Large component flatness automatic measurement system based on high-precision real-time visual positioning

Abstract

The invention discloses a large component flatness automatic measurement system based on high-precision real-time visual localization, which relates to the technical field of automatic measurement and comprises a measurer reference building module, a measurement area planning module, an image acquisition and analysis module and a flatness analysis module. The method comprises the following steps: firstly, acquiring spatial state data of a target large component through a measurer reference establishment module, and establishing an actual measurement platform in combination with historical use data of a measurement platform in a database; the measurement area planning module collects component space measurement data, divides a measurement area and generates a platform movement track; the image acquisition and analysis module acquires pictures according to the motion trail, generates a re-detection scheme through analysis, and acquires effective use pictures; and the flatness analysis module calculates the flatness based on the used picture and carries out early warning, so that the automation and precision of the flatness measurement of the large component are improved, and the measurement efficiency and reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of automatic measurement technology, and more specifically to an automatic measurement system for the flatness of large components based on high-precision real-time visual positioning. Background Technology

[0002] In high-end equipment fields such as aerospace, shipbuilding, and heavy machinery, the flatness of large components is a key indicator affecting product assembly accuracy, operational performance, and service life. These components are typically large in size, heavy in weight, and complex in structure, requiring extremely high flatness control, often within the micrometer or sub-millimeter range. Therefore, an automatic flatness measurement system for large components based on high-precision real-time visual positioning is needed.

[0003] Traditional methods for measuring the flatness of large components mainly rely on manual operation, such as using tools like levels, dial indicators, and straightedges for contact measurement. These traditional methods are not only time-consuming and labor-intensive with extremely low measurement efficiency, but they are also greatly affected by human operating experience and subjective judgment, making it difficult to guarantee the consistency and reliability of measurement accuracy. In addition, contact measurement may cause scratches on the surface of the component, making it particularly unsuitable for large components with high surface accuracy requirements.

[0004] With the development of industrial automation technology, some semi-automatic measuring equipment has emerged, such as laser trackers and total stations. Although semi-automatic measuring equipment has improved the measurement accuracy to a certain extent, it still requires manual assistance to adjust the measurement posture and measurement point position. The degree of automation is limited. For large components with complex shapes or obstructions, it is difficult to achieve automatic full-surface coverage measurement. Moreover, the splicing and integration of measurement data requires complex manual processing, which can easily introduce cumulative errors. It cannot meet the needs of modern production for rapid and accurate detection of the flatness of large components.

[0005] In the field of traditional visual measurement technology, the ultra-large size of large components exceeds the effective measurement range of a single vision sensor, requiring multi-view image stitching or mobile measurement platform to work together. Traditional technologies cannot perform high-precision real-time positioning and trajectory planning, and cannot ensure the spatial consistency of image data. Summary of the Invention

[0006] To address the aforementioned technical shortcomings, the present invention aims to provide an automatic measurement system for the flatness of large components based on high-precision real-time visual positioning.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The present invention provides an automatic measurement system for the flatness of large components based on high-precision real-time visual positioning, including the following modules: a measuring instrument reference construction module, used to collect spatial state data of the target large component, analyze the spatial state data of the target large component, generate a measurement platform construction scheme, generate each digital measurement platform according to the measurement platform construction scheme, select the best digital measurement platform from each digital measurement platform based on the historical usage data of the measurement platform in the database, and build the target measurement platform according to the best digital measurement platform.

[0008] The measurement area planning module is used to collect spatial measurement data of the target large component, analyze the spatial measurement data of the large component, divide the target large component into various measurement areas, and then generate the motion trajectory of the target measurement platform.

[0009] The image acquisition and analysis module is used to acquire signal data of one type of detection images of each large target component based on the motion trajectory of the target measurement platform, analyze the signal data of one type of detection images of each large target component to obtain the measurement area of ​​two types of detection images of each large target component, and then generate a re-detection scheme to obtain each usable image of the large target component.

[0010] The flatness analysis module is used to calculate the flatness of the target large component based on various images of the target large component, and to issue warnings based on the calculation results.

[0011] Preferably, the specific generation process of the motion trajectory of the target measurement platform is as follows: The structural feature parameters of each identification point on the surface of the target large component are similar to the various connection structure feature parameters in the database to obtain the similarity of various connection structures of each identification point on the surface of the target large component. Each identification point on the surface of the target large component with a maximum structural similarity greater than a preset structural similarity is recorded as a connection identification point, thus obtaining the similarity of various connection structures of each connection identification point. The connection structure type with the maximum connection similarity is selected as the connection structure type of each connection identification point, thus obtaining the connection structure type of each connection identification point. Based on the three-dimensional coordinates of each connection identification point, connection identification points with consecutive three-dimensional coordinates are recorded as the same group of connection identification points, thus obtaining each group of connection identification points. If the connection structure type of a group of connection identification points is the same, the group of connection identification points is recorded as a continuous identification point group, thus obtaining each continuous identification point group.

[0012] By merging the component regions containing identification points from the same continuous identification point group, we obtain the measurement regions of the target large component.

[0013] The beneficial effects of this invention are as follows: 1. This invention first collects the spatial state data of the target large component through the measuring instrument reference construction module, and combines it with the historical usage data of the measuring platform in the database to build an actual measuring platform; the measuring area planning module collects the spatial measurement data of the component, divides the measuring area and generates the platform motion trajectory; the image acquisition and analysis module collects images based on the motion trajectory, generates a re-detection scheme after analysis, and obtains effective usage images; the flatness analysis module calculates the flatness based on the usage images and provides early warning, which improves the automation and accuracy of the flatness measurement of large components and enhances the measurement efficiency and reliability.

[0014] 2. The system adopts high-precision real-time visual positioning technology. The image acquisition and analysis module performs multi-dimensional analysis on the detection image, combines the SIFT algorithm to extract feature points and uses the triangulation principle to calculate the three-dimensional coordinates, and then uses the ICP algorithm to stitch the feature point cloud. Finally, the minimum region method is used to calculate the flatness, which effectively reduces the errors caused by manual operation and data stitching in traditional measurement and increases the accuracy range of flatness measurement.

[0015] 3. Significantly improves measurement efficiency: The measurement area planning module reduces redundant measurements and invalid movements by dividing the measurement area and generating the optimal motion trajectory; the re-detection scheme performs targeted detection only on the measurement area of ​​the second-class detection image, reducing redundant operations and significantly shortening the measurement time, which is suitable for the rapid detection needs in the mass production of large components.

[0016] 4. Enhanced automation and intelligence in the measurement process: The system can complete a series of operations such as measurement platform setup, trajectory planning, image acquisition and analysis, and flatness calculation and early warning without manual intervention. The measuring instrument reference setup module can automatically select the best measurement platform based on the spatial state data of the target large component. The measurement area planning module can adaptively divide the area and generate motion trajectory, demonstrating strong intelligent adaptive capabilities and reducing reliance on the professional skills of operators.

[0017] 5. Reduce the risk of damage to component surfaces: The non-contact visual measurement method avoids direct contact between traditional contact measurement tools and component surfaces, effectively preventing scratches on component surfaces, and is especially suitable for measuring large components with high surface accuracy requirements. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0020] 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.

[0021] according to Figure 1 As shown, the present invention provides an automatic measurement system for the flatness of large components based on high-precision real-time visual positioning, including the following modules: a measuring instrument reference construction module, a measurement area planning module, an image acquisition and analysis module, a flatness analysis module, and a database.

[0022] The measurement area planning module is connected to the measuring instrument reference construction module and the image acquisition and analysis module, respectively. The flatness analysis module is connected to the image acquisition and analysis module. The measuring instrument reference construction module, the measurement area planning module, the image acquisition and analysis module, and the flatness analysis module are all connected to the database.

[0023] The measuring instrument reference construction module is used to collect spatial state data of the target large component, analyze the spatial state data of the target large component, generate a measurement platform construction plan, generate various digital measurement platforms based on the measurement platform construction plan, select the best digital measurement platform from the various digital measurement platforms based on the historical usage data of the measurement platform in the database, and build the target measurement platform based on the best digital measurement platform.

[0024] In one specific embodiment, the spatial state data of the target large component is collected in the following process: the spatial state data of the target large component includes the geometric parameter data, spatial posture data and surface state data of the target large component. Macroscopic images are collected by a camera, and then the spatial state data is collected by image recognition technology.

[0025] Geometric parameter data includes, but is not limited to, the overall dimensions, local structural features, and overall shape of the component, and is usually represented by the coordinates of the lidar point cloud.

[0026] Spatial attitude data includes, but is not limited to, translation parameters, rotation parameters, relative distance to surrounding equipment, relative perpendicularity to surrounding equipment, and relative parallelism to surrounding equipment in the coordinate system.

[0027] Surface condition data includes, but is not limited to, material reflectivity, material roughness, type of environmentally attached obstructions, area of ​​environmentally attached obstructions, and area covered by oil stains.

[0028] In one specific embodiment, the analysis of the spatial state data of the target large component is carried out as follows: The utilization rate of the geometric parameters of each standard component of each first-class platform structure is obtained from the database; the similarity between the geometric parameters of each standard component and the geometric parameters of the target large component is calculated to obtain the geometric similarity of each standard component to the target large component; the geometric correction factor corresponding to each geometric similarity is obtained from the database to obtain the geometric correction factor of each standard component; the utilization rate of the geometric parameters of each standard component of each first-class platform structure is multiplied by the geometric correction factor to obtain the utilization index of the target large component of each first-class platform structure; the first-class platform structures are arranged in descending order of the utilization index of the target large component to obtain a sequence of first-class platform structures; and a predetermined number of first-class platform structures at the top of the first-class platform structure sequence are selected as pre-constructed first-class platform structures.

[0029] The utilization rate of the spatial attitude of each standard component of each second-class platform structure and the attitude correction factor corresponding to the similarity of each attitude are obtained from the data. Based on the acquisition method of each pre-constructed first-class platform structure, each pre-constructed second-class platform structure is obtained.

[0030] The utilization rate of the surface state of each standard component of each second-class platform structure and the surface correction factor corresponding to each surface similarity are obtained from the data. Based on the acquisition method of each prefabricated first-class platform structure, each prefabricated third-class platform structure is obtained.

[0031] The measurement platform construction scheme is as follows: each of the pre-constructed first-class platform structures, each of the pre-constructed second-class platform structures, and each of the pre-constructed third-class platform structures are combined to obtain each digital measurement platform.

[0032] In one specific embodiment, the process of selecting the best digital measurement platform from each digital measurement platform is as follows: the historical usage data of the measurement platform includes the measurement platform parameters, component parameters and usage effect index of each historical measurement. The similarity between the component parameters of each historical measurement and the target large component is calculated to obtain the construction similarity of each historical measurement. Each historical measurement with a construction similarity greater than the preset construction similarity is recorded as a similar historical measurement.

[0033] It should be noted that the measurement platform parameters are the characteristic data of the measurement platform, including but not limited to the dimensions, gantry dimensions and camera accuracy. The component parameters include but are limited to geometric parameter data, spatial attitude data and surface condition data. The usage effect index is: the area of ​​the measurement area of ​​each type II detection image collected divided by the total area of ​​the corresponding collection area.

[0034] The component parameters are the spatial state data from the historical measurement process.

[0035] The similarity between the measurement platform parameters of each similar historical measurement and the feature parameters of each digital measurement platform is calculated to obtain the platform similarity of each digital measurement platform with similar historical measurements. A certain digital measurement platform is selected as the target digital measurement platform. Each similar historical measurement with a platform similarity greater than the preset platform similarity is recorded as a valid historical measurement of the target digital measurement platform. The similarity of each valid historical measurement of the target digital measurement platform is divided by the sum of the similarities of each valid historical measurement to obtain the weight factor of each valid historical measurement. Then, the usage effect index of each valid historical measurement of the target digital measurement platform is weighted and calculated to obtain the usage effect index of the target digital measurement platform. Thus, the usage effect index of each digital measurement platform is obtained.

[0036] The digital measurement platform with the highest usage effectiveness index is selected as the optimal digital measurement platform.

[0037] The measurement area planning module is used to collect spatial measurement data of the target large component, analyze the spatial measurement data of the large component, divide the target large component into various measurement areas, and then generate the motion trajectory of the target measurement platform.

[0038] In one specific embodiment, the spatial measurement data of the target large component is collected in the following way: the spatial measurement data of the target large component includes the three-dimensional coordinates and structural feature parameters of each identification point on the surface of the target large component, which are collected by equipment such as lidar, structured light sensor, and total station.

[0039] Structural characteristic parameters include, but are not limited to, the position coordinates, dimensions, and angles of the raised stiffeners.

[0040] In one specific embodiment, the analysis of the spatial measurement data of the large component is carried out as follows: a digital twin model is constructed based on the three-dimensional coordinates of each identification point on the surface of the target large component, and the digital twin model of the target large component is obtained. The structural feature parameter range of the reference edge points is obtained from the database. The similarity between the structural feature parameter range of the reference edge points and the structural feature parameters of each identification point on the surface of the target large component is calculated to obtain the reference edge similarity of each identification point on the surface of the target large component. The identification points whose reference edge similarity is greater than the preset standard reference edge similarity are recorded as reference edge identification points, and then the three-dimensional coordinates of each reference edge identification point are obtained. Reference edges are constructed in the digital twin model of the target large component based on the three-dimensional coordinates of each reference edge identification point.

[0041] A two-dimensional coordinate system is established with the reference edge of the target large component as the axis. Rectangular grid areas are divided according to a preset interval to obtain the regions of each component. Then, the regions of each component are integrated into each measurement region to obtain the measurement regions of the target large component.

[0042] In one specific embodiment, the motion trajectory of the target measurement platform is generated as follows: The structural feature parameters of each identification point on the surface of the target large component are similar to the various connection structure feature parameters in the database to obtain the similarity of various connection structures of each identification point on the surface of the target large component. Each identification point on the surface of the target large component with a maximum structural similarity greater than a preset structural similarity is recorded as a connection identification point, thus obtaining the similarity of various connection structures of each connection identification point. The connection structure type with the maximum connection similarity is selected as the connection structure type of each connection identification point, thus obtaining the connection structure type of each connection identification point. Based on the three-dimensional coordinates of each connection identification point, connection identification points with consecutive three-dimensional coordinates are recorded as the same group of connection identification points, thus obtaining each group of connection identification points. If the connection structure type of a group of connection identification points is the same, the group of connection identification points is recorded as a continuous identification point group, thus obtaining each continuous identification point group.

[0043] It should be noted that the types of connection structures include, but are not limited to, bolted structures and welded structures.

[0044] By merging the component regions containing identification points from the same continuous identification point group, we obtain the measurement regions of the target large component.

[0045] In one specific embodiment, the motion trajectory of the target measurement platform is generated as follows: the center coordinates of each measurement area are calculated, and the measurement areas with spatial distances less than the standard spatial distance are divided into the same group using the K-means clustering algorithm to obtain each measurement area group. A weight factor is set according to the number of connection identification points in each measurement area. Combined with the spatial coordinates of the area center, the weighted path is calculated using the shortest path algorithm: starting from the starting point, areas with high weights and close distances are visited first to obtain the area access order.

[0046] It should be noted that the K-means clustering algorithm is an existing technology that can be found on the Internet, so it will not be described in detail here. The weight factor of each measurement region is obtained by connecting the number of identification points. The larger the number of identification points, the larger the weight factor.

[0047] Measurement areas without connection identification points are designated as Class I areas, and measurement areas with connection identification points are designated as Class II areas. A serpentine scanning path is used to scan Class I areas. In Class II areas, a main path is generated along the distribution direction of continuous identification point groups, and intersecting scan lines perpendicular to the main path are generated at preset intervals. In Class II areas, the measurement platform moves along the main path. When it reaches the intersection of each intersecting scan line and the main path, it moves along the corresponding intersecting scan line. After regressing, it continues to move along the main path until the main path detection is completed, thus obtaining the movement path within the area.

[0048] It should be noted that the snake scanning path is an existing technology that can be found on the Internet, and will not be elaborated further.

[0049] The target measurement platform's movement trajectory is as follows: it visits each measurement area in the order of area access, and moves within each measurement area according to the movement path within that area.

[0050] The image acquisition and analysis module is used to acquire signal data of one type of detection images of each large target component based on the motion trajectory of the target measurement platform, analyze the signal data of one type of detection images of each large target component to obtain the measurement area of ​​two types of detection images of each large target component, and then generate a re-detection scheme to obtain each usable image of the large target component.

[0051] In one specific embodiment, the signal data of each type of detection image of the target large component is acquired as follows: the signal data of each type of detection image of the target large component includes descriptor data, grayscale data and phase information data of each type of detection image of the target large component, which is obtained through image recognition technology.

[0052] In one specific embodiment, the analysis of the signal data of each type of detection image of the target large component is carried out as follows: Select a target type detection image from each type of detection image of the target large component, and then obtain each adjacent type detection image of the target type detection image. Through image recognition technology, obtain each overlapping area image of the target type detection image and the corresponding adjacent overlapping area image.

[0053] It should be noted that the overlapping areas between the target class I detection image and the adjacent class I detection image are denoted as each overlapping area. The image of each overlapping area of ​​the target class I detection image is: the image corresponding to each overlapping area in the target class I detection image, and the image of the corresponding adjacent overlapping area is: the image of the overlapping area in the adjacent class I detection image.

[0054] The feature point matching rate of each overlapping region image of the target-class detection image is obtained by using the descriptor data of each overlapping region image and the corresponding adjacent overlapping region image. The grayscale curve matching rate of each overlapping region image of the target-class detection image is obtained by using the grayscale value data of each overlapping region image and the corresponding adjacent overlapping region image. The phase matching rate of each overlapping region image of the target-class detection image is obtained by using the phase information data of each overlapping region image and the corresponding adjacent overlapping region image.

[0055] It should be noted that the feature point matching rate acquisition process is as follows: A hierarchical clustering tree index is constructed using the descriptor data of each overlapping region image and its corresponding adjacent overlapping region images of the target class detection image. The neighbor distances from each descriptor of each overlapping region image to its corresponding adjacent overlapping region images are obtained through norm calculation. The minimum distance is recorded as the nearest neighbor distance, and the second smallest distance is recorded as the second nearest neighbor distance. This yields the nearest and second nearest neighbor distances from each descriptor of each overlapping region image to its corresponding adjacent overlapping region images. If the ratio of the nearest neighbor distance to the second nearest neighbor distance is less than a preset standard ratio, it is considered a valid match. This yields the effective matching quantity of each overlapping region image of the target class detection image. The effective matching quantity of each overlapping region image of the target class detection image is divided by the total number of descriptors to obtain the feature point matching rate of the target class detection image.

[0056] The standard ratio is the threshold for the normal ratio. Exceeding the threshold indicates that the descriptor does not match. The specific value is set by the staff.

[0057] The grayscale curve matching rate acquisition process is as follows: The repeated areas of the image are divided into image grids, the average grayscale value of each grid is calculated, and the grids are arranged in row and column order to obtain two grayscale curves. The corresponding distance of each element in the two grayscale curve sequences is calculated, and a grayscale distance matrix is ​​constructed. The elements of the grayscale distance matrix are the absolute values ​​of the grayscale differences of the corresponding distances of each element. A cumulative cost matrix is ​​constructed, and the elements are the sum of the minimum absolute values ​​of the grayscale differences of the preceding sequence elements in the grayscale curve sequence. The cumulative cost matrix is ​​traced back from the lower right corner to the upper left corner to find the path with the minimum cumulative distance. The points on the path indicate that the corresponding elements in the two grayscale curve sequences are aligned. The sum of the cumulative distances of the optimal path is the DTW distance. When the DTW distance is less than the preset standard DTW distance, it indicates that the grids match. The grayscale curve matching rate is obtained by dividing the number of matching grids by the total number of image grids. Thus, the grayscale curve matching rate of the target class detection image is obtained.

[0058] The standard DTW distance is the normal DTW distance threshold. Exceeding the threshold indicates that the grayscale curves do not match. The specific value is set by the staff.

[0059] Phase matching rate acquisition process: For corresponding pixels in the overlapping regions of the image, Gabor filters at various scales and directions are used for filtering. After filtering, each pixel obtains a complex response at each scale and direction. The complex response includes a real part and an imaginary part, corresponding to the cosine and sine components, respectively. For pixels at the same spatial position in adjacent images, their complex responses at the same scale and direction are extracted. The complex response of the target image is subtracted from the complex response of the adjacent images at the same scale and direction. The absolute value of the difference is then used to obtain the phase difference. If the phase difference is less than the preset phase difference, the phase is considered to be consistent at that scale and in that direction. The phase consistency count of each pixel is obtained in this way. The phase consistency count of each pixel is divided by the sum of the types at each scale and in each direction to obtain the phase consistency rate of each pixel. If the phase consistency rate is greater than the preset phase consistency rate, it indicates that the pixels are consistent. The pixel consistency count is divided by the total number of pixels to obtain the phase matching rate. This is used to obtain the phase matching rate of each overlapping region image of the target class detection image.

[0060] The preset phase difference is the normal phase difference threshold. Exceeding the threshold indicates phase consistency. The specific value is set by the staff. The preset phase consistency rate is the normal phase consistency rate threshold. Exceeding the threshold indicates phase matching. The specific value is set by the staff.

[0061] The standard matching rate range of feature points, the standard gray-scale curve matching rate range, and the standard phase matching rate range of the matching images are obtained from the database. If the feature point matching rate of a certain overlapping area of ​​the target Class I detection images does not belong to the standard matching rate range of feature points, the gray-scale curve matching rate of a certain overlapping area image does not belong to the standard gray-scale curve matching rate range, or the phase matching rate of a certain overlapping area image does not belong to the standard phase matching rate range, the target Class I detection images are recorded as Class II detection images. In this way, each Class II detection image is obtained, and the measurement area containing the Class II detection images is recorded as the measurement area of ​​each Class II detection image of the target large component.

[0062] In one specific embodiment, the generation process of the re-detection scheme is as follows: for each type of detection image measurement region of the target large component, a weighted path is calculated using the shortest path algorithm based on the corresponding weight factor and the spatial coordinates of the region center to obtain the region access order, and detection is performed according to the region access order.

[0063] In one specific embodiment, the process of obtaining the various usage images of the target large component is as follows: the first-class detection images that are not second-class detection images are recorded as third-class detection images, and so on to obtain various third-class detection images; the images obtained by the measurement areas of each second-class detection image during the re-detection scheme are recorded as various fourth-class detection images; and the various third-class detection images and the various fourth-class detection images are merged to obtain various usage images of the target large component.

[0064] The flatness analysis module is used to calculate the flatness of the target large component based on various images of the target large component, and to issue warnings based on the calculation results.

[0065] In one specific embodiment, the warning process is as follows: feature points of each image are extracted using the SIFT algorithm, the three-dimensional coordinates of the feature points are calculated using the triangulation principle to generate a feature point cloud, the feature point cloud is stitched to a unified coordinate system using the ICP algorithm, and after fitting an ideal plane using the minimum region method, the distance from each feature point to the plane is calculated, and the maximum distance is recorded as the flatness. If the flatness is greater than the preset flatness, a warning is issued.

[0066] It should be noted that the SIFT algorithm, triangulation principle, ICP algorithm and minimum region method are all existing technologies and will not be elaborated further.

[0067] The preset flatness is the flatness threshold of a normal component. When the flatness exceeds the threshold, it indicates that the component is not flat, which will have an adverse effect on the use of the component. The specific value is set by the staff.

[0068] The database is used to store historical usage data of the measurement platform, usage rates of geometric parameters of standard components of each type of platform structure, geometric correction factors corresponding to each geometric similarity, structural feature parameter ranges of reference edge points, feature parameters of various connection structures, standard matching rate ranges of feature points of matched images, standard grayscale curve matching rate ranges of matched images, and standard phase matching rate ranges of matched images.

[0069] The above description is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined in this specification, they should all fall within the protection scope of the present invention.

Claims

1. An automatic measurement system for the flatness of large components based on high-precision real-time visual positioning, characterized in that, Includes the following modules: The measuring instrument reference construction module is used to collect spatial state data of the target large component, analyze the spatial state data of the target large component, generate a measurement platform construction plan, generate various digital measurement platforms according to the measurement platform construction plan, select the best digital measurement platform from the various digital measurement platforms based on the historical usage data of the measurement platform in the database, and build the target measurement platform according to the best digital measurement platform. The measurement area planning module is used to collect spatial measurement data of the target large component, analyze the spatial measurement data of the large component, divide the target large component into various measurement areas, and then generate the motion trajectory of the target measurement platform. The image acquisition and analysis module is used to acquire signal data of one type of detection images of each large target component based on the motion trajectory of the target measurement platform, analyze the signal data of one type of detection images of each large target component to obtain the measurement area of ​​two types of detection images of each large target component, and then generate a re-detection scheme to obtain each use image of the large target component. The flatness analysis module is used to calculate the flatness of the target large component based on various images of the target large component, and to issue warnings based on the calculation results.

2. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 1, characterized in that, The analysis of the spatial state data of the target large component is carried out in the following specific process: The spatial state data of the target large component includes the geometric parameter data, spatial attitude data, and surface state data of the target large component. The utilization rate of the geometric parameters of each standard component of each first-class platform structure is obtained from the database. The similarity between the geometric parameters of each standard component and the geometric parameter data of the target large component is calculated to obtain the geometric similarity of each standard component to the target large component. The geometric correction factor corresponding to each geometric similarity is obtained from the database to obtain the geometric correction factor of each standard component. The utilization rate of the geometric parameters of each standard component of each first-class platform structure is multiplied by the geometric correction factor to obtain the utilization index of the target large component of each first-class platform structure. The first-class platform structures are arranged in descending order of the utilization index of the target large component to obtain the first-class platform structure sequence. The first-class platform structures at the top of the first-class platform structure sequence are selected as each pre-constructed first-class platform structure. The utilization rate of the spatial attitude of each standard component of each second-class platform structure and the attitude correction factor corresponding to the similarity of each attitude are obtained from the data. Based on the acquisition method of each pre-constructed first-class platform structure, each pre-constructed second-class platform structure is obtained. The utilization rate of the surface state of each standard component of each second-class platform structure and the surface correction factor corresponding to each surface similarity are obtained from the data. Based on the acquisition method of each prefabricated first-class platform structure, each prefabricated third-class platform structure is obtained. The measurement platform construction scheme is as follows: each of the pre-constructed first-class platform structures, each of the pre-constructed second-class platform structures, and each of the pre-constructed third-class platform structures are combined to obtain each digital measurement platform.

3. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 2, characterized in that, The specific analysis process for selecting the optimal digital measurement platform from among various digital measurement platforms is as follows: The historical usage data of the measurement platform includes the measurement platform parameters, component parameters and usage effect index of each historical measurement. The similarity between the component parameters of each historical measurement and the target large component is calculated to obtain the construction similarity of each historical measurement. Each historical measurement with a construction similarity greater than the preset construction similarity is recorded as a similar historical measurement. The similarity between the measurement platform parameters of each similar historical measurement and the feature parameters of each digital platform to be measured is calculated to obtain the platform similarity of each digital platform to be measured for each similar historical measurement, and then the usage effect index of each digital platform to be measured is obtained. The digital measurement platform with the highest usage effectiveness index is selected as the optimal digital measurement platform.

4. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 1, characterized in that, The analysis of spatial measurement data for large components is carried out in the following specific process: Spatial measurement data of the target large component includes the three-dimensional coordinates and structural feature parameters of each identification point on the surface of the target large component; A digital twin model is constructed based on the three-dimensional coordinates of each identification point on the surface of the target large component. The structural feature parameter range of the reference edge points is obtained from the database. The similarity between the structural feature parameter range of the reference edge points and the structural feature parameters of each identification point on the surface of the target large component is calculated to obtain the reference edge similarity of each identification point on the surface of the target large component. The identification points whose reference edge similarity is greater than the preset standard reference edge similarity are recorded as reference edge identification points. Then, the three-dimensional coordinates of each reference edge identification point are obtained. Reference edges are constructed in the digital twin model of the target large component based on the three-dimensional coordinates of each reference edge identification point. A two-dimensional coordinate system is established with the reference edge of the target large component as the axis. Rectangular grid areas are divided according to a preset interval to obtain the regions of each component. Then, the regions of each component are integrated into each measurement region to obtain the measurement regions of the target large component.

5. The automatic flatness measurement system for large components based on high-precision real-time visual positioning according to claim 4, characterized in that, The specific process for generating the motion trajectory of the target measurement platform is as follows: The structural feature parameters of each identification point on the surface of the target large component are similar to the various connection structure feature parameters in the database to obtain the similarity of various connection structures of each identification point on the surface of the target large component. The identification points on the surface of the target large component with the maximum structural similarity greater than the preset structural similarity are recorded as each connection identification point, and the similarity of various connection structures of each connection identification point is obtained in this way. The connection structure type with the maximum connection similarity is selected as the connection structure type of each connection identification point, and the connection structure type of each connection identification point is obtained in this way. According to the three-dimensional coordinates of each connection identification point, the connection identification points with continuous three-dimensional coordinates are recorded as the same group of connection identification points, and the groups of connection identification points are obtained in this way. If the connection structure type of a group of connection identification points is the same, the group of connection identification points is recorded as a continuous identification point group, and the continuous identification point groups are obtained in this way. By merging the component regions containing identification points from the same continuous identification point group, we obtain the measurement regions of the target large component.

6. The automatic flatness measurement system for large components based on high-precision real-time visual positioning according to claim 5, characterized in that, The specific process for generating the motion trajectory of the target measurement platform is as follows: Calculate the center coordinates of each measurement area, and use the K-means clustering algorithm to divide the measurement areas with spatial distances less than the standard spatial distance into the same group to obtain each measurement area group. Set the weight factor according to the number of connection identification points in each measurement area, and combine the spatial coordinates of the area center to use the shortest path algorithm to calculate the weighted path: starting from the starting point, prioritize visiting areas with high weights and close distances to obtain the area visiting order. Each measurement area without a connection identification point is designated as the first type of area, and each measurement area with a connection identification point is designated as the second type of area. A serpentine scanning path is used to scan the first type of area. In the second type of area, a main path is generated along the distribution direction of the continuous identification point group, and each cross scan line perpendicular to the main path is generated at preset intervals. In the second type of area, the measurement platform moves along the main path. When it reaches the intersection of each cross scan line and the main path, it moves along the corresponding cross scan line. After returning, it continues to move along the main path until the main path detection is completed, thus obtaining the movement path within the area. The target measurement platform's movement trajectory is as follows: it visits each measurement area in the order of area access, and moves within each measurement area according to the movement path within that area.

7. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 6, characterized in that, The signal data of each type of detection image of the target large component is analyzed, and the specific analysis process is as follows: The signal data of each type of detection image of the target large component includes descriptor data, grayscale data and phase information data of each type of detection image of the target large component; From the target large component, select the target type detection image from each type of detection image, and then obtain the target type detection image and its adjacent type detection images. Through image recognition technology, obtain the overlapping area images of the target type detection images and the corresponding adjacent overlapping area images. Based on the signal data of each overlapping region image and the corresponding adjacent overlapping region image of the target class detection image, the feature point matching rate, grayscale curve matching rate and phase matching rate of each overlapping region image of the target class detection image are obtained. The database is used to obtain the standard matching rate range of feature points, the standard gray-scale curve matching rate range, and the standard phase matching rate range of the matching images. If the feature point matching rate, gray-scale curve matching rate, or phase matching rate of a certain overlapping area of ​​the target Class I detection images does not belong to the standard matching rate range of feature points, the standard gray-scale curve matching rate range, or the standard phase matching rate range, it is recorded as a Class II detection image. In this way, each Class II detection image is obtained, and the measurement area containing the Class II detection images is recorded as the measurement area of ​​each Class II detection image of the target large component.

8. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 7, characterized in that, The specific generation process of the re-detection scheme is as follows: For each type of large target component, the measurement area of ​​each detection image is calculated using the shortest path algorithm based on the corresponding weight factor and the spatial coordinates of the area center to obtain the area access order, and detection is performed according to the area access order.

9. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 8, characterized in that, The specific analysis process for issuing the early warning is as follows: The SIFT algorithm is used to extract feature points from each image. The three-dimensional coordinates of the feature points are calculated using the triangulation principle to generate a feature point cloud. The ICP algorithm is used to stitch the feature point cloud to a unified coordinate system. The minimum region method is used to fit an ideal plane. The distance from each feature point to the plane is calculated. The maximum distance is recorded as the flatness. If the flatness is greater than the preset flatness, an early warning is issued.

10. The automatic measurement system for the flatness of large components based on high-precision real-time visual positioning according to claim 1, characterized in that, It also includes a database for storing historical usage data of the measurement platform, the usage rate of geometric parameters of standard components of each type of platform structure, the geometric correction factor corresponding to each geometric similarity, the range of structural feature parameters of the reference edge points, the feature parameters of various connection structures, the range of standard matching rate of feature points of the matching images, the range of standard grayscale curve matching rate of the matching images, and the range of standard phase matching rate of the matching images.

Images