Method for automatically detecting levelness of radio frequency coaxial connector based on point cloud data

Abstract

The invention belongs to the technical field of radar antenna automatic assembly, and particularly relates to a radio frequency coaxial connector levelness automatic detection method based on point cloud data, and the method comprises the steps: extracting all local plane features in the point cloud data of a radio frequency coaxial connector through employing a local plane detection algorithm with prior shape constraint based on RANSAC; the automatic detection capability can be modularly embedded into the automatic assembly process of the phased array radar antenna to form an assembly-measurement integrated phased array antenna automatic assembly unit, so that the assembly efficiency and the assembly consistency of the radar antenna assembly are greatly improved. Compared with a traditional manual measurement or three-coordinate measurement mode, the method has the advantage that the detection efficiency of the levelness of the radio frequency coaxial connector can be improved by 5-10 times by combining surface structured light scanning with the algorithm provided by the invention. The antenna does not need to be frequently moved and adjusted due to limitation of measurement hardware, and in-situ measurement is achieved. And meanwhile, automatic data analysis and calculation do not need manual participation, so that the labor cost is greatly reduced.

Description

Technical Field

[0001] This application belongs to the field of automatic assembly technology of radar antennas, and specifically relates to an automatic detection method for the flushness of radio frequency coaxial connectors based on point cloud data. Background Technology

[0002] The assembly quality of the TR (Transmission Terminal) assembly in a phased array radar antenna is a crucial parameter for evaluating the antenna's detection performance. The RF coaxial connectors on the TR assembly connect the TR assembly to components such as the antenna and signal processing unit, ensuring reliable high-frequency signal transmission and providing a stable connection between the TR assembly and the antenna. Radar assembly processes place high demands on the consistency of the end face height of the RF coaxial connectors. However, phased array antennas typically require the installation of hundreds of TR assemblies, each with several RF coaxial connectors. Due to the cumulative effects of manufacturing tolerances and assembly errors, it is difficult to guarantee that the end faces of all RF coaxial connectors are in the ideal position after assembly.

[0003] Due to the sheer number of RF coaxial connectors in the overall antenna array, traditional methods involve measuring each one individually with vernier calipers. This approach is extremely inefficient and struggles to establish a unified height measurement benchmark, failing to meet the requirements for RF coaxial connector flushness testing. While coordinate measuring machines (CMMs) can currently provide high-precision measurements, the extremely thin walls of RF coaxial connectors necessitate extensive manual assistance for positioning. Furthermore, the large size and weight of the antenna require its transport to the CMM platform. Therefore, this method is both inefficient and cumbersome in practical application, making it unsuitable for the demands of modern digital and agile manufacturing.

[0004] Currently, with advancements in digital measurement technology, a high-precision structured light scanner can be used to digitally measure the flatness of all RF coaxial connectors on the antenna array. By processing and analyzing the point cloud data obtained from the scan, the height of all RF coaxial connectors can be quickly measured. However, the point cloud data obtained from a single measurement includes not only the RF coaxial connectors but also other structural data from the TR assembly. Furthermore, because the wall thickness of the RF coaxial connectors is only on the order of millimeters, extracting the end-face data of the RF coaxial connectors from the point cloud data is extremely difficult. Therefore, to solve these problems...

[0005] This invention proposes a digital detection method for the flushness of RF coaxial connectors in large-scale phased array antennas, enabling automatic extraction and calculation of the height of all RF coaxial connectors on the antenna array surface. Summary of the Invention

[0006] The purpose of this application is to provide an automatic detection method for the flushness of RF coaxial connectors based on point cloud data, so as to solve the problem of low efficiency in the existing RF coaxial connector height calculation.

[0007] The technical solution of this application is: an automatic detection method for the flushness of RF coaxial connectors based on point cloud data, comprising:

[0008] Scan to acquire point cloud data of a single TR component of the antenna;

[0009] Statistical filtering algorithms are used to remove outliers from the point cloud data of the TR component;

[0010] The RANSAC algorithm is used to extract the planar data of the upper surface of the TR component point cloud data after removing outliers, obtain the normal vector of the planar data of the upper surface, and then segment and remove the extracted planar data of the upper surface to obtain the point cloud data of the RF coaxial connector.

[0011] A RANSAC-based local plane detection algorithm with prior shape constraints is used to extract all local plane features from the point cloud data of the RF coaxial connector.

[0012] Extract the normal vectors from all local planar features, find the local planar features corresponding to the normal vectors that are parallel to the normal vectors of the planar data on the upper surface, and use them as the point cloud of the RF coaxial connector end face.

[0013] The continuity of point cloud data is used to segment it into a single RF coaxial connector end face point cloud. The height value of the single RF coaxial connector end face point cloud is calculated by combining the planar data of the upper surface of the TR component point cloud data.

[0014] The point cloud data of other TR components are scanned sequentially, and the height value of the point cloud of the end face of each individual RF coaxial connector is calculated repeatedly until the height value of the point cloud of the end face of all individual RF coaxial connectors is calculated.

[0015] Preferably, the method for extracting all local planar features from the point cloud data of the RF coaxial connector is as follows:

[0016] Set the number of sampling times. In each sampling time, randomly collect 3 points. When the normal vectors of the 3 points are the same, they are fitted into a plane. Set a first threshold and calculate the points within the range of the first threshold as the interior points of the plane. Calculate the largest set of connected interior points as the set of points supporting the plane, which is used as the local plane feature. Repeat the collection and extraction of local plane features until the set number of sampling times is reached, and count all the local plane features found.

[0017] Preferably, the number of sampling times ,in The ratio is the ratio of interior points.

[0018] Preferably, the method for obtaining the planar data of the upper surface of the TR component point cloud data is as follows:

[0019] Three points are randomly collected from the point cloud data of the RF coaxial connector. The plane formed by these three points is used to calculate the distance from all points in the TR component point cloud data after removing outliers to this plane. A second distance threshold is set, and a plane formed by a certain proportion of points within the second distance threshold range is taken as the target plane. The proportion of the number of points in the target plane to the total number of points is determined. If it is greater than 90%, the target plane is considered to be the plane data of the upper surface of the TR component point cloud data.

[0020] Preferably, the method for removing outliers from the point cloud data of the TR component is as follows:

[0021] Calculate the neighborhood distance of each point in the point cloud data of a single TR component, and calculate the average neighborhood distance of all points; calculate the mean of the average neighborhood distances of all points. and standard deviation ;

[0022] Set a neighborhood determination threshold to determine whether the average distance of a point's neighborhood exceeds the neighborhood determination threshold. If so, the point is determined to be an outlier and is removed.

[0023] Preferably, when calculating the neighborhood distance of each point, for each point in the point cloud... Apply K-nearest neighbor search to find the k nearest points, and calculate the distance between the k nearest points and the point. The distance is denoted as ,in A point in the neighborhood;

[0024] The formula for calculating the average neighborhood distance is:

[0025] ;in, For point The corresponding average neighborhood distance;

[0026] The mean of the neighborhood average distance and standard deviation for:

[0027] , ;

[0028] The formula for determining whether the average neighborhood distance of a point exceeds the neighborhood determination threshold is:

[0029] .

[0030] The automatic detection method for RF coaxial connector flushness based on point cloud data in this application can modularly embed automated detection capabilities into the automated assembly process of phased array radar antennas, forming an automated assembly unit for phased array antennas that integrates assembly and testing, thereby significantly improving the assembly efficiency and consistency of radar antenna components.

[0031] Compared to traditional manual or coordinate measuring machine (CMM) methods, the use of structured light scanning combined with the algorithm of this invention can improve the efficiency of RF coaxial connector flushness detection by 5-10 times. It eliminates the need for frequent antenna adjustments due to hardware limitations, enabling in-situ measurement. Furthermore, its automated data analysis and calculations require no manual intervention, significantly reducing labor costs. Attached Figure Description

[0032] To more clearly illustrate the technical solutions provided in this application, the accompanying drawings will be briefly described below. Obviously, the drawings described below are merely some embodiments of this application.

[0033] Figure 1 This is a schematic diagram of the overall process of this application;

[0034] Figure 2 This is a diagram showing the working scene of the structured light scanner in this application and its positional relationship with the antenna.

[0035] Figure 3 This application describes the structure of the TR component and its RF coaxial connector.

[0036] 1. Robotic arm; 2. Surface structured light scanner; 3. Antenna; 4. TR assembly; 5. RF coaxial connector. Detailed Implementation

[0037] 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.

[0038] An automatic detection method for the flushness of RF coaxial connectors based on point cloud data, such as Figure 1 It includes the following steps:

[0039] Step S100: Scan and acquire point cloud data of a single TR component of the antenna.

[0040] The RF coaxial connectors are located on the upper surface of the TR components and are arranged in a cylindrical array. Hundreds of TR components are mounted on a single antenna. To automate the measurement of the RF coaxial connector height, a robotic arm equipped with a structured light scanner is used to acquire point cloud data of each TR component individually. The robotic arm equipped with the structured light scanner is described in detail below. Figure 2 As shown, it includes a robotic arm, a surface structured light scanner, and an antenna; the surface structured light scanner is located at the top of the robotic arm, and the antenna is located on one side of the surface structured light scanner.

[0041] Step S200: Use a statistical filtering algorithm to remove outliers from the point cloud data of the TR component.

[0042] Because the scanner's scanning direction is aligned with the axial direction of the RF coaxial connector, the point cloud data of the TR component acquired by the scan has good data quality in the XY plane, such as... Figure 3 These data include the upper surface plane of TR component 4 and the RF coaxial connector, and the end face of 5. Because the scanning angle is almost parallel to the feature shape, the point cloud data located on the sidewalls of the TR component and the RF coaxial connector has poor data quality. These point clouds are considered outliers and need to be removed.

[0043] Preferably, the method for removing outliers from the point cloud data of the TR component is as follows:

[0044] Calculate the neighborhood distance of each point in the point cloud data of a single TR component, and calculate the average neighborhood distance of all points; calculate the mean of the average neighborhood distances of all points. and standard deviation ;

[0045] Set a neighborhood determination threshold to determine whether the average distance of a point's neighborhood exceeds the neighborhood determination threshold. If so, the point is determined to be an outlier and is removed.

[0046] When calculating the neighborhood distance of each point, for each point in the point cloud Apply K-nearest neighbor search to find the k nearest points, and calculate the distance between the k nearest points and the point. The distance is denoted as ,in A point in the neighborhood;

[0047] The formula for calculating the average neighborhood distance is:

[0048] ;in, For point The corresponding average neighborhood distance;

[0049] For a TR component containing N points, the mean of the average distances between all points in the neighborhood of the point cloud computing system is... and standard deviation for:

[0050] , ;

[0051] Determine the threshold for identifying outliers. Typically, the neighborhood mean of points in a point cloud generally follows a normal distribution, meaning that the neighborhood mean of most points is distributed within a certain range. In the vicinity, only a small number of outliers have neighborhood mean values ​​much greater than [missing value]. This aligns with the characteristics of actual scanned point cloud data: the point cloud data distribution is regular and uniform on the upper surface of the TR component facing the scanner and the end face of the RF coaxial connector, while the point cloud data density is sparse on the sidewalls of the TR component and the sidewalls of the RF coaxial connector parallel to the scanning direction. Therefore, utilizing this characteristic, a setting can be made for the average neighborhood distance of a certain point... When the density of points near a given point is considered to be lower than normal, it is identified as an outlier and removed. This is the standard deviation multiple, which is set to 1.5 in this case.

[0052] The TR component planar data is segmented. After removing outliers, the TR component point cloud data mainly contains two types of feature data: one is the planar data constituting the upper surface of the TR component, and the other is the point cloud data constituting the RF coaxial connector. The RANSAC (Random Sample Consensus) algorithm is used to extract the planar data from the point cloud data, and this planar data is segmented and removed to obtain point cloud data containing only the RF coaxial connector.

[0053] Step S300: Perform planar detection on the point cloud data obtained in the previous step. The traditional RANSAC algorithm can only perform shape detection on the overall point cloud, and cannot perform individual detection on point cloud data containing multiple shapes, and its computational efficiency is low.

[0054] This application employs a RANSAC-based local plane detection algorithm with prior shape constraints to extract all regions that may constitute a plane in each local area. By comparing the vector information of the sampling points with the prior shape in advance, the extraction efficiency is greatly improved.

[0055] The RANSAC algorithm is used to extract the planar data of the upper surface of the TR component point cloud data after removing outliers, obtain the normal vector of the planar data of the upper surface, and then segment and remove the extracted planar data of the upper surface to obtain the point cloud data of the RF coaxial connector.

[0056] Preferably, the method for obtaining the planar data of the upper surface of the TR component point cloud data is as follows:

[0057] Three points are randomly collected from the point cloud data of the RF coaxial connector. The plane formed by these three points is used to calculate the distance from all points in the TR component point cloud data after removing outliers to this plane. A second distance threshold is set, and a plane formed by a certain proportion of points within the second distance threshold range is taken as the target plane. The proportion of the number of points in the target plane to the total number of points is determined. If it is greater than 90%, the target plane is considered to be the plane data of the upper surface of the TR component point cloud data.

[0058] Step S400: Use a RANSAC-based local plane detection algorithm with prior shape constraints to extract all local plane features from the RF coaxial connector point cloud data.

[0059] Preferably, the method for extracting all local planar features from the point cloud data of the RF coaxial connector is as follows:

[0060] Set the number of sampling iterations. In each iteration, randomly collect at least 3 points. When the normal vectors of the 3 points are the same, they are fitted to a plane. The principle for setting the number of sampling iterations is to find the correct shape with a 99% probability. ,in The interior point ratio refers to the proportion of points contained in the detected plane to the total amount of input point cloud data.

[0061] Set a first threshold, and calculate the points within the range of the first threshold as the interior points of the plane; if the orientation deviation of the normal vectors of these three points is greater than the first threshold, return to the previous step to resample.

[0062] The points in the plane extracted in the previous step may be composed of multiple discontinuous point clouds. Calculate the largest connected internal point set as the point set supporting the plane, and use it as a local plane feature. Repeat the collection and extraction of local plane features until the set number of samplings is reached, and count all the local plane features found.

[0063] Step S500: Extract the normal vectors from all local planar features, find the local planar features corresponding to the normal vectors that are parallel to the normal vectors of the planar data on the upper surface, and use them as the point cloud of the RF coaxial connector end face.

[0064] Since the point cloud contains not only the point cloud data of the RF coaxial connector end face, but also the sidewalls and some outliers with higher density, the plane normals obtained from the previous plane detection step have different orientations. However, since the planes formed by the RF coaxial connector end faces all have the same orientation and are all along the Z-axis, this characteristic is utilized to extract the plane whose plane orientation is along the Z-axis and whose bounding box size supporting the plane is larger than the diameter of the RF coaxial connector as the RF coaxial connector end face point cloud.

[0065] Step S600: The continuity of the point cloud data is used to segment the point cloud into a single RF coaxial connector end face point cloud, and the height value of the single RF coaxial connector end face point cloud is calculated by combining the planar data of the upper surface of the TR component point cloud data.

[0066] Preferably, calculate the standard deviation of the end face height for all RF coaxial connectors: Where n is the number of RF coaxial connectors, This is the average end-face height of all RF coaxial connectors. This represents the end face height of a single RF coaxial connector. According to the formula above, A smaller value indicates better uniformity in the end face height of the RF coaxial connector on the antenna, with the end face approaching a uniform horizontal plane. Otherwise, it indicates that the end face height of the RF coaxial connector on the antenna fluctuates significantly, requiring maintenance to ensure the electrical performance of the radar antenna.

[0067] Step S700: Scan the point cloud data of other TR components in sequence, and repeat the calculation of the height value of the point cloud of the end face of the individual RF coaxial connector until the height value of the point cloud of the end face of all individual RF coaxial connectors is calculated.

[0068] In summary, this application has the following advantages:

[0069] Automated testing capabilities can be modularly embedded into the automated assembly process of phased array radar antennas, forming an integrated automated assembly unit for phased array antennas that combines assembly and testing, thereby significantly improving the assembly efficiency and consistency of radar antenna components.

[0070] Compared to traditional manual or coordinate measuring machine (CMM) methods, the use of structured light scanning combined with the algorithm of this invention can improve the efficiency of RF coaxial connector flushness detection by 5-10 times. It eliminates the need for frequent antenna adjustments due to hardware limitations, enabling in-situ measurement. Furthermore, its automated data analysis and calculations require no manual intervention, significantly reducing labor costs.

[0071] The entire scanning, analysis, and calculation of the antenna array can be fully automated by using a robotic arm equipped with a structured light scanner. At the same time, the flushness calculation of multiple types of antenna RF coaxial connectors can be achieved by changing a few parameters according to the type of antenna and RF coaxial connector, which has high compatibility.

[0072] Finally, it should be noted that the accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0073] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An automatic detection method for the flushness of RF coaxial connectors based on point cloud data, characterized in that, The method comprises the following steps: scanning to obtain a single TR component point cloud data of an antenna; adopting a statistical filtering algorithm to remove outliers in the TR component point cloud data; adopting a RANSAC algorithm to extract the plane data of the upper surface of the TR component point cloud data after removing the outliers, to obtain the normal vector of the plane data of the upper surface, and to remove the extracted plane data of the upper surface to obtain the point cloud data of the radio frequency coaxial connector; adopting a local plane detection algorithm based on RANSAC with prior shape constraint to extract all local plane features in the point cloud data of the radio frequency coaxial connector; extracting the normal vector in all local plane features, finding the local plane feature corresponding to the normal vector parallel to the normal vector of the plane data of the upper surface as the end surface point cloud of the radio frequency coaxial connector; segmenting the single radio frequency coaxial connector end surface point cloud by using the continuity of the point cloud data, and combining the plane data of the upper surface of the TR component point cloud data to calculate the height value of the single radio frequency coaxial connector end surface point cloud; sequentially scanning other TR component point cloud data, and repeating the calculation of the height value of the single radio frequency coaxial connector end surface point cloud until the calculation of the height value of all single radio frequency coaxial connector end surface point clouds is completed.

2. The method for automatic detection of the flushness of a radio frequency coaxial connector based on point cloud data according to claim 1, wherein, The method for extracting all local plane features in the point cloud data of the radio frequency coaxial connector is as follows: setting a sampling number, in each sampling number, randomly collecting 3 points, when the normal vectors of the 3 points are the same, fitting a plane; setting a first threshold, calculating the points within the range of the first threshold as the inliers of the plane; calculating the maximum connected inlier set as the point set supporting the plane, as the local plane feature; repeating the collection and extraction of the local plane feature until the set sampling number is reached, and all local plane features found are counted.

3. The method of claim 2, wherein the method further comprises: Number of samples wherein is the inlier ratio.

4. The method for automatic detection of the flushness of a radio frequency coaxial connector based on point cloud data according to claim 1, wherein, The method for obtaining the plane data of the upper surface of the TR component point cloud data is as follows: in the point cloud data of the radio frequency coaxial connector, randomly collecting 3 points, counting the distances from all points in the TR component point cloud data after removing outliers to the plane formed by the 3 points; setting a second distance threshold, taking the plane formed by a certain proportion of points within the second distance threshold as the target plane, judging the proportion of the number of points in the target plane to all points, if greater than 90%, the target plane is considered as the plane data of the upper surface of the TR component point cloud data.

5. The method for automatic detection of the flushness of a radio frequency coaxial connector based on point cloud data according to claim 1, wherein, The method for removing outliers in the TR component point cloud data is as follows: Calculate the neighborhood distance of each point in the single TR component point cloud data, count and calculate the neighborhood average distance of all points; calculate the mean of the neighborhood average distance of all points and standard deviation ; setting a neighborhood judgment threshold, judging whether the neighborhood average distance of a certain point exceeds the neighborhood judgment threshold, if yes, judging that the point is an outlier, and removing the point.

6. The method for automatic detection of the flushness of a radio frequency coaxial connector based on point cloud data according to claim 5, wherein, When calculating the neighborhood distance of each point, for each point in the point cloud The k-nearest neighbors are obtained by applying K-Nearest Neighbors to find the k points closest to it, and the distances between the k nearest neighbors and the point are calculated, denoted as where is a point in the neighborhood; The neighborhood average distance calculation formula is as follows: ; wherein, is the distance at point the corresponding neighborhood average distance; the mean of the neighborhood average distances and the standard deviation is: ， ； the calculation formula for judging whether the neighborhood average distance of a certain point exceeds the neighborhood judgment threshold is as follows: 。

Images