image acquisition module 5 acquires image information of visual target 3, and control module 6 receives the image information of visual target, performs image processing, and calculates the real-time relative pose of array surface 4 and docking frame 2 based on the image information of visual target. .

[0155] Specifically, the first image acquisition module 501 captures the image of the first visual target 301, and the second image acquisition module 502 captures the image of the second visual target 302. The real-time relative pose of the array surface 4 and the docking frame 2 is obtained from the visual target images. The real-time relative pose of array surface 4 and docking frame 2 was obtained. This should include translation and rotation information of the array surface 4 relative to the docking frame 2. The real-time relative pose of the array surface 4 and the docking frame 2 is obtained from the image information of the visual target. The method is as follows.

[0156] like Figure 5 As shown, image recognition technology is used to obtain the coordinates of the center points of feature graphics 9, 10, 11, and 12 on visual targets 301 and 302, respectively, which are p1, p2, p3, and p4. Each coordinate includes the x, y, and z spatial information of the center point in the static platform reference frame. The midpoint of p1 and p3 is used to obtain the coordinates of the upper feature point 11, and the midpoint of p2 and p4 is used to obtain the coordinates of the lower feature point 12. The coordinates of the upper and lower feature points obtained by the processing module on the first image acquisition module 501 are defined as follows: , The coordinates of the upper and lower feature points obtained by the processing module of the second image acquisition module 502 are respectively , After the control module 6 of the second image acquisition module 502 calculates the coordinate data of the upper and lower feature points of the second target 3, it transmits them to the processing module of the first image acquisition module 501 to calculate the real-time relative pose of the array surface 4 and the docking frame 2. .

[0157] The coordinates of the lower feature point are now obtained by the second image acquisition module 502. (Right now Taking (e.g.) as an example, this illustrates how to calculate the real-time relative pose between the array surface 4 and the docking frame 2 based on feature point coordinate data. The specific calculation formulas are Equations (4), (5) and (6).

[0158] (4)

[0159] (5)

[0160] (6)

[0161] in, , represents the real-time pose change relative to the static platform coordinate system, and is a 3x3 matrix. express point to spatial vectors, Indicates to Modulo operation on vectors This represents the cross product of vectors. For the real-time pose relative to the static platform coordinate system, where ,Right now , respectively, represent translations of x, y, and z.

[0162] It is understandable that the coordinate data of any feature point can be used as... When using When substituting into the above formula, the right side The subscript i should correspond to the subscript q, indicating the displacement data of the corresponding feature point.

[0163] Preferably, in another embodiment, when , , , When a single feature point is missing, use Substitute into equation (5) Calculation, use Substitute into equation (5) Calculation; Use or Substitute into equation (6) Computation can improve the reliability when a single feature point is missing.

[0164] Preferably, in another embodiment, when the center points of all feature patterns are identified, the large-scale array stitching visual positioning method of the present invention utilizes the redundancy of the number of feature points to improve calculation accuracy. and Substituting the mean into equation (5) Calculation, use and Substituting the mean into equation (5) The calculation reduced random errors.

[0165] Preferably, in another embodiment, in the step of acquiring real-time relative pose by capturing images of corresponding visual targets using the image acquisition module, the large-scale array stitching visual positioning method of the present invention further includes a combined preprocessing method during the visual target group image recognition process to reduce the influence of ambient light on recognition. The combined preprocessing step during the visual target group image recognition process is described in the relevant section of Embodiment 1 and will not be repeated here.

[0166] The steps to calculate real-time pose deviation are as follows:

[0167] Based on the real-time relative pose, calculate the real-time pose deviation between the array surface and the docking frame. .

[0168] Real-time pose deviation should include information on the translational and rotational movements required for the array to reach the stitching completion state. This invention proposes a pose deviation... The calculation method is as follows:

[0169] (7)

[0170] Among them, pose deviation Represented in the form of a homogeneous transformation matrix, it is derived from the real-time relative pose. Compared with the reference pose The inverse matrix is ​​obtained by multiplying the inverse matrix.

[0171] Steps to determine whether the real-time pose deviation is within the set allowable range:

[0172] Determine real-time pose deviation Is it within the allowable range? If there is a real-time pose deviation? If the position is within acceptable limits, the positioning process is complete; if the real-time pose deviation is within acceptable limits... If the position deviation is outside the allowable range, the real-time pose deviation is sent as a guiding command to the stitching mechanism. The stitching mechanism then performs stitching according to the translational and rotational motion amounts given by the pose deviation, repeating the steps of obtaining the real-time relative pose, calculating the real-time pose deviation, and determining whether the real-time pose deviation is within the set allowable range, until the process ends. The calculation methods for the translational and rotational motion amounts are as follows:

[0173] (8)

[0174] (9)

[0175] (10)

[0176] in, For real-time pose deviation The rotational motion relative to the reference coordinate system includes ( ), Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. For real-time pose deviation The translational motion relative to the reference coordinate system includes ( ), Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis. The allowable range is determined based on the physical structure of the array and the docking frame, meaning it does not exceed the tolerance range of the mechanical structure at the docking point between the array and the docking frame, in order to meet the splicing accuracy requirements, and is the same as the allowable range used in the initial teaching.

[0177] It is understood that when the image acquisition module 5 contains three or more cameras and the visual target 3 contains three or more visual targets, the coordinates of the four feature points can still be defined by defining the average value of the coordinates of each feature point, and the visual positioning can still be performed according to the steps described in Embodiment 2.

[0178] While the present invention has been disclosed above with reference to preferred embodiments, these embodiments are not intended to limit the invention. Any equivalent changes or modifications made without departing from the spirit and scope of the invention are also within the scope of protection of the invention. Therefore, the scope of protection of the present invention should be determined by the claims of this application.

Claims

1. A visual positioning method for a large array splicing mechanism, characterized in that, A large array splicing mechanism visual positioning system is used to splice the array surfaces with the docking frame according to the set pose. The system includes an array splicing mechanism, a visual target, an image acquisition module, and a control module. The array splicing mechanism includes a static platform that is stationary relative to the workspace and a moving platform that moves relative to the workspace. The array is stably connected to the moving platform. The array splicing mechanism receives instructions from the control module to adjust the relative position of the array and the docking frame. The visual target is mounted on the end face of the docking frame facing the array surface; An image acquisition module is installed on the array splicing mechanism to acquire image information of the visual target; The control module receives image information of the visual target acquired by the image acquisition module, performs image processing, obtains the real-time relative pose of the array surface and the docking frame from the image information of the visual target, calculates the real-time pose deviation between the array surface and the docking frame based on the real-time relative pose, and sends a command to the array surface stitching mechanism based on the real-time pose deviation. The array surface stitching mechanism adjusts the relative pose of the array surface and the docking frame so that the real-time pose deviation is within the set allowable range, thereby enabling the array surface to be stitched with the docking frame according to the set pose. The visual positioning method for the large array splicing mechanism includes the following steps: The steps for calculating the real-time relative pose are as follows: The image acquisition module acquires image information of the visual target; the control module receives the image information of the visual target, performs image processing, and calculates the real-time relative pose between the array surface and the docking frame based on the image information of the visual target. ; The steps for calculating the real-time pose deviation are as follows: based on the real-time relative pose... Calculate the real-time pose deviation between the array surface and the docking frame. ; The steps for determining whether the real-time pose deviation is within the set allowable range are as follows: If the real-time pose deviation is within the set allowable range, the process ends; if the real-time pose deviation is not within the allowable range, a command is sent to the array splicing mechanism based on the real-time pose deviation. The array splicing mechanism adjusts the relative pose of the array and the docking frame according to the translational and rotational motion amounts in the command. The steps of calculating the real-time relative pose, calculating the real-time pose deviation, and determining whether the real-time pose deviation is within the set allowable range are repeated until the process ends.

2. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, The real-time relative pose between the array surface and the docking frame is calculated using the image information of the visual target. Specifically, it includes: Obtain the coordinates of the center points of the first, second, third, and fourth feature images on the visual target, respectively. , , , All of these include the x, y, and z spatial information of the center point of the graphic in the static platform reference frame; the control module calculates the real-time relative pose of the array surface and the docking frame according to the following formula. , （4） （5） （6） in, , represents the real-time pose change of the array relative to the static platform coordinate system, and is a 3x3 matrix. express point to spatial vectors, Indicates to Modulo operation on vectors Represents the cross product of vectors. This represents the real-time pose of the array relative to the static platform coordinate system, where... ,Right now , where x, y, and z represent the translation amounts, respectively.

3. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, The real-time pose deviation between the array surface and the docking frame is calculated based on the real-time relative pose. Specifically, it includes: （7） Among them, real-time pose deviation Represented in the form of a homogeneous transformation matrix, it is derived from the real-time relative pose. Compared with the reference pose The inverse matrix is ​​obtained by multiplying the inverse matrix.

4. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, The calculation methods for the translational and rotational motion are as follows: （8） （9） （10） in, For real-time pose deviation The rotational motion relative to the reference coordinate system includes ( ), Let be the amount of rotation about the x-axis. Let be the amount of rotation about the y-axis. Let be the rotation about the z-axis. For real-time pose deviation The translational motion relative to the reference coordinate system includes ( ), Let x be the displacement along the x-axis. This represents the displacement along the y-axis. This represents the displacement along the z-axis.

5. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, The step of calculating the real-time relative pose also includes an initial teaching step, specifically: Step 2.1: Use instruments to detect the pose deviation of the array surface relative to the docking frame. The pose deviation should include translational deviations (x, y, z) and rotational deviations (Rx, Ry, Rz) required for the array surface to reach the completed assembly state. The translational deviations (x, y, z) are the lengths of the projections of the lines connecting the current position and the completed assembly position on the mounting frame, measured in the x, y, and z directions with the static platform coordinate system as the reference direction. The rotational deviation is the deviation obtained by subtracting the current rotational values ​​(Rx, Ry, Rz) of the array surface in the Rx, Ry, and Rz directions from the corresponding values ​​at the completed assembly position on the mounting frame, measured in the static platform coordinate system as the reference direction. Step 2.2: Determine the pose deviation. If the pose deviation is outside the set allowable range, it is determined that the splicing of the array surface and the docking frame is incomplete. The control module outputs the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the pose deviation to the splicing mechanism moving platform. The splicing mechanism moving platform moves according to the translational motion x, y, z and rotational motion Rx, Ry, Rz values ​​given by the pose deviation to adjust the attitude of the array surface. Repeat steps 2.1 and 2.

2. If the pose deviation is within the set allowable range, it is determined that the splicing of the array surface and the docking frame is complete, and proceed to step 2.

3. Step 2.3: Record the pose deviation between the array surface and the docking frame at this time as the initial teaching reference pose to complete the initial teaching. In the step of calculating the real-time relative pose, the initial teaching reference pose is used as the initial value of the real-time relative pose.

6. The visual positioning method for a large array splicing mechanism according to claim 2, characterized in that, When the , , , When a single feature point is missing, use Substitute into equation (5) Calculation, use Substitute into equation (5) Calculation; Use or Substitute into equation (6) calculate.

7. The visual positioning method for a large array splicing mechanism according to claim 2, characterized in that, When the center point of all feature graphics has been identified, use and Substituting the mean into equation (5) Calculation, use and Substituting the mean into equation (5) calculate.

8. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, It also includes a combined preprocessing method during the image acquisition module's acquisition of image information of the visual target. The combined preprocessing steps are as follows: S102, Gaussian blur processing is performed on the original image to filter high-frequency information in the original image; S103 utilizes the gradient information of the grayscale image to separate the parts with large grayscale gradients and detect local feature edges. S104, perform morphological transformation; S105 performs contour detection to obtain the contours of feature points.

9. The visual positioning method for a large array splicing mechanism according to claim 1, characterized in that, The system includes multiple visual targets; the number of image acquisition modules is the same as the number of visual targets, and each image acquisition module acquires the image information of the corresponding visual target; each image acquisition module is equipped with a control module, and the image information of each visual target is processed in the corresponding control module. The control modules exchange visual target data, and one of the control modules sends a command to the stitching mechanism based on the real-time pose deviation.

Images