A method and system for identification detection of radio frequency connectors

Abstract

The application provides a kind of identification detection method and system for radio frequency connector, method includes: obtaining the initial appearance image data of radio frequency connector surface, and the first appearance image is obtained by processing image data;From the appearance feature set and identification area image in the first appearance image;The character segmentation and feature extraction are carried out to identification area image, and the preliminary manufacturing information text is obtained by combining the text recognition model established in advance, the preliminary manufacturing information text is checked and is associated with appearance feature set, and generates associated dataset;Appearance feature set is compared with preset standard appearance database, and generates defect mark data, and the integrated analysis data is generated by fusing associated dataset and defect mark data, and visual chart is generated according to integrated analysis data;According to the visual chart, further analyze performance index change and defect severity, generate risk assessment data, and generate final detection analysis file according to risk assessment data.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method and system for identifying and detecting radio frequency connectors. Background Technology

[0002] Radio frequency (RF) connectors are indispensable core components in modern communication and electronic equipment, playing a crucial role in ensuring signal transmission quality and system stability. Their performance directly impacts the operational efficiency and security of multiple key areas such as communication networks, aerospace, and medical equipment. Therefore, accurate testing and performance evaluation of RF connectors have become a top priority for the industry.

[0003] However, current mainstream testing methods largely rely on manual inspection and basic testing tools, which are not only inefficient but also susceptible to human error, making it difficult to guarantee the accuracy and consistency of test results. This traditional approach often falls short when faced with the complex and diverse types of connectors and the demands of large-scale production, failing to meet the high precision and efficiency requirements of modern industry.

[0004] Against this backdrop, the field faces significant technical challenges. The most pressing is how to rapidly acquire connector appearance features and manufacturing information through automation to replace the inefficient manual identification method. The failure to effectively address this issue further hinders the comprehensive coverage of connector geometric parameters and operating environment information during data acquisition, thus affecting the accurate analysis and evaluation of performance indicators. These interconnected challenges collectively constitute the technical bottlenecks in the intelligence and comprehensiveness of the testing system, urgently requiring innovative solutions.

[0005] Therefore, how to construct an automated method and system based on image recognition and data analysis to efficiently acquire the appearance features and manufacturing information of RF connectors has become a key research issue. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for identifying and detecting radio frequency connectors, enabling accurate detection and risk assessment of appearance defects in radio frequency connectors, and providing effective support for product quality control.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] A method for identifying and detecting radio frequency connectors, the method comprising the following steps:

[0009] Step 1: Acquire initial appearance image data of the RF connector surface using an image acquisition device, and process the initial appearance image data to obtain the first appearance image;

[0010] Step 2: Extract the set of appearance features and the image of the marked area from the first appearance image;

[0011] Step 3: Perform character segmentation and feature extraction on the image of the identified area, combine it with the pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate the first associated dataset;

[0012] Step 4: Compare the set of appearance features with the preset standard appearance database to generate defect marking data. Merge the first associated dataset with the defect marking data to generate comprehensive analysis data. Generate the first visualization chart based on the comprehensive analysis data to present the distribution and trend relationship of defects.

[0013] Step 5: Further analyze the changes in performance indicators and the severity of defects based on the first visualization chart, generate risk assessment data, and generate the final detection analysis file based on the risk assessment data for archiving and feedback.

[0014] Preferably, the step of acquiring initial appearance image data of the RF connector surface through an image acquisition device and processing the initial appearance image data to obtain a first appearance image includes:

[0015] The surface of the RF connector is scanned by a high-resolution image acquisition device to obtain initial appearance image data. For the initial appearance image data, illumination correction processing technology is used to adjust the areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured.

[0016] Based on the corrected initial appearance image data, determine whether the image clarity meets the preset requirements. If not, re-acquire and process the image, and obtain a first appearance image that meets the quality requirements through multiple processing iterations.

[0017] Preferably, the step of extracting the set of appearance features and the image of the marked region from the first appearance image includes:

[0018] For the first appearance image, the character region positioning method is used to identify the surface manufacturing information marking area, determine the marking area image, and perform region segmentation on the marking area image to separate the specific range of the manufacturing information marking, generating clear marking area image data;

[0019] Edge contour features are extracted from the first appearance image using edge detection technology. Based on the edge contour features, geometric feature data of surface defects are determined. The edge contour features and the geometric feature data are then integrated into an appearance feature set.

[0020] Based on the set of appearance features and the image data of clearly marked areas, a feature mapping relationship is established for subsequent defect comparison and text parsing.

[0021] Preferably, the step of performing character segmentation and feature extraction on the identified region image, combining it with a pre-established text recognition model to obtain preliminary manufacturing information text, verifying the preliminary manufacturing information text and associating it with the appearance feature set to generate a first associated dataset includes:

[0022] For the image of the identified region, a character segmentation algorithm is used to split the connected characters into individual character units. Then, character feature extraction technology is used to obtain the stroke width and spacing feature data of the individual character units. Based on the stroke width and spacing feature data, a pre-established text recognition model is used for classification and matching. Based on the classification and matching results, preliminary manufacturing information text is generated.

[0023] For the preliminary manufacturing information text, a text format verification method is used to verify whether the batch number and production date conform to the preset encoding rules. Based on the verification results, erroneous data that does not conform to the rules are removed. Based on the verified text data, a data association mapping technology is used to uniquely bind it to the set of appearance features. Based on the binding results, a first associated dataset is generated.

[0024] Preferably, the step of comparing the set of appearance features with a preset standard appearance database to generate defect marking data includes:

[0025] For the set of appearance features, a template comparison method is used to match it with a preset standard appearance database. Based on the matching results, the deviation value between the set of appearance features and the standard data is calculated. If the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, defect marking data is generated.

[0026] Preferably, the step of fusing the first associated dataset with the defect marker data to generate comprehensive analysis data, and generating a first visualization chart based on the comprehensive analysis data, includes:

[0027] For the first associated dataset and the defect marking data, multidimensional data overlay technology is used to perform feature matching between the appearance defect information in the defect marking data and the manufacturing information in the first associated dataset to determine the preliminary fused data structure. Through data structure analysis tools, field mapping processing is performed on the preliminary fused data structure to obtain intermediate layer data containing defect location and batch number. If there are missing fields or inconsistent formats in the intermediate layer data, the missing values ​​are filled in and the format is unified through data cleaning algorithms to determine the standardized data that meets the requirements. Based on the standardized data, a data integration framework is used to deeply associate the defect location and batch number to obtain the basic version of comprehensive analysis data. Through data verification mechanism, the completeness and accuracy of the basic version of comprehensive analysis data are checked to determine the reliable data set after it is error-free. If outliers or data redundancy are found in the reliable data set, records that do not conform to logic are removed through anomaly detection algorithms to obtain comprehensive analysis data.

[0028] Based on the comprehensive analysis data, a spatial distribution graph of surface defects of the RF connector is drawn using a defect distribution heatmap, and a batch correlation line graph is used to show the trend relationship between different batch numbers and defect rates. A first visualization chart is generated by combining the defect distribution heatmap and the batch correlation line graph.

[0029] Preferably, the step of further analyzing performance indicator changes and defect severity through visualization charts to generate risk assessment data includes:

[0030] For the first visualization chart, time series scatter plot technology is used to arrange the distribution of performance index changes by production date. A color coding method is used to distinguish the severity of defects by using different colors. Based on the marking results, a second visualization chart is generated. For the second visualization chart, dynamic interactive interface technology is used to allow users to view specific defect location information. A bar chart of performance indexes is used to present the comparison results between the predicted performance value and the preset threshold. If the predicted performance value is lower than the preset threshold, it is marked as a high-risk item. Based on the marking results, risk assessment data is generated.

[0031] Preferably, the step of generating the final detection and analysis file based on the risk assessment data includes:

[0032] Based on the risk assessment data, the defect distribution heatmap and batch correlation line chart are saved as vector format files using the chart export function. Based on the saved results, the final detection analysis file is generated.

[0033] For the final detection and analysis archive, an archive index is established to support fast retrieval. The integrity of the archive data is determined by verifying the final detection and analysis archive. If the integrity does not meet the requirements, the archive data is regenerated.

[0034] Based on the final detection and analysis files, a data backup mechanism is established to ensure data security. By classifying and managing the final detection and analysis files, the storage priority of the files is determined, and the storage location and access permissions of the files are adjusted according to the storage priority.

[0035] An identification and detection system applicable to the identification and detection method of radio frequency connectors, the identification and detection system comprising:

[0036] Image acquisition and preprocessing module: used to acquire initial appearance image data of the surface of the RF connector, and process the initial appearance image data to obtain a first appearance image;

[0037] Multimodal feature extraction and decoding module: used to extract the set of appearance features and the image of the marked region from the first appearance image;

[0038] Data fusion and verification module: used to perform character segmentation and feature extraction on the image of the identified area, and combine it with a pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate a first associated dataset;

[0039] Defect analysis and visualization decision module: used to compare the set of appearance features with a preset standard appearance database to generate defect marking data, and to merge the first associated dataset with the defect marking data to generate comprehensive analysis data, and generate a first visualization chart based on the comprehensive analysis data;

[0040] Risk assessment and record management module: used to further analyze the performance index changes and defect severity of the first visualization chart, generate risk assessment data, and generate the final detection analysis record based on the risk assessment data.

[0041] Preferably, the image acquisition and preprocessing module includes:

[0042] Image acquisition unit: scans the surface of the RF connector to acquire initial appearance image data;

[0043] Illumination correction preprocessing unit: For the initial appearance image data, illumination correction processing technology is used to adjust the areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured.

[0044] Image quality iterative optimization unit: verifies the clarity of the corrected initial appearance image data. If it does not meet the requirements, the image is re-acquired and processed. Through multiple processing iterations, a first appearance image with the required quality is obtained.

[0045] The multimodal feature extraction and decoding module includes:

[0046] Character region positioning unit: used to identify the manufacturing information marking area on the surface of the acquired first appearance image and determine the marking area image;

[0047] Edge detection and feature extraction unit: used to extract edge contour features from the first appearance image, determine the geometric feature data of surface defects, and integrate the edge contour features and the geometric feature data into an appearance feature set;

[0048] Region segmentation and boundary optimization unit: used to segment the identification region image to separate the specific range of the manufacturing information identification, and optimize the boundary of the identification region image to generate clear identification region image data;

[0049] Feature mapping relationship establishment unit: used to establish a feature mapping relationship between the set of appearance features and the image data of clearly marked areas, for subsequent defect comparison and text parsing;

[0050] The data fusion and verification module includes:

[0051] Character segmentation and feature advance unit: used to split connected characters into individual character units and obtain the stroke width and spacing feature data of the individual character units;

[0052] Text recognition and preliminary generation unit: used to classify and match the stroke width and spacing feature data of the acquired individual character units in combination with a pre-established text recognition model, and generate the preliminary manufacturing information text based on the classification and matching results;

[0053] Text format verification unit: used to verify whether the batch number and production date of the preliminary manufacturing information text conform to the preset coding rules, and to remove erroneous data that does not conform to the rules based on the verification results;

[0054] Cross-modal data association unit: uniquely binds the verified text data to the appearance feature set, and generates the first associated dataset based on the binding result;

[0055] The defect analysis and visualization decision-making module includes:

[0056] Standard comparison and defect detection unit: used to match the set of appearance features with a preset standard appearance database, calculate the deviation value between the set of appearance features and the standard data based on the matching result, and if the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, the defect marking data is generated.

[0057] Multidimensional data fusion unit: used to fuse information between the first associated dataset and the defect marking data using multidimensional data overlay technology, and to associate the appearance defect information with the manufacturing information through the fusion result, generating the comprehensive analysis data containing the defect location and batch number;

[0058] Visualization chart generation unit: used to draw the spatial distribution of defects on the surface of RF connectors using a defect distribution heatmap, and to show the trend relationship between different batch numbers and defect rates using a batch correlation line chart, generating the first visualization chart;

[0059] The risk assessment and record management module includes:

[0060] Time series visualization analysis unit: Used to arrange the distribution of performance index changes by production date using time series scatter plot technology on the first visualization chart, and to mark the severity of defects with different colors using a color coding method. Based on the marking results, a second visualization chart is generated.

[0061] Interactive Defect Location Unit: Used to support users in viewing specific defect location information using dynamic interactive interface technology on the second visualization chart, and to present the comparison results between the predicted performance value and the preset threshold through a performance index bar chart. If the predicted performance value is lower than the preset threshold, it is marked as a high-risk item, and the risk assessment data is generated based on the marking results.

[0062] Multi-format chart export unit: This unit exports risk assessment data to vector format files, including defect distribution heatmaps and batch correlation line graphs. Based on the exported data, it generates a final inspection and analysis archive. An index is then created for this final inspection and analysis archive to support rapid retrieval.

[0063] Intelligent record management unit: For the final inspection and analysis records, a record index is established to support fast retrieval. The integrity of the record data is determined by verifying the final inspection and analysis records. If the integrity does not meet the requirements, the record data is regenerated.

[0064] Enterprise-level data governance unit: Used to establish a data backup mechanism for the final test and analysis files to ensure data security, as well as to classify and manage the final test and analysis files, determine the storage priority of the files, and adjust the storage location and access permissions of the files according to the storage priority.

[0065] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0066] This invention discloses a method for detecting and assessing the appearance defects of radio frequency connectors. It acquires an initial appearance image using a high-resolution image acquisition device, obtains a first appearance image after illumination correction, identifies the surface manufacturing information marking area using a character region localization method, extracts appearance features, and obtains the manufacturing information text by combining a character segmentation algorithm and a text recognition model. The verified text is uniquely bound to the appearance feature set to obtain a first associated dataset. This invention utilizes a template comparison method to detect appearance defects, generates defect marking data, and integrates appearance defect information with manufacturing information to form comprehensive analysis data. The defect distribution and trends are presented through visualization methods such as defect distribution heatmaps and batch correlation line graphs, ultimately generating risk assessment data. This method achieves accurate detection and risk assessment of appearance defects in radio frequency connectors, providing effective support for product quality control. Attached Figure Description

[0067] Figure 1 This is a flowchart of the identification and detection method for radio frequency connectors according to the present invention;

[0068] Figure 2 This is a structural block diagram of the identification and detection system for radio frequency connectors according to the present invention. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0070] like Figure 1 This embodiment provides a method for identifying and detecting radio frequency connectors, which includes the following steps:

[0071] Step 1: Acquire initial appearance image data of the RF connector surface using an image acquisition device, and process the initial appearance image data to obtain a first appearance image;

[0072] Step 2: Extract the set of appearance features and the image of the marked area from the first appearance image;

[0073] Step 3: Perform character segmentation and feature extraction on the image of the identified area, combine it with the pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate the first associated dataset;

[0074] Step 4: Compare the set of appearance features with a preset standard appearance database to generate defect marking data. Merge the first associated dataset with the defect marking data to generate comprehensive analysis data. Generate a first visualization chart based on the comprehensive analysis data to present the distribution and trend relationship of defects.

[0075] Step 5: Further analyze the changes in performance indicators and the severity of defects based on the first visualization chart, generate risk assessment data, and generate a final detection analysis file based on the risk assessment data for archiving and feedback.

[0076] In one embodiment, acquiring initial appearance image data of the RF connector surface through an image acquisition device and processing the initial appearance image data to obtain a first appearance image includes:

[0077] The surface of the RF connector is scanned by a high-resolution image acquisition device to obtain initial appearance image data. For the initial appearance image data, illumination correction processing technology is used to adjust the areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured.

[0078] Based on the corrected initial appearance image data, determine whether the image clarity meets the preset requirements. If not, re-acquire and process the image, and obtain a first appearance image that meets the quality requirements through multiple processing iterations.

[0079] For example, during image acquisition, a high-resolution image acquisition device can scan the surface of the RF connector at a resolution of 50 million pixels, and a linear CCD sensor can be used to acquire initial appearance image data with an accuracy of 0.05mm. For the initial appearance image data, the MSRCR algorithm based on Retinex theory is used for illumination correction. The Gaussian kernel scale parameter is set to [15,80,200], and the brightness uneven area is adjusted until the gray value deviation is less than 5%. After the correction is completed, the first appearance image is obtained.

[0080] In one embodiment, extracting the set of appearance features and the image of the identified region from the first appearance image includes:

[0081] For the first appearance image, the character region positioning method is used to identify the surface manufacturing information marking area, determine the marking area image, and perform region segmentation on the marking area image to separate the specific range of the manufacturing information marking, generating clear marking area image data;

[0082] In this embodiment, when extracting the identification area image, a character region localization method based on a convolutional neural network (CNN) can be used to extract features from the first appearance image, identify the surface manufacturing information identification area, and determine the identification area image. A threshold-based image segmentation technique is then used to separate the text portion from the identification area image, obtaining clear identification area image data. When using the threshold-based image segmentation technique, a global thresholding method can be used, where a threshold is manually selected based on experience by observing the image's grayscale histogram. For example, if the grayscale values ​​of most text pixels in the image are concentrated between 100 and 200, and the grayscale values ​​of background pixels are concentrated between 0 and 50, a threshold of 75 can be selected. Alternatively, a local thresholding method, such as an adaptive thresholding method, can be used, which can achieve adaptive threshold segmentation by calculating the threshold based on the characteristics of local regions of the image. The Tesseract optical character recognition engine is then used to parse the clear identification area image data and extract the manufacturing information text data, including information such as model number and batch number.

[0083] Edge contour features are extracted from the first appearance image using edge detection technology. Based on the edge contour features, geometric feature data of surface defects are determined. The edge contour features and the geometric feature data are then integrated into an appearance feature set.

[0084] Based on the set of appearance features and the image data of clearly marked areas, a feature mapping relationship is established for subsequent defect comparison and text parsing.

[0085] In one embodiment, the image of the identified region is segmented and its features are extracted. A preliminary manufacturing information text is obtained by combining this text with a pre-established text recognition model. The preliminary manufacturing information text is then verified and associated with the appearance feature set to generate a first associated dataset, including:

[0086] For the image of the identified region, a character segmentation algorithm is used to split the connected characters into individual character units. Then, character feature extraction technology is used to obtain the stroke width and spacing feature data of the individual character units. Based on the stroke width and spacing feature data, a pre-established text recognition model is used for classification and matching. Based on the classification and matching results, the preliminary manufacturing information text is generated.

[0087] In this embodiment, a character segmentation algorithm based on projection is used to split connected characters. For example, vertical projection analysis is performed on the image, and a threshold of 5 pixels is set to segment the characters into individual character units. Feature extraction is performed on each individual character unit, and an edge detection algorithm is used to obtain stroke width and spacing feature data. For example, the Canny algorithm is used to extract edges, calculating an average stroke width of 3 pixels and an average spacing of 2 pixels. A pre-established text recognition model based on convolutional neural networks is used to classify and match the extracted stroke width and spacing feature data. For example, a ResNet model is used to classify the characters, resulting in the preliminary manufacturing information text "2023-09-15-B123".

[0088] For the preliminary manufacturing information text, a text format verification method is used to verify whether the batch number and production date conform to the preset encoding rules. Based on the verification results, erroneous data that does not conform to the rules are removed. Based on the verified text data, a data association mapping technology is used to uniquely bind it to the appearance feature set. Based on the binding results, the first associated dataset is generated.

[0089] In this embodiment, a regular expression validation method is used to verify the batch number and production date in the preliminary manufacturing information text. For example, the regular expression "^\d{4}-\d{2}-\d{2}-B\d{3}" is used. "It is determined whether the batch number or production date conforms to the preset encoding rules. If the verification result shows that the batch number or production date does not conform to the preset encoding rules, such as "2023-09-15-B12", the erroneous data is removed, and the verified text data "2023-09-15-B123" is obtained. Through data association mapping technology, the verified text data is uniquely bound to the set of appearance features, for example, by using a hash algorithm to generate a unique identifier, thus obtaining the first associated dataset."

[0090] In one embodiment, the set of appearance features is compared with a preset standard appearance database to generate defect marking data, including:

[0091] For the set of appearance features, a template comparison method is used to match it with a preset standard appearance database. Based on the matching result, the deviation value between the set of appearance features and the standard data is calculated. If the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, the defect marking data is generated.

[0092] In this embodiment, a template comparison method based on Euclidean distance is used to match the set of appearance features with a standard appearance database. For example, the Euclidean distance is calculated to be 0.8, and the comparison deviation value is obtained. If the comparison deviation value exceeds a preset threshold of 0.5, it is judged as an appearance defect, and defect marker data "Defect type: scratch, location: upper right corner" is generated.

[0093] By integrating the defect marker data with the first associated dataset and processing it using a decision tree-based comprehensive analysis algorithm, the final manufacturing information and defect analysis results are obtained: "Batch: B123, Production Date: 2023-09-15, Defect: Scratch".

[0094] In one embodiment, the step of fusing the first associated dataset with the defect marker data to generate comprehensive analysis data, and generating a first visualization chart based on the comprehensive analysis data, includes:

[0095] For the first associated dataset and the defect marking data, multidimensional data overlay technology is used to perform feature matching between the appearance defect information in the defect marking data and the manufacturing information in the first associated dataset to determine the preliminary fused data structure. Through data structure analysis tools, field mapping processing is performed on the preliminary fused data structure to obtain intermediate layer data containing defect location and batch number. If there are missing fields or inconsistent formats in the intermediate layer data, the missing values ​​are filled in and the format is unified through data cleaning algorithms to determine the standardized data that meets the requirements. Based on the standardized data, a data integration framework is used to deeply associate the defect location and batch number to obtain the basic version of comprehensive analysis data. Through data verification mechanism, the completeness and accuracy of the basic version of comprehensive analysis data are checked to determine the reliable data set after it is error-free. If outliers or data redundancy are found in the reliable data set, records that do not conform to logic are removed through anomaly detection algorithms to obtain comprehensive analysis data.

[0096] In this example, when fusing the first associated dataset with the defect marker data, the original data combination containing appearance defect information and preliminary manufacturing information is extracted from both datasets. For example, manufacturing records with batch numbers B20231001 to B20231010 and their corresponding defect marker data are retrieved from the database. Based on the original data combination, multidimensional data overlay technology is used to perform feature matching between the appearance defect information in the defect marker data and the manufacturing information in the first associated dataset. For example, the KNN algorithm is used to associate the defect location coordinates with the manufacturing batch number to determine the preliminary fused data structure. Data structure analysis tools are then used to perform field mapping processing on the preliminary fused data structure. For example, the defect location field is mapped to latitude and longitude coordinates, and the batch number field is mapped to the production date, resulting in intermediate layer data containing defect location and batch number. If there are missing fields or inconsistent formats in the intermediate layer data, data cleaning algorithms are used to fill in the missing values ​​and standardize the format. For example, the mean imputation method is used to complete the missing defect location data, and standardized data that meets the requirements is identified. Based on the standardized data, a data integration framework is used to deeply associate the defect location with the batch number. For example, association rule mining algorithms are used to analyze the correlation between the defect location and the batch number to obtain the basic version of the comprehensive analysis data. The basic version of the comprehensive analysis data is checked for completeness and accuracy through a data verification mechanism. For example, the CRC check algorithm is used to verify the integrity of the data to determine the set of reliable data that is error-free. If outliers or data redundancy are found in the reliable data set, records that do not conform to logic are removed using anomaly detection algorithms. For example, the isolated forest algorithm is used to identify and remove abnormal batch data to obtain the comprehensive analysis data.

[0097] Based on the comprehensive analysis data, a spatial distribution analysis algorithm is used to divide the defect locations into grids, obtaining preliminary spatial data of defect distribution. Thermal values ​​are calculated from the preliminary spatial data, and a defect distribution heatmap is generated using color mapping technology to identify densely populated defect areas on the RF connector surface. Based on the results of the defect distribution heatmap, the statistical value of the number of defects in each grid area is obtained, yielding quantitative index data of spatial distribution. Defect rate data for different batch numbers is collected, and a time series analysis method is used to organize the relationship between batches and defect rates, obtaining batch-related trend data. Based on the batch-related trend data, a batch-related line chart is generated using line chart plotting technology to determine the trend characteristics of defect rate changes with batches. Finally, the defect distribution heatmap and the batch-related line chart are integrated into a single view, and a first visualization chart is generated using multi-layer overlay technology.

[0098] In this embodiment, a spatial distribution analysis algorithm is used to divide the defect locations into a grid, dividing the surface into 10×10 grid cells to obtain preliminary spatial data of the defect distribution. Thermal values ​​are then calculated from this preliminary spatial data, and a Gaussian kernel density estimation algorithm is used to calculate the defect density value for each grid cell. Color mapping technology is then used to map the density values ​​to a gradient heatmap from blue to red, identifying the defect-dense areas on the RF connector surface. Based on the results of the defect distribution heatmap, the defect quantity statistics for each grid region are obtained; for example, grid A1 has 15 defects, and grid B2 has 8 defects. The quantitative index data of spatial distribution is obtained. By collecting defect rate data of different batch numbers, the relationship between batch and defect rate is organized using time series analysis. For example, the defect rate of batch 001 is 2.5% and the defect rate of batch 002 is 3.8%, and the trend data of batch association is obtained. Based on the trend data of batch association, a batch association line chart is generated using line chart drawing technology. The horizontal axis is the batch number and the vertical axis is the defect rate. The trend characteristics of defect rate change with batch are determined. By integrating the defect distribution heat map and the batch association line chart, the first visualization chart is generated using multi-layer overlay technology.

[0099] In one embodiment, the step of further analyzing performance indicator changes and defect severity through visualization charts to generate risk assessment data includes:

[0100] The process involves acquiring the production date and corresponding performance index data of the RF connectors to obtain a preliminary time-series dataset. Based on this dataset, a time-series scatter plot technique is used to arrange the distribution of performance index changes in chronological order of production date, generating a basic visualization. A defect detection algorithm is then used to analyze the defect severity data for each data point, obtaining a classification result for defect severity. Based on this classification, a color-coding method is used to assign different colors to each data point, resulting in a color-coded time-series scatter plot. A data integration module then associates the color-coded scatter plot with the original performance index data to generate a second visualization chart. Based on this second visualization chart, data storage technology is used to bind each data point in the chart with its corresponding detailed defect information, resulting in an interactive data structure. This interactive data structure is rendered using a dynamic interactive interface, allowing users to extract and display specific defect location information when the chart is clicked. User query results are then retrieved, and a performance index bar chart generation technique is used to create a comparison graph between the predicted performance value and a preset threshold, resulting in a comparative analysis chart. If the predicted performance value is lower than the preset threshold, a risk labeling algorithm is used to mark the corresponding data point as a high-risk item, determining the final risk assessment data.

[0101] In this embodiment, the production date and corresponding performance indicators of the RF connector, such as insertion loss (range 0.5dB-2.0dB) and voltage standing wave ratio (range 1.2-1.8), are extracted to form a CSV format dataset containing timestamps and numerical values. The time-series scatter plot technique utilizes Python's Matplotlib library to plot a scatter plot with the production date as the horizontal axis and the performance indicators as the vertical axis. The horizontal axis interval is set to 7 days, and the vertical axis is divided into 0.1dB precision levels. The defect detection algorithm uses the YOLOv5 model to analyze the quality inspection image associated with each data point, outputting a defect severity level (0-no defect, 1-minor, 2-moderate, 3-severe), with an accuracy of 98%. The color coding module maps the level results to green (level 0), yellow (level 1), orange (level 2), and red (level 3), and uses O... The penCV library modifies the color channel values ​​of the scatter plot. The data integration module uses Pandas to merge color-marked data with the original performance metrics, generating a JSON structure containing RGB fields. The data storage technology uses MongoDB to build a document-oriented database, with the production date as the primary key to associate defect coordinates (e.g., X: 12.5mm, Y: 8.3mm) and the original optical inspection data. The dynamic interactive interface is developed based on D3.js, which listens for mouse click events to trigger an AJAX request to retrieve a 200dpi microscopic image of the corresponding coordinates from the database. The performance metric bar chart is generated using Echarts, with a preset threshold of 1.5dB insertion loss. If the predicted value of 1.7dB exceeds the threshold, the risk labeling algorithm is called to update the risk_flag field to 1 in MySQL and trigger an alarm push on WeChat.

[0102] In one embodiment, generating the final detection analysis file based on the risk assessment data includes:

[0103] Based on the risk assessment data, the defect distribution heatmap and batch correlation line chart are saved as vector format files using the chart export function. Based on the saved results, the final detection analysis file is generated.

[0104] For the final detection and analysis archive, an archive index is established to support fast retrieval. The integrity of the archive data is determined by verifying the final detection and analysis archive. If the integrity does not meet the requirements, the archive data is regenerated.

[0105] Based on the final detection and analysis files, a data backup mechanism is established to ensure data security. By classifying and managing the final detection and analysis files, the storage priority of the files is determined, and the storage location and access permissions of the files are adjusted according to the storage priority.

[0106] See Figure 2 As shown, an identification and detection system for an identification and detection method of an RF connector is provided, the identification and detection system comprising:

[0107] Image acquisition and preprocessing module: used to acquire initial appearance image data of the RF connector surface, and process the initial appearance image data to obtain a first appearance image; the image acquisition and preprocessing module includes:

[0108] Image acquisition unit: scans the surface of the RF connector to acquire initial appearance image data;

[0109] Illumination correction preprocessing unit: For the initial appearance image data, illumination correction processing technology is used to adjust the areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured.

[0110] Image quality iterative optimization unit: verifies the clarity of the corrected initial appearance image data. If it does not meet the requirements, the image is re-acquired and processed. Through multiple processing iterations, a first appearance image with the required quality is obtained.

[0111] Multimodal feature extraction and decoding module: used to extract a set of appearance features and an image of the marked region from the first appearance image; the multimodal feature extraction and decoding module includes:

[0112] Character region positioning unit: used to identify the manufacturing information marking area on the surface of the acquired first appearance image and determine the marking area image;

[0113] Edge detection and feature extraction unit: used to extract edge contour features from the first appearance image, determine the geometric feature data of surface defects, and integrate the edge contour features and the geometric feature data into an appearance feature set;

[0114] Region segmentation and boundary optimization unit: used to segment the identification region image to separate the specific range of the manufacturing information identification, and optimize the boundary of the identification region image to generate clear identification region image data;

[0115] Feature mapping relationship establishment unit: used to establish a feature mapping relationship between the set of appearance features and the image data of clearly marked areas, for subsequent defect comparison and text parsing.

[0116] Data fusion and verification module: used to perform character segmentation and feature extraction on the image of the identified area, and combine it with a pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate a first associated dataset; the data fusion and verification module includes:

[0117] Character segmentation and feature advance unit: used to split connected characters into individual character units and obtain the stroke width and spacing feature data of the individual character units;

[0118] Text recognition and preliminary generation unit: used to classify and match the stroke width and spacing feature data of the acquired individual character units in combination with a pre-established text recognition model, and generate the preliminary manufacturing information text based on the classification and matching results;

[0119] Text format verification unit: used to verify whether the batch number and production date of the preliminary manufacturing information text conform to the preset coding rules, and to remove erroneous data that does not conform to the rules based on the verification results;

[0120] Cross-modal data association unit: uniquely binds the verified text data to the appearance feature set, and generates the first associated dataset based on the binding result.

[0121] The defect analysis and visualization decision module is used to compare the set of appearance features with a preset standard appearance database to generate defect marking data, and to fuse the first associated dataset with the defect marking data to generate comprehensive analysis data, and to generate a first visualization chart based on the comprehensive analysis data; the defect analysis and visualization decision module includes:

[0122] Standard comparison and defect detection unit: used to match the set of appearance features with a preset standard appearance database, calculate the deviation value between the set of appearance features and the standard data based on the matching result, and if the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, the defect marking data is generated.

[0123] Multidimensional data fusion unit: used to fuse information between the first associated dataset and the defect marking data using multidimensional data overlay technology, and to associate the appearance defect information with the manufacturing information through the fusion result, generating the comprehensive analysis data containing the defect location and batch number;

[0124] Visualization chart generation unit: used to draw the spatial distribution of defects on the surface of RF connectors using a defect distribution heatmap, and to display the trend relationship between different batch numbers and defect rates using a batch correlation line chart, generating the first visualization chart.

[0125] Risk assessment and record management module: used to further analyze performance indicator changes and defect severity from the first visualization chart, generate risk assessment data, and generate the final detection analysis record based on the risk assessment data; the risk assessment and record management module includes:

[0126] Time series visualization analysis unit: Used to arrange the distribution of performance index changes by production date using time series scatter plot technology on the first visualization chart, and to mark the severity of defects with different colors using a color coding method. Based on the marking results, a second visualization chart is generated.

[0127] Interactive Defect Location Unit: Used to support users in viewing specific defect location information using dynamic interactive interface technology on the second visualization chart, and to present the comparison results between the predicted performance value and the preset threshold through a performance index bar chart. If the predicted performance value is lower than the preset threshold, it is marked as a high-risk item, and the risk assessment data is generated based on the marking results.

[0128] Multi-format chart export unit: This unit exports risk assessment data to vector format files, including defect distribution heatmaps and batch correlation line graphs. Based on the exported data, it generates a final inspection and analysis archive. An index is then created for this final inspection and analysis archive to support rapid retrieval.

[0129] Intelligent record management unit: For the final inspection and analysis records, a record index is established to support fast retrieval. The integrity of the record data is determined by verifying the final inspection and analysis records. If the integrity does not meet the requirements, the record data is regenerated.

[0130] Enterprise-level data governance unit: Used to establish a data backup mechanism for the final test and analysis files to ensure data security, as well as to classify and manage the final test and analysis files, determine the storage priority of the files, and adjust the storage location and access permissions of the files according to the storage priority.

[0131] The comprehensive application of the above systems has greatly improved detection efficiency, reduced interference from human factors, and significantly enhanced the accuracy and reliability of detection results.

[0132] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identification detection of a radio frequency connector, characterized in that, The method includes the following steps: Step 1: Acquire initial appearance image data of the RF connector surface using an image acquisition device, and process the initial appearance image data to obtain the first appearance image; Step 2: Extract the set of appearance features and the identification area image from the first appearance image, including: for the first appearance image, use the character region positioning method to identify the surface manufacturing information identification area, determine the identification area image, and perform region segmentation on the identification area image to separate the specific range of the manufacturing information identification, and generate clear identification area image data. Edge contour features are extracted from the first appearance image using edge detection technology. Based on the edge contour features, the geometric feature data of the surface defects are determined. The edge contour features and geometric feature data are then integrated into an appearance feature set. Based on the set of appearance features and the image data of clearly marked areas, a feature mapping relationship is established for subsequent defect comparison and text parsing. Step 3: Perform character segmentation and feature extraction on the identified region image, combine it with a pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate the first associated dataset, including: For the image of the identified region, a character segmentation algorithm is used to break down connected characters into individual character units. Then, character feature extraction technology is used to obtain the stroke width and spacing feature data of each individual character unit. Based on the stroke width and spacing feature data, a pre-established text recognition model is used for classification and matching. Based on the classification and matching results, preliminary manufacturing information text is generated. For the initial manufacturing information text, a text format verification method is used to verify whether the batch number and production date conform to the preset coding rules. Based on the verification results, erroneous data that does not conform to the rules is removed. Based on the verified text data, a data association mapping technology is used to uniquely bind it to the appearance feature set. Based on the binding results, the first associated dataset is generated. Step 4: Compare the set of appearance features with the preset standard appearance database to generate defect marking data. Merge the first associated dataset with the defect marking data to generate comprehensive analysis data. Generate the first visualization chart based on the comprehensive analysis data to present the distribution and trend relationship of defects. Step 5: Further analyze the changes in performance indicators and the severity of defects based on the first visualization chart, generate risk assessment data, and generate the final detection analysis file based on the risk assessment data for archiving and feedback.

2. The identification and detection method for radio frequency connectors as described in claim 1, characterized in that, The process of acquiring initial appearance image data of the RF connector surface through an image acquisition device and processing the initial appearance image data to obtain a first appearance image includes: The surface of the RF connector is scanned using a high-resolution image acquisition device to obtain initial appearance image data. For the initial appearance image data, illumination correction processing technology is used to adjust areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured. Based on the corrected initial appearance image data, determine whether the image clarity meets the preset requirements. If not, re-acquire and process the image, and obtain a first appearance image that meets the quality requirements through multiple processing iterations.

3. The identification and detection method for radio frequency connectors as described in claim 1, characterized in that, The step of comparing the set of appearance features with a preset standard appearance database to generate defect marking data includes: For the set of appearance features, a template comparison method is used to match it with a preset standard appearance database. Based on the matching results, the deviation value between the set of appearance features and the standard data is calculated. If the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, defect marking data is generated.

4. The identification and detection method for radio frequency connectors as described in claim 3, characterized in that, The process of fusing the first associated dataset with defect marker data to generate comprehensive analysis data, and generating a first visualization chart based on the comprehensive analysis data, includes: For the first associated dataset and the defect marking data, multidimensional data overlay technology is used to fuse information. Through the fusion result, the appearance defect information is associated with the manufacturing information. Based on the association result, comprehensive analysis data containing defect location and batch number is generated. Based on the comprehensive analysis data, a spatial distribution graph of surface defects of the RF connector is drawn using a defect distribution heatmap. A batch correlation line graph is used to show the trend relationship between different batch numbers and defect rates. The first visualization chart is generated by combining the defect distribution heatmap and the batch correlation line graph.

5. The identification and detection method for radio frequency connectors as described in claim 4, characterized in that, The step involves further analyzing performance indicator changes and defect severity based on the first visualization chart to generate risk assessment data, including: For the first visualization chart, time series scatter plot technology is used to arrange the distribution of performance index changes by production date. Different colors are used to mark the severity of defects through color coding. Based on the marking results, a second visualization chart is generated. For the second visualization chart, dynamic interactive interface technology is used to allow users to view specific defect location information. The performance index bar chart presents the comparison results between the predicted performance value and the preset threshold. If the predicted performance value is lower than the preset threshold, it is marked as a high-risk item. Based on the marking results, risk assessment data is generated.

6. The identification and detection method for radio frequency connectors as described in claim 5, characterized in that, The process of generating the final detection and analysis file based on the risk assessment data includes: For the risk assessment data, the defect distribution heatmap and batch correlation line chart are saved as vector format files using the chart export function. Based on the saved results, the final inspection and analysis file is generated. For the final test and analysis archives, an archive index is established to support fast retrieval. The integrity of the archive data is determined by verifying the final test and analysis archives. If the integrity does not meet the requirements, the archive data is regenerated. Based on the final test and analysis files, a data backup mechanism is established to ensure data security. By classifying and managing the final test and analysis files, the storage priority of the files is determined, and the storage location and access permissions of the files are adjusted according to the storage priority.

7. An identification and detection system applicable to the identification and detection method for radio frequency connectors according to any one of claims 1-6, characterized in that, The identification and detection system includes: Image acquisition and preprocessing module: used to acquire initial appearance image data of the surface of the RF connector, and process the initial appearance image data to obtain a first appearance image; Multimodal feature extraction and decoding module: used to extract the set of appearance features and the image of the marked region from the first appearance image; Data fusion and verification module: used to perform character segmentation and feature extraction on the image of the identified area, and combine it with a pre-established text recognition model to obtain preliminary manufacturing information text, verify the preliminary manufacturing information text and associate it with the appearance feature set to generate a first associated dataset; Defect analysis and visualization decision module: used to compare the set of appearance features with a preset standard appearance database to generate defect marking data, and to merge the first associated dataset with the defect marking data to generate comprehensive analysis data, and generate a first visualization chart based on the comprehensive analysis data; Risk assessment and record management module: used to further analyze the performance index changes and defect severity of the first visualization chart, generate risk assessment data, and generate the final detection analysis record based on the risk assessment data.

8. The identification and detection system for an identification and detection method for radio frequency connectors as described in claim 7, characterized in that, The image acquisition and preprocessing module includes: Image acquisition unit: scans the surface of the RF connector to acquire initial appearance image data; Illumination correction preprocessing unit: For the initial appearance image data, illumination correction processing technology is used to adjust the areas with uneven brightness. By correcting the brightness of the initial appearance image data, the uniformity of the image data is ensured. Image quality iterative optimization unit: verifies the clarity of the corrected initial appearance image data. If it does not meet the requirements, the image is re-acquired and processed. Through multiple processing iterations, a first appearance image with the required quality is obtained. The multimodal feature extraction and decoding module includes: Character region positioning unit: used to identify the manufacturing information marking area on the surface of the acquired first appearance image and determine the marking area image; Edge detection and feature extraction unit: used to extract edge contour features from the first appearance image, determine the geometric feature data of surface defects, and integrate the edge contour features and the geometric feature data into an appearance feature set; Region segmentation and boundary optimization unit: used to segment the identification region image to separate the specific range of the manufacturing information identification, and optimize the boundary of the identification region image to generate clear identification region image data; Feature mapping relationship establishment unit: used to establish a feature mapping relationship between the set of appearance features and the image data of clearly marked areas, for subsequent defect comparison and text parsing; The data fusion and verification module includes: Character segmentation and feature advance unit: used to split connected characters into individual character units and obtain the stroke width and spacing feature data of the individual character units; Text recognition and preliminary generation unit: This unit is used to classify and match the stroke width and spacing feature data of the acquired individual character units with a pre-established text recognition model, and generate preliminary manufacturing information text based on the classification and matching results. Text format verification unit: used to verify whether the batch number and production date of the preliminary manufacturing information text conform to the preset coding rules, and to remove erroneous data that does not conform to the rules based on the verification results; Cross-modal data association unit: uniquely binds the verified text data to the appearance feature set, and generates a first association dataset based on the binding result; The defect analysis and visualization decision-making module includes: Standard comparison and defect detection unit: used to match the set of appearance features with a preset standard appearance database, calculate the deviation value between the set of appearance features and the standard data based on the matching result, and if the deviation value exceeds a preset threshold, it is judged as an appearance defect. Based on the judgment result, defect marking data is generated. Multidimensional data fusion unit: used to fuse information between the first associated dataset and the defect marking data using multidimensional data overlay technology, and to associate the appearance defect information with the manufacturing information through the fusion result, generating comprehensive analysis data containing defect location and batch number; Visualization chart generation unit: used to draw the spatial distribution of defects on the surface of RF connectors using a defect distribution heatmap, and to show the trend relationship between different batch numbers and defect rates using a batch correlation line chart, generating the first visualization chart; The risk assessment and record management module includes: Time series visualization analysis unit: Used to arrange the distribution of performance index changes by production date using time series scatter plot technology on the first visualization chart, and to mark the severity of defects with different colors using a color coding method. Based on the marking results, a second visualization chart is generated. Interactive Defect Location Unit: Used to support users in viewing specific defect location information using dynamic interactive interface technology on the second visualization chart, and presents the comparison results of predicted performance value and preset threshold through performance index bar chart. If the predicted performance value is lower than the preset threshold, it is marked as a high-risk item. Based on the marking results, risk assessment data is generated. Multi-format chart export unit: This unit exports risk assessment data to vector format files, including defect distribution heatmaps and batch correlation line graphs. Based on the exported data, it generates a final inspection and analysis archive. An index is then created for this final inspection and analysis archive to support rapid retrieval. Intelligent record management unit: For the final inspection and analysis records, a record index is established to support fast retrieval. The integrity of the record data is determined by verifying the final inspection and analysis records. If the integrity does not meet the requirements, the record data is regenerated. Enterprise-level data governance unit: Used to establish a data backup mechanism for the final test and analysis files to ensure data security, as well as to classify and manage the final test and analysis files, determine the storage priority of the files, and adjust the storage location and access permissions of the files according to the storage priority.

Images