Dynamic production line measurement and control method and system based on visual recognition

Abstract

The present invention relates to the field of visual recognition technology, and in particular to a dynamic production line measurement and control method and system based on visual recognition. The method comprises the following steps: collecting production line product data through visual recognition equipment to obtain production line product image data; performing dynamic calibration based on production line product image data to obtain product image calibration data; performing product status analysis based on product image calibration data to obtain product status data; performing production line adjustment processing based on product status data to obtain production line adjustment data to perform auxiliary operations for production line parameter adjustment. The present invention improves the automation level of the production line, reduces manual intervention, improves the accuracy and consistency of product quality control, and realizes dynamic adjustment and optimization through a closed-loop feedback system.

Description

Technical Field

[0001] The present invention relates to the field of visual recognition technology, and in particular to a dynamic production line measurement and control method and system based on visual recognition. Background Art

[0002] With the rapid development of intelligent manufacturing technology, the automation level of production lines is constantly improving, especially in industries such as high-precision manufacturing, electronic manufacturing and automotive parts, where the requirements for product quality are becoming more and more stringent. Traditional production line control and quality inspection methods mostly rely on manual inspection and mechanical sensors, which not only have limited detection accuracy, but also have low efficiency and are easily affected by human errors and environmental changes. They cannot meet the requirements of modern industrial production for high precision, real-time monitoring and full automation.

[0003] The introduction of visual recognition technology has provided new possibilities for the intelligent and automated management of production lines. The production line inspection system based on computer vision and image processing technology can realize non-contact, high-precision inspection of products. It can not only detect the shape and size of products, but also identify surface defects of products and detect internal structures. At the same time, the visual recognition system can adapt to the complex environmental changes of the production line, such as lighting conditions and product position offset, through dynamic calibration and automatic adjustment, which significantly improves the reliability of inspection and production efficiency. In the production environment, due to the influence of factors such as lighting, angle and equipment vibration, the visual inspection system is easily disturbed, resulting in a decrease in inspection accuracy. In addition, when the product shape is complex or there is deformation or occlusion, it is difficult for traditional visual inspection methods to achieve stable and high-precision inspection. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a dynamic production line measurement and control method based on visual recognition to solve at least one of the above technical problems.

[0005] The present application provides a dynamic production line measurement and control method based on visual recognition, comprising the following steps:

[0006] Step S1: collect production line product data through visual recognition equipment to obtain production line product image data;

[0007] Step S2: Perform dynamic calibration according to the product image data of the production line to obtain product image calibration data;

[0008] Step S3: Perform product status analysis based on the product image calibration data to obtain product status data;

[0009] Step S4: Perform production line adjustment processing according to the product status data to obtain production line adjustment data to perform auxiliary operations for production line parameter adjustment.

[0010] The visual recognition in the present invention belongs to non-contact detection, which reduces the physical interference to the product and is suitable for high-speed assembly line operation. By dynamically calibrating the product image, the deviation caused by mechanical equipment errors or production line jitter can be automatically detected and corrected to ensure production accuracy. Dynamic calibration can be adjusted in real time according to environmental factors such as lighting and angle changes, so that the visual recognition system can maintain high accuracy in complex environments. By analyzing the product status through image analysis technology, it is possible to quickly detect whether the product meets the production standards and identify problems such as surface defects and shape deviations. By automatically adjusting the parameters of the production line, the system can quickly respond to product quality problems or equipment errors and reduce the generation of defective products. Through real-time adjustment and optimization, the time for production line downtime and adjustment is reduced, ensuring the efficiency and consistency of production.

[0011] Preferably, step S1 specifically includes:

[0012] Scan the production line through visual recognition equipment to obtain production line image data;

[0013] Obtain standard product image data;

[0014] The product visual area is extracted according to the standard product image data and the production line image data to obtain the production line product image data.

[0015] In the present invention, by combining standard product image data with production line image data, the visual area of ​​the product can be accurately extracted, background interference can be reduced, and recognition accuracy can be improved. Scanning operations are performed through visual recognition equipment without contacting the product, which reduces interference with the physical shape of the product and is suitable for high-speed assembly lines and fragile products. Standard product images are used as comparison objects to effectively distinguish between non-product areas and product areas on the production line, reduce false detection and missed detection, and improve the accuracy of visual recognition. By combining standard product images with production line image data, the visual area where the product is located can be accurately extracted, and irrelevant parts such as background and mechanical parts can be filtered out, thereby improving the accuracy of subsequent detection and analysis. By extracting the exact product area, the need for comprehensive analysis of the entire image is reduced, the processing time is shortened, and the efficiency of visual processing is improved.

[0016] Preferably, step S2 specifically includes:

[0017] Selecting a calibration mode according to the product image data of the production line to obtain calibration mode data;

[0018] Perform calibration point detection based on the product image data of the production line to obtain calibration point data;

[0019] Perform spatial calibration according to the calibration pattern data, the calibration point data and the production line product image data to obtain product spatial calibration data;

[0020] Perform color calibration according to the calibration pattern data and the product space calibration data to obtain product color calibration data;

[0021] Perform dynamic environmental compensation based on calibration mode data and product color calibration data to obtain product environmental factor compensation data;

[0022] The posture calibration is performed according to the calibration mode data and the product environmental factor compensation data to obtain the product image calibration data.

[0023] The calibration mode selection in the present invention automatically selects the appropriate calibration process according to the type of product, production environment and process requirements to ensure the optimal match of the calibration scheme. Selecting different calibration modes for different products or environments can avoid unnecessary calibration operations and improve efficiency. Accurate calibration point data helps to reduce the accumulation of spatial errors that may occur in subsequent steps, thereby improving the accuracy of the overall calibration. Through calibration point data and calibration mode, it is ensured that the production line products are accurately calibrated in space under different angles, positions, distances and other conditions, avoiding deviations caused by equipment mechanical errors or product movement. Color calibration ensures that the color data of the product is consistent under different lighting and reflection conditions, and is particularly suitable for color-sensitive production scenarios such as printing, textiles and other industries. Dynamic environmental compensation can compensate for environmental factors (such as lighting changes, temperature fluctuations, vibrations, etc.) in the production line in real time, ensuring that the visual recognition system can adapt to environmental changes and maintain high accuracy. Posture calibration can accurately detect and correct posture deviations (such as rotation, tilt, etc.) of products in the production line, ensuring that each product is inspected, processed or assembled in a standard posture.

[0024] Preferably, the calibration point detection is specifically:

[0025] Perform sub-pixel precision calibration point detection based on the product image data of the production line to obtain first calibration point data;

[0026] Performing geometric invariance feature calibration point detection based on the production line product image data to obtain second calibration point data;

[0027] Perform self-calibration calibration point detection according to the product image data of the production line to obtain third calibration point data;

[0028] Calibration point matching is performed according to the first calibration point data, the second calibration point data and the third calibration point data to obtain calibration point matching data.

[0029] The sub-pixel precision calibration point detection in the present invention effectively reduces the error caused by the limitation of the device resolution, improves the accuracy of the calibration point position, and ensures that the spatial calibration process is more accurate. In a complex environment where geometric features are not easy to change, the geometric invariance calibration point detection can effectively ensure the accurate identification of the calibration point, ensuring that the product can be correctly positioned even under non-ideal conditions. The self-calibration technology can automatically adjust the calibration point detection algorithm according to different product characteristics and environmental changes, adapt to the dynamic changes of the production line, and reduce manual intervention and additional settings. The self-calibration technology reduces the dependence on external factors, can effectively deal with environmental interference such as illumination, background debris, vibration, etc., and ensure the stability of calibration point detection. By matching the calibration point data from different detection methods, the limitations of each method can be eliminated, and the advantages can be taken to improve the accuracy of the calibration point through fusion and matching. By matching the multi-source calibration point data, the correction and compensation of the detection error can be realized, ensuring that the calibration point matching data has high reliability and accuracy.

[0030] Preferably, the sub-pixel precision calibration point detection is specifically:

[0031] Perform rough positioning processing of the calibration point area according to the product image data of the production line to obtain rough positioning data of the calibration point area;

[0032] Perform sub-pixel edge detection based on the coarse positioning data of the calibration point area to obtain sub-pixel edge detection data;

[0033] Perform sub-pixel centroid calculation according to sub-pixel edge detection data to obtain sub-pixel centroid data;

[0034] Sub-pixel shape fitting is performed according to the sub-pixel centroid data to obtain first calibration point data.

[0035] The coarse positioning in the present invention reduces the computational burden of the image processing algorithm, and only further refines the area of ​​the calibration point, thereby improving the efficiency of the system. Sub-pixel-level edge detection can accurately detect edge contours that are more detailed than pixel resolution, and the precise edge point positions are calculated through interpolation algorithms, which is suitable for scenes with high precision requirements. The centroid calculation further refines the position of the calibration point, so that the detection accuracy can reach the sub-pixel level, which is suitable for application scenarios that require high-precision positioning, such as precision machining or detection equipment. Through sub-pixel shape fitting (such as parabola fitting or spline curve fitting), the shape of the calibration point can be accurately fitted within the minimum error range, which not only ensures the accuracy of the calibration point position, but also retains its true geometric shape.

[0036] Preferably, the geometric invariance feature calibration point detection is specifically:

[0037] Extract geometric invariant features based on product image data of the production line to obtain geometric invariant feature data;

[0038] Performing geometric invariant feature point matching on the geometric invariant feature data to obtain candidate calibration point matching pair data;

[0039] According to the matching of candidate calibration points, the data is randomly sampled and consistent outliers are eliminated to obtain valid calibration point data;

[0040] Estimating affine transformation parameters according to valid calibration point data to obtain affine transformation parameter data;

[0041] The position of the calibration point is recalculated according to the affine transformation parameter data to obtain second calibration point data.

[0042] The present invention can extract stable local feature points. Even if there are changes in illumination, complex background or noise interference on the production line, the geometric invariance feature can still reliably extract important calibration point features. Based on the geometric invariance feature, the feature points in the standard template and the feature points actually detected can be quickly matched using algorithms such as Euclidean distance and Hamming distance to ensure the matching accuracy of the calibration points. The feature point matching results are verified by a random sampling consistency algorithm to eliminate those erroneous matching points that do not meet the geometric consistency to ensure the accuracy of the calibration points. Through affine transformation parameter estimation (such as rotation angle, scaling ratio, translation vector), the geometric transformation of the calibration points in the product image can be accurately restored to ensure that the calibration points can still be accurately located even when the product undergoes spatial changes. Through the affine transformation parameters, the actual spatial position of the calibration point can be accurately calculated, the displacement, rotation or scaling errors caused by the geometric transformation can be compensated, and the position accuracy of the calibration point can be ensured. Regardless of the transformation of the product on the production line, the system recalculates the calibration point position through this step to ensure that consistent and accurate calibration point data can be obtained for each detection.

[0043] Preferably, the self-calibration calibration point detection is specifically:

[0044] Perform edge detection and contour extraction based on the product image data of the production line to obtain the main contour data of the product;

[0045] Perform corner point detection based on the main contour data of the product to obtain product corner point data;

[0046] Perform repetitive feature analysis based on product corner point data to obtain product structural feature data;

[0047] Perform symmetry detection and extract candidate calibration points based on product structure feature data to obtain candidate calibration point data;

[0048] Performing geometric invariance analysis on candidate calibration point data to obtain geometric invariance data of the candidate calibration points;

[0049] The self-calibration feature point verification is performed according to the geometric indeformation data of the candidate calibration points to obtain the third calibration point data.

[0050] The present invention can cope with production environments with different lighting conditions, complex backgrounds and noise interference, and ensure the accuracy and stability of contour extraction. By analyzing the geometric characteristics of the product and the repeatability of the local structure, parts of the product with the same or similar features can be detected, and these parts are usually stable and repeatable candidate areas for calibration points. Since the repeatability features have high stability on the product, by selecting these feature areas as calibration points, the stability and robustness of the calibration points in the dynamic changes of the production line can be improved. The symmetry of the product is a stable and reliable geometric feature. Through symmetry detection, the exact position of the calibration points can be determined, which is particularly suitable for products with symmetrical structures. Geometric invariance analysis ensures that the calibration points can remain unchanged under geometric transformations such as rotation, scaling, and translation, and adapt to changes caused by product movement, equipment jitter, etc. on the production line. Through self-calibration feature point verification, the system can automatically check and confirm the accuracy of the calibration points, eliminate calibration points that do not meet the standards or are misjudged, and ensure that the output calibration points have high accuracy.

[0051] Preferably, step S3 is specifically:

[0052] Perform product morphology analysis according to product image calibration data to obtain product morphology analysis data;

[0053] Perform surface quality inspection according to product image calibration data to obtain surface quality inspection data;

[0054] Generate X-ray images through X-ray equipment to obtain product X-ray image data;

[0055] Perform internal structure analysis based on product X-ray image data to obtain internal structure analysis data;

[0056] Performing dimensional accuracy measurement on product image calibration data to obtain dimensional accuracy measurement data;

[0057] Product morphology analysis data, surface quality inspection data, internal structure analysis data and dimensional accuracy measurement data are integrated to obtain product status data.

[0058] In the present invention, by performing morphological analysis on the calibration data of the product image, the appearance characteristics of the product can be accurately identified, ensuring that the shape meets the design standards, and is suitable for production scenarios that require high-precision shape control. Automated surface quality detection can quickly identify problems such as flaws, scratches, pits or contamination on the surface of the product to ensure the high quality and consistency of the product appearance. X-rays can detect hidden defects inside the product, such as cracks, cavities, bubbles, foreign matter, etc. These internal problems cannot be found in conventional visual inspections, greatly improving the comprehensiveness of product quality control. By analyzing the X-ray image, it is possible to accurately detect whether the internal geometric structure of the product meets the design requirements, ensuring the correct assembly and structural integrity of the internal parts of the product. The dimensional measurement is performed through the calibrated image to ensure that the size of the product meets the design specifications, which is particularly suitable for industrial manufacturing fields with high dimensional accuracy requirements. By integrating various test data together, the quality status of each product can be fully grasped to ensure that any potential problems can be discovered and handled before entering the downstream process, reducing the accumulation and spread of quality problems.

[0059] Preferably, step S4 is specifically:

[0060] Perform product status parameter conversion according to the product status data to obtain product status parameter data;

[0061] Match the product status parameter data with the preset product quality inspection index library to obtain product status quality inspection data;

[0062] The production line is adjusted based on the product status quality inspection data and the preset production line expert knowledge engine to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

[0063] In the present invention, by converting product status data into specific product status parameters, the system can simplify the processing of complex multi-dimensional data, extract key quality parameters (such as shape, size, internal structure, etc.), and ensure that various status information of the product is converted into standardized parameter values ​​through status parameter conversion, which is convenient for consistency comparison with quality inspection indicators and improves data processing efficiency. By matching with the preset quality inspection indicator library, the system can automatically determine whether the product meets the quality standards, realize fully automatic quality inspection and evaluation, and reduce errors and subjectivity in manual quality inspection. The expert knowledge engine can quickly determine the status of the production line according to the changes in quality inspection data, adjust the parameters of the production line in time, prevent the expansion of quality problems, and reduce the appearance of defective products. The expert knowledge engine can gradually optimize its own decision-making logic through the accumulation and learning of historical data, improve the accuracy and efficiency of production line adjustment, and make the system more intelligent and efficient in long-term operation.

[0064] Preferably, the present application also provides a dynamic production line measurement and control system based on visual recognition, which is used to execute the dynamic production line measurement and control method based on visual recognition as described above. The dynamic production line measurement and control system based on visual recognition includes:

[0065] The production line product data collection module is used to collect production line product data through visual recognition equipment to obtain production line product image data;

[0066] A product image calibration module, used to perform dynamic calibration according to the product image data of the production line to obtain product image calibration data;

[0067] A product status analysis module, used to perform product status analysis based on product image calibration data to obtain product status data;

[0068] The production line adjustment processing module is used to perform production line adjustment processing according to product status data to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

[0069] The beneficial effects of the present invention are: from product data collection to production line adjustment, a closed-loop automated data flow is formed, errors and lags introduced by human intervention are avoided, and production efficiency is improved. Visual recognition, as a contactless detection method, will not cause physical interference to the product and is suitable for industries requiring high-precision production such as precision manufacturing and electronic equipment. Through calibration in multiple dimensions such as space, color, environment and posture, it is ensured that the product detection results remain high-precision even under different conditions of illumination, viewing angle and position changes. Dynamic environmental compensation enables the system to make adaptive adjustments according to real-time environmental changes (such as illumination, temperature, vibration, etc.), ensuring that the production line detection remains stable in complex environments. Morphological analysis can accurately identify the geometric shape of the product, ensure that its appearance meets the design standards, and promptly discover appearance defects caused by mold deviation or production error. Surface quality inspection quickly detects product surface defects through visual recognition equipment, while X-ray images provide non-destructive detection of internal structures, so that both surface and internal defects can be fully detected, ensuring all-round quality control of the product. Through real-time product data analysis and quality inspection matching, the system can provide instant feedback based on different product status, adjust production line parameters, quickly respond to quality issues, and prevent quality issues from spreading to the entire production batch. Using the preset expert system, the adjustment of the production line is based on historical data and professional experience, which can achieve refined and intelligent production control. The expert system can automatically adjust common problems and optimize decision logic through learning. The status data of each product can be fed back to the production line for adjustment and optimization, forming a closed-loop control and reducing quality fluctuations in production. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0071] Figure 1 A flowchart showing a method for dynamic production line measurement and control based on visual recognition according to an embodiment of the present invention is provided;

[0072] Figure 2 A flowchart showing a method for collecting product data of a production line according to an embodiment of the present invention is provided;

[0073] Figure 3 A flowchart showing the steps of a product image calibration method according to an embodiment is shown;

[0074] Figure 4 A flowchart showing a method for analyzing product status according to an embodiment of the present invention is provided;

[0075] Figure 5 A flowchart of a production line adjustment processing method according to an embodiment is shown. DETAILED DESCRIPTION

[0076] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0077] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0078] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0079] Use visual recognition equipment to scan products on the production line in real time and generate high-definition image data of the production line products. The image data includes information such as the shape, surface features, and spatial position of the product on the production line. The production line produces some precision mechanical parts. The standard shape of each part is a cylinder with a height of 10 mm and a diameter of 5 mm. The system uses visual recognition equipment to capture the appearance image of the part, obtain preliminary product shape data, and measure the appearance dimensions of the part to be 9.8 mm high and 5.1 mm in diameter.

[0080] The system is dynamically calibrated based on the product image data to correct for product position on the production line, color differences caused by lighting conditions, and posture changes. After calibration, the product image data is processed to ensure subsequent detection accuracy. Select the appropriate calibration mode according to different product types. Perform sub-pixel precision calibration point detection on the image to ensure positioning accuracy. Color calibration eliminates the impact of ambient lighting on the image and dynamically adjusts the system to adapt to environmental changes. During image acquisition, the shape of the detected parts was slightly distorted due to the different angles of the visual recognition equipment. Through dynamic calibration, the system adjusts the spatial position of the object in the image, corrects the distorted cylinder, and obtains accurate data with a height of 9.9 mm and a diameter of 5.0 mm.

[0081] The system analyzes the status of the product from multiple dimensions, including: Analyzing the geometric shape of the product to determine whether it meets the design requirements. Identifying surface defects such as scratches and dents. Performing non-destructive testing of the internal structure through X-rays to ensure that there are no internal defects (such as bubbles, cracks, etc.). Accurately measure the dimensions of the product and compare them with the design specifications. Through morphological analysis, the system detected that the part is 9.9 mm in height and 5.0 mm in diameter, which meets the design tolerance requirement of ±0.1 mm. Surface inspection found that there were slight scratches on the surface of the part, measuring 0.05 mm, which did not affect the performance and met the quality standards. X-ray images show that the internal structure of the part is intact, without bubbles and cracks. Dimension measurement confirms that the overall size of the part meets the requirements.

[0082] Based on the product status data, the system converts these test results into product status parameters and matches them with the preset product quality inspection index library. Based on the matching results, the system automatically adjusts the parameters of the production line through the expert knowledge engine. Convert the morphological analysis, surface quality, internal structure and dimensional accuracy data into quantifiable parameters. Match with the preset quality inspection standards to determine whether the product is qualified. If a deviation is detected, the system will adjust the production line parameters such as equipment accuracy, temperature, pressure, etc. through the expert knowledge engine to ensure that subsequent products meet the quality requirements. According to the quality inspection standards, the product size, surface quality and internal structure all meet the standards, so the product is judged to be qualified. If the dimensional deviation of the part exceeds the tolerance range (for example, the height is detected to be 9.6 mm), the system will identify the cause of the problem based on the expert knowledge engine and find that it may be due to the wear of the cutting tool of the cutting equipment. The system will automatically adjust the tool parameters of the cutting equipment to compensate for this deviation and ensure the dimensional accuracy of subsequent parts.

[0083] The production line is responsible for manufacturing small metal connectors for electronic devices, which require a height of 2.0 mm, a width of 1.5 mm, a thickness of 0.5 mm, and a tolerance of ±0.05 mm. 5,000 such connectors are produced daily. The visual recognition equipment captures image data for each connector. The preliminary test results show that the dimensions of some connectors are slightly deviated, such as 1.98 mm (height), 1.52 mm (width), and 0.48 mm (thickness). The system performs dynamic calibration to correct the detection errors caused by changes in the device angle or lighting. After calibration, it is confirmed that the actual dimensions of the connector are 2.01 mm (height), 1.50 mm (width), and 0.50 mm (thickness), which are within the tolerance range. It is confirmed that the geometry of the connector meets the design requirements and the detected morphological deviation is within an acceptable range. The visual recognition equipment found that there were tiny scratches on the surface of some connectors. After analysis, the scratch depth was 0.02 mm, which met the surface quality standards. X-rays show that there are no defects such as bubbles and cracks inside the connector, and the internal structure is intact. Confirm that the dimensions of the final product are within the tolerance range. The width deviation of some products is +0.02 mm and the thickness deviation is -0.01 mm, which still meets the precision requirements.

[0084] The system converts the test results into product status parameters and matches them with quality inspection standards. If a small number of connectors are found to have widths outside the tolerance range (e.g., the width reaches 1.56 mm), the system will identify equipment wear or material shrinkage issues and automatically adjust the cutting speed and pressure of the equipment through the expert system. The adjusted production line continues to run, and the width of the next batch of products is stabilized at 1.50 mm, ensuring that the dimensions of all connectors meet the requirements.

[0085] See also Figures 1 to 5 The present application provides a dynamic production line measurement and control method based on visual recognition, comprising the following steps:

[0086] Step S1: collect production line product data through visual recognition equipment to obtain production line product image data;

[0087] Specifically, multiple visual recognition devices, such as cameras or image sensors, are arranged on the production line to ensure that the devices can cover the key parts of the production line. When a product passes through the production line, the visual recognition device starts working and captures the image data of the product, which contains information such as the product's shape, color, and size. The image is segmented to separate the product from the background and extract the area of ​​interest.

[0088] Step S2: Perform dynamic calibration according to the product image data of the production line to obtain product image calibration data;

[0089] Specifically, extract the main feature information of the product, such as edges, shapes, textures, etc., from the preprocessed image to ensure that the status of the product can be accurately described. Compare these features with the preset standard template or reference model. The reference model is usually based on known standard product information and covers the ideal form and parameters of the product. According to the comparison results, adjust the recognition parameters of the system in real time to cope with subtle changes in the product during the production process. For example, changes in production line speed, changes in lighting conditions, etc., all of which will affect the quality of the image, so it is necessary to ensure consistency by adjusting the calibration.

[0090] Step S3: Perform product status analysis based on the product image calibration data to obtain product status data;

[0091] Specifically, based on the image calibration data of the product, the system builds a state analysis model that includes a series of key indicators such as the product's size, shape, color, etc. The extracted product features are compared with the analysis model. For example, it detects whether the product has defects (such as damage, deformation, etc.), or determines whether the product is assembled correctly (such as whether parts are missing or in the wrong position). Based on the comparison results, the product is classified as "qualified" or "unqualified", and the specific reasons for the unqualified are marked for subsequent processing.

[0092] Step S4: Perform production line adjustment processing according to the product status data to obtain production line adjustment data to perform auxiliary operations for production line parameter adjustment.

[0093] Specifically, based on the product status data, the system analyzes the current operating status of the production line and determines whether the production parameters need to be adjusted. For example, if the number of unqualified products increases, it may be due to equipment wear, improper parameter settings, and other problems. Based on the product status analysis results, the system gives specific production line adjustment suggestions. These suggestions include: adjusting the production line speed, adjusting the product position positioning accuracy, or modifying the parameter settings of the production equipment. In some cases, the system will automatically make real-time parameter adjustments based on the adjustment suggestions, such as adjusting the conveyor belt speed, the operating accuracy of the equipment, or the recognition sensitivity of the visual recognition equipment to optimize the operating efficiency of the production line.

[0094] Preferably, step S1 specifically includes:

[0095] Scan the production line through visual recognition equipment to obtain production line image data;

[0096] Specifically, multiple visual recognition devices (such as high-resolution cameras or 3D scanners) are installed and configured to ensure that the devices can cover key areas of the production line. The angle and focal length of the device need to be adjusted so that it can capture a complete image of the product. When the product passes through the production line, the visual recognition device continuously shoots the product to generate production line image data, which usually includes the entire production environment and the background around the product. When the product passes by, the camera continuously shoots multiple images per second (such as 10 frames per second) to generate a series of pictures containing the product.

[0097] Obtain standard product image data;

[0098] Specifically, data of standard product images is acquired from a preset database, wherein the standard product images are generated by photographing the standard product at multiple angles to capture complete feature information. The acquired standard images are product images with high resolution and clean background.

[0099] The product visual area is extracted according to the standard product image data and the production line image data to obtain the production line product image data.

[0100] Specifically, the real-time image data on the production line is compared with the standard product image data. The boundaries and shapes of the products are found through feature matching technology, and the background parts of the image that are not related to the product are removed to ensure that only the areas that match the standard products are extracted. The product area is extracted from the production line image using shape features, color distribution and other information, and the product is separated from the background through image segmentation technology to extract the outline and shape of the product. The extracted product area is finely compared with the standard product to confirm whether the extracted area is accurate. If there are errors in position or size, calibration is performed to make the product image data match the standard image as much as possible.

[0101] Preferably, step S2 specifically includes:

[0102] Selecting a calibration mode according to the product image data of the production line to obtain calibration mode data;

[0103] Specifically, based on the characteristics of the product (such as shape, material, color, etc.) and the operating conditions of the production line, the system automatically selects the corresponding calibration mode. Different calibration modes are optimized for different dimensions of the product (such as space, color or posture). Based on the image data of the product on the production line, the system analyzes the current environment, such as light conditions, product movement speed, etc., and selects the corresponding calibration mode. The calibration mode can be divided into standard mode, enhanced precision mode or fast mode, depending on the requirements of the production line. The system first determines the most appropriate calibration mode by analyzing the quality of the image (such as clarity, contrast) and the speed of the production line. If the product passes at high speed, the system will select the fast mode; if the ambient lighting is complex, the enhanced precision mode is selected.

[0104] Perform calibration point detection based on the product image data of the production line to obtain calibration point data;

[0105] Specifically, some key feature points are extracted from the product image. These feature points are one of the product's corners, marking points, or specific contours of the shape. These calibration points can be used as a reference for calibration. The system accurately locates the calibration points through image processing algorithms to ensure their accuracy, such as using edge detection, corner detection or other feature extraction techniques.

[0106] Perform spatial calibration according to the calibration pattern data, the calibration point data and the production line product image data to obtain product spatial calibration data;

[0107] Specifically, based on the data of the calibration points and the calibration mode, the system performs a geometric transformation on the image to correct the relative position deviation between the camera and the product, compensate for the error in the camera angle or installation position, and ensure that the product in the image maintains the correct spatial proportion. In some cases, the system needs to convert the two-dimensional information in the image into three-dimensional spatial information and determine the position and posture of the product in the actual space through the calibration point data.

[0108] Perform color calibration according to the calibration pattern data and the product space calibration data to obtain product color calibration data;

[0109] Specifically, based on the spatially calibrated image data, the system first converts the image's color space to make it more suitable for color analysis (for example, converting the RGB color space to the HSV or Lab color space). The system uses the color model of standard products (including the color's hue, brightness, saturation, etc.) to compare with the image of the current product. Through comparison, the system corrects the color deviation in the image caused by environmental factors such as light and reflection.

[0110] Perform dynamic environmental compensation based on calibration mode data and product color calibration data to obtain product environmental factor compensation data;

[0111] Specifically, the system monitors factors such as light intensity, color changes, and background noise in the production line environment in real time. These environmental factors can cause changes in product color and features in the image. Based on the real-time environmental data, the system dynamically compensates the image, including adjusting brightness, contrast, or applying filtering technology to compensate for the impact of the environment on the image.

[0112] The posture calibration is performed according to the calibration mode data and the product environmental factor compensation data to obtain the product image calibration data.

[0113] Specifically, the system uses previously extracted calibration points, spatial calibration data, and color calibration data to detect the posture of the product in the image (including rotation, tilt, displacement, etc.). The system determines whether correction is required by comparing the standard posture model with the posture in the current image. If the posture of the product is detected to be different from the standard posture (such as the product is tilted or rotated at a different angle), the system will rotate, translate, or scale the image to ensure that the product image is consistent with the standard posture.

[0114] Preferably, the calibration point detection is specifically:

[0115] Perform sub-pixel precision calibration point detection based on the product image data of the production line to obtain first calibration point data;

[0116] Specifically, edge detection technology (such as gradient method or difference method) is first used to locate the key feature edge of the product. The gradient method or difference method used identifies the edge by detecting significant changes in pixel intensity in the image. In order to improve accuracy, a sub-pixel interpolation method is used to increase the accuracy of the calibration points to the sub-pixel level between pixels. The system locates more accurate calibration point coordinates at the edge point by analyzing the changing trend of the pixel grayscale value. Geometric features with special significance on the edge (such as corners or places with significant curvature) are extracted as the first batch of calibration point data.

[0117] Performing geometric invariance feature calibration point detection based on the production line product image data to obtain second calibration point data;

[0118] Specifically, geometric invariant features such as corner points, edge points or local features of the region (such as the label edge of the bottle, the unique shape of the bottle cap, etc.) are extracted from the image. These features remain stable under geometric transformations such as rotation and scaling. Using the invariant feature detection method, invariant feature points are detected through template matching or specific feature descriptors (such as edge-based shape features or local textures). The detected invariant features are compared with the invariant features in the standard product image, and these feature points are determined as the second calibration points.

[0119] Perform self-calibration calibration point detection according to the product image data of the production line to obtain third calibration point data;

[0120] Specifically, the system automatically detects additional calibration points through self-calibration technology. These points are generated by analyzing the symmetry, proportion or other internal geometric relationships of the image. Self-calibration is usually used to correct image distortion and improve the accuracy of calibration points. According to the symmetry or repetitive characteristics of the product (such as the symmetry axis of the bottle body, the equally spaced marks on the bottle body, etc.), the system automatically generates a series of third calibration points. By analyzing the geometric distribution of these calibration points, the system self-calibrates them to ensure that the positions of these points are consistent with the standard model.

[0121] Calibration point matching is performed according to the first calibration point data, the second calibration point data and the third calibration point data to obtain calibration point matching data.

[0122] Specifically, the data of the first batch of sub-pixel precision calibration points, the second batch of geometric invariance feature calibration points, and the third batch of self-calibration calibration points are integrated. Each type of calibration point provides different calibration information, and the system combines this information to improve the calibration accuracy. Through feature matching technology, these calibration points are matched with the calibration points of standard products and the differences between them are calculated. Based on the differences, the system fine-tunes and matches the calibration points to ensure that they can be accurately aligned with the calibration points in the standard template. If there are errors in the matching process, the system will perform error correction based on the characteristics of the three types of calibration points and output accurate calibration point matching data.

[0123] Preferably, the sub-pixel precision calibration point detection is specifically:

[0124] Perform rough positioning processing of the calibration point area according to the product image data of the production line to obtain rough positioning data of the calibration point area;

[0125] Specifically, the production line product images are first processed with basic methods, such as grayscale, denoising, and contrast enhancement, to improve the accuracy of subsequent calibration point detection. The main feature areas of the products are found in the image through threshold segmentation, morphological operations, or edge detection techniques, which are used to identify the approximate calibration point areas, but the accuracy is relatively rough. The rough position of the calibration points is locked by analyzing the geometric features in the image. For example, a rough center position is identified in the product area.

[0126] Perform sub-pixel edge detection based on the coarse positioning data of the calibration point area to obtain sub-pixel edge detection data;

[0127] Specifically, within the rough positioning area of ​​the calibration point, edge detection technology is applied to further extract detailed edge information. Gradient detection, Canny edge detection and other methods can be used to accurately find the pixel position of the edge. In order to achieve sub-pixel accuracy, interpolation processing is performed near the detected edge point. By analyzing the grayscale gradient or difference of the pixels around the edge point, the edge position finer than the pixel is calculated.

[0128] Perform sub-pixel centroid calculation according to sub-pixel edge detection data to obtain sub-pixel centroid data;

[0129] Specifically, the centroid of the calibration point area is calculated based on the detected sub-pixel edge points. The centroid calculation is based on the coordinates of the edge points, and the centroid of the sub-pixel accuracy is calculated by counting the geometric center of the edge points. On the basis of sub-pixel accuracy, the accuracy of the centroid is further improved. The centroid position is determined by combining the density and grayscale information of the edge through methods such as weighted averaging.

[0130] Specifically, more importantly, multi-channel fusion is performed on the sub-pixel edge detection data to obtain edge detection multi-channel fusion data; local spatial resampling is performed on the edge detection multi-channel fusion data to obtain local resampled edge point data; dynamic edge weight allocation is performed on the local resampled edge point data to obtain weighted edge point data; dynamic centroid clustering is performed on the weighted edge point data to obtain centroid candidate area data; weighted centroid calculation is performed on the centroid candidate area data to obtain sub-pixel centroid data.

[0131] Sub-pixel edge detection data contains image feature information in multiple dimensions such as brightness and color. In order to improve the accuracy of centroid calculation, these data can be processed in multiple channels, that is, different channel information (such as RGB color channel, brightness channel, texture channel, etc.) is weighted and fused according to their importance. Through the weight distribution algorithm based on gradient and edge strength (such as adaptive channel weight distribution algorithm), each channel is dynamically weighted to generate fused high-precision sub-pixel edge point data. Since sub-pixel edge points have high precision but sparse distribution, it is necessary to perform local spatial resampling on them before centroid calculation. A high-density edge point set is regenerated in the neighborhood of sub-pixel edge points through spatial interpolation algorithms (such as bilinear interpolation or high-order B-spline interpolation), which further refines the distribution of edge points to increase the accuracy of centroid calculation.

[0132] For each resampled edge point, different weights are assigned according to its distance from the candidate centroid region, edge strength, and gradient direction. A dynamic edge weighting mechanism is used to ensure that points closer to the centroid have greater weights by combining the dual factors of spatial distance and edge strength (such as weighted regression based on spatial distance and edge strength to obtain the corresponding weights). For each edge point, the distance between it and the center of the candidate centroid region (or the initial estimated position) is calculated. The calculation of spatial distance can reflect the relative position of the edge point and the centroid position. Points close to the center of the candidate centroid region should have greater influence in the centroid calculation. For each edge point, its edge strength is calculated by the gradient of the image. The gradient is the rate of change of image brightness, which can reflect the clarity and significance of the edge. The level of edge strength represents the significance of the edge point. Edge points with higher strength should have greater weight in the centroid calculation, especially in complex edge or noisy environments, where points with higher strength have higher credibility. The core of the dynamic edge weighting mechanism is to assign a weight coefficient based on spatial distance and edge strength to each edge point. The weighting formula is: , is the weight data / dynamic weighting coefficient corresponding to the weighted edge point data, is the edge strength, is the spatial distance, A small positive number is used to prevent the divisor from being 0. Based on the dynamic weighting coefficient, after the weight of each edge point is calculated, the weight will be directly used to calculate the centroid. At this time, the coordinates of each edge point will be multiplied by its corresponding weight to obtain the weighted edge point data used for centroid calculation.

[0133] At the same time, the edge point removal algorithm based on noise analysis can remove noise points that deviate from the standard. Based on the weighted edge point data, all edge points are first subjected to a preliminary clustering analysis to find the candidate centroid region. Through a density-based clustering algorithm (such as the DBSCAN algorithm), densely distributed edge points are grouped together and sparse noise points are excluded. The generated candidate centroid region further optimizes the selection of the centroid position. Based on the candidate centroid region data, the weighted coordinates of all edge points in the local area are calculated by using a weighted centroid formula (such as an improved version of the classic centroid formula) to obtain the centroid position. Here, the centroid coordinates (x, y) are calculated using the following formula: , , is the horizontal coordinate data of the sub-pixel centroid data, For the The weight of the edge points, For the The horizontal coordinate data of the edge points, is the ordinate data of the sub-pixel centroid data, For the At this time, the centroid calculation not only considers the spatial position, but also integrates multiple image features to make the centroid positioning more accurate.

[0134] Sub-pixel shape fitting is performed according to the sub-pixel centroid data to obtain first calibration point data.

[0135] Specifically, based on the obtained sub-pixel centroid data, a geometric fitting method (such as least squares method to fit a circle or ellipse) is applied to fit the shape of the calibration point, which can not only correct the centroid data, but also obtain the accurate shape features of the calibration point. Through the fitted shape features, the system determines the position and shape parameters (such as the center of the circle, radius, etc.) of the calibration point as the first calibration point data.

[0136] Preferably, the geometric invariance feature calibration point detection is specifically:

[0137] Extract geometric invariant features based on product image data of the production line to obtain geometric invariant feature data;

[0138] Specifically, basic image processing such as grayscale, denoising, and contrast enhancement is performed on the product images of the production line. In this way, the clarity of the image and the recognizability of the calibration points are improved. The approximate calibration point area is identified from the product image through threshold segmentation, edge detection or morphological operations. Coarse positioning usually extracts larger geometric features (such as bottle caps, labels, etc.) to determine the approximate location of the calibration points. The system uses a coarse positioning algorithm to frame the calibration point area into a larger rectangular or circular frame as the basis for subsequent sub-pixel precision processing.

[0139] Performing geometric invariant feature point matching on the geometric invariant feature data to obtain candidate calibration point matching pair data;

[0140] Specifically, a more sophisticated edge detection algorithm (such as Canny, Sobel, etc.) is applied in the coarse positioning area of ​​the calibration point to detect the edge of the calibration point. The edge detection accuracy at this time is at the pixel level. In order to improve the accuracy of edge points, a sub-pixel interpolation method is used. By interpolating the change in the gray value of the neighborhood, edge points that are more accurate than the pixel grid are found. Common interpolation methods include linear interpolation or parabolic interpolation. The sub-pixel precision edge points after interpolation are extracted for further calculation.

[0141] According to the matching of candidate calibration points, the data is randomly sampled and consistent outliers are eliminated to obtain valid calibration point data;

[0142] Specifically, random sampling is performed from candidate matching pairs to gradually remove abnormal points or noise points. Random sampling consistency outlier detection finds point pairs that satisfy a certain geometric relationship (such as an affine transformation model) from matching point pairs through multiple iterations, and removes abnormal points that do not conform to the model. Point pairs with large deviations (wrong matches caused by occlusion, illumination or noise) are removed, and only valid calibration point matching pairs are retained. Several point pairs are randomly selected from candidate matching point pairs for consistency evaluation to find the maximum consensus set that satisfies a certain geometric transformation model (such as affine transformation). In the iterative process, those abnormal points (noise points) that are inconsistent with the mainstream matching results are removed. In each iteration, by estimating the affine model and checking whether each candidate matching point conforms to the model, those wrong matching points with large deviations are removed. The system randomly selects several points from a set of candidate calibration point matching pairs, and removes abnormal points that do not conform to the transformation model by randomly sampling consistent outliers. For example, the system finds that some noise points do not conform to the expected affine transformation, leaving valid calibration point pairs, such as (120, 180) and (115, 175).

[0143] Estimating affine transformation parameters according to valid calibration point data to obtain affine transformation parameter data;

[0144] Specifically, the parameters of the affine transformation are calculated using the retained valid calibration point pairs. Affine transformation includes parameters such as translation, rotation, scaling, and shearing, which can be estimated by the least squares method and other methods. The system compares the positions of the calibration points in the standard image and the current image, calculates the affine relationship between these points, and then obtains the transformation matrix.

[0145] The position of the calibration point is recalculated according to the affine transformation parameter data to obtain second calibration point data.

[0146] Specifically, the calibration point positions in the standard image are mapped to the current image through affine transformation parameters to obtain the recalculated positions of the calibration points. The calibration points adjusted by the affine transformation are used as the second calibration point data.

[0147] Preferably, the self-calibration calibration point detection is specifically:

[0148] Perform edge detection and contour extraction based on the product image data of the production line to obtain the main contour data of the product;

[0149] Specifically, edge detection techniques (such as Canny, Sobel, etc.) are used to extract the edge contour of the product. Edge detection identifies the outer contour of the product by capturing significant changes in pixel values ​​in the image (such as grayscale gradients). Based on the detected edge points, the complete contour of the product is extracted to form a closed curve. Through morphological operations (such as expansion, erosion, etc.), the integrity and continuity of the contour are improved.

[0150] Perform corner point detection based on the main contour data of the product to obtain product corner point data;

[0151] Specifically, the significant corners on the contour are identified by corner detection technology (such as Harris corner detection). Corner points are usually the locations where the edge direction changes dramatically in the image, which can represent the geometric features of the product. From the extracted main contour, find all the corner points and record their coordinates as corner point data. These corner points will be used as feature points in the subsequent steps.

[0152] Perform repetitive feature analysis based on product corner point data to obtain product structural feature data;

[0153] Specifically, by analyzing the corner points in multiple image frames, it is determined whether these corner points remain stable under different viewing angles and zoom conditions. Repeatable features usually refer to feature points that appear repeatedly in multiple images and whose positions do not change significantly. Those corner points that are repeatable and stable in multiple frames of images are screened out to form the structural feature data of the product. These points are usually key parts of the product structure, such as label edges, symmetrical points on the bottle body, etc.

[0154] Perform symmetry detection and extract candidate calibration points based on product structure feature data to obtain candidate calibration point data;

[0155] Specifically, the symmetry of the product is detected by analyzing the geometric distribution of the product's structural features. Symmetry can be achieved by calculating whether the distribution of feature points about a certain axis is symmetrical, such as whether the corner points on the left and right sides of the bottle are symmetrical. Based on the symmetry detection, feature points that are located near the symmetry axis or can describe the symmetry of the product are extracted as candidate calibration points.

[0156] Performing geometric invariance analysis on candidate calibration point data to obtain geometric invariance data of candidate calibration points;

[0157] Specifically, by analyzing whether the candidate calibration points maintain geometric relationships under different transformations (such as rotation and scaling). The geometric invariance property can be determined by comparing the relative position changes of feature points after affine transformation or other geometric operations. Verify whether the candidate calibration points can maintain consistency under different viewing angles or scale changes. If the position relationship of the calibration points remains unchanged under multiple conditions, its geometric invariance can be determined.

[0158] The self-calibration feature point verification is performed according to the geometric indeformation data of the candidate calibration points to obtain the third calibration point data.

[0159] Specifically, based on the geometric invariant data of the candidate calibration points, verify whether they meet the self-calibration characteristics, mainly to determine whether these points can be calibrated through internal structural relationships. The feature points of self-calibration need to meet the requirements of symmetry, repeatability and geometric invariance. The calibration points verified by the self-calibration characteristics will be determined as valid calibration points, and the system will output these points as the third calibration point data.

[0160] Preferably, step S3 specifically includes:

[0161] Perform product morphology analysis according to product image calibration data to obtain product morphology analysis data;

[0162] Specifically, the system uses image processing technology to analyze the overall shape of the product, mainly including the recognition and calculation of the product's outline, shape, structural features, etc. Morphological operations (such as corrosion, expansion) or feature extraction algorithms are used to identify the geometric shape of the product. The key geometric features of the product (such as height, width, angle, curve, etc.) are extracted and compared with the morphological parameters in the standard model to determine whether the product meets the expected geometric shape.

[0163] Perform surface quality inspection according to product image calibration data to obtain surface quality inspection data;

[0164] Specifically, the surface of the product is analyzed through high-resolution image processing technology to detect surface defects (such as scratches, dents, bubbles, pollution, etc.). Texture analysis and pattern recognition technology are used to find abnormal areas. By comparing the texture and grayscale distribution of the normal surface, possible defect areas are automatically detected. Then the defects are classified and different types of surface defects are marked.

[0165] Generate X-ray images through X-ray equipment to obtain product X-ray image data;

[0166] Specifically, an X-ray device is used to scan the product to generate an image of the product's internal structure. X-rays will produce different attenuations when passing through materials of different densities, and the system generates X-ray images based on this attenuation information. X-ray images are collected through sensors and preprocessed (such as denoising and contrast enhancement) to improve image clarity and resolution.

[0167] Perform internal structure analysis based on product X-ray image data to obtain internal structure analysis data;

[0168] Specifically, based on the X-ray image data, the system analyzes the internal structure of the product, such as material distribution, internal defects (such as bubbles, cracks), and the connection of internal parts. Image segmentation and pattern recognition technology are used to distinguish and analyze areas of different densities. By comparing the normal internal structure with the current X-ray image, internal abnormal points or potential defect areas are automatically identified and marked.

[0169] Performing dimensional accuracy measurement on product image calibration data to obtain dimensional accuracy measurement data;

[0170] Specifically, based on the calibrated image data, the system automatically measures the key dimensional parameters of the product, such as length, width, thickness, etc. By comparing with the standard dimensional model, it evaluates whether the dimensional accuracy of the product is within the allowable error range. The system performs error analysis on the measurement results to determine whether the deviation between the actual measured size and the standard size meets the design requirements.

[0171] Product morphology analysis data, surface quality inspection data, internal structure analysis data and dimensional accuracy measurement data are integrated to obtain product status data.

[0172] Specifically, the system summarizes and integrates the data obtained from morphological analysis, surface quality inspection, internal structure analysis and dimensional accuracy measurement. Through multi-dimensional data fusion, a complete product status report is formed to evaluate whether the product meets the standards. Based on data from various aspects, the system evaluates the overall status of the product, gives a final judgment on whether it is qualified, and lists the test results of each data in detail.

[0173] Preferably, step S4 is specifically:

[0174] Perform product status parameter conversion according to the product status data to obtain product status parameter data;

[0175] Specifically, the product status data (morphological analysis, surface quality inspection, internal structure analysis, dimensional accuracy measurement, etc.) obtained from the previous stage are converted into standardized status parameters. The status parameters are expressed as digital indicators, such as qualified / unqualified judgment results, specific defect types and quantities, dimensional errors, etc. Through data cleaning and conversion, different types of status data are uniformly converted into a format that is easy to process, such as quantification of dimensional deviation into millimeters, classification of the area or severity of surface defects, etc.

[0176] Match the product status parameter data with the preset product quality inspection index library to obtain product status quality inspection data;

[0177] Specifically, the system compares the standardized product status parameters with the preset product quality inspection index library. The quality inspection index library contains information such as product quality requirements, qualification standards and tolerance ranges. According to the standards in the quality inspection index library, the system automatically determines whether the product status parameters meet the requirements. For status parameters that do not meet the standards, the system will mark the reasons for non-compliance, generate quality inspection data, and point out the specific problem areas.

[0178] The production line is adjusted based on the product status quality inspection data and the preset production line expert knowledge engine to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

[0179] Specifically, using the preset production line expert knowledge engine, the system analyzes whether the current production line parameters need to be adjusted based on the product status quality inspection data. The expert knowledge engine contains production processes and operating rules, and can generate adjustment suggestions based on the test results, such as adjusting production speed, changing equipment settings, calibrating equipment, etc. Based on the quality inspection data, the system analyzes the reasons for non-conformity and generates corresponding production line adjustment suggestions. Adjustment suggestions include adjusting equipment accuracy, reducing production speed, optimizing operating procedures, etc., to improve the product qualification rate.

[0180] Preferably, the present application also provides a dynamic production line measurement and control system based on visual recognition, which is used to execute the dynamic production line measurement and control method based on visual recognition as described above. The dynamic production line measurement and control system based on visual recognition includes:

[0181] The production line product data collection module is used to collect production line product data through visual recognition equipment to obtain production line product image data;

[0182] A product image calibration module, used to perform dynamic calibration according to the product image data of the production line to obtain product image calibration data;

[0183] A product status analysis module, used to perform product status analysis based on product image calibration data to obtain product status data;

[0184] The production line adjustment processing module is used to perform production line adjustment processing according to product status data to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

[0185] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0186] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A dynamic production line measurement and control method based on visual recognition, characterized in that: The following steps are involved: Step S1: collect production line product data through visual recognition equipment to obtain production line product image data; Step S2: Perform dynamic calibration according to the product image data of the production line to obtain product image calibration data; Step S3: Perform product status analysis based on the product image calibration data to obtain product status data; Step S4: Perform production line adjustment processing according to the product status data to obtain production line adjustment data to perform auxiliary operations for production line parameter adjustment; Wherein step S2 is specifically: Selecting a calibration mode according to the product image data of the production line to obtain calibration mode data; Perform calibration point detection based on the product image data of the production line to obtain calibration point data; Perform spatial calibration according to the calibration pattern data, the calibration point data and the production line product image data to obtain product spatial calibration data; Perform color calibration according to the calibration pattern data and the product space calibration data to obtain product color calibration data; Perform dynamic environmental compensation based on calibration mode data and product color calibration data to obtain product environmental factor compensation data; Perform posture calibration according to the calibration mode data and product environmental factor compensation data to obtain product image calibration data; The specific calibration point detection is as follows: Perform sub-pixel precision calibration point detection based on the product image data of the production line to obtain first calibration point data; Performing geometric invariance feature calibration point detection based on the production line product image data to obtain second calibration point data; Perform self-calibration calibration point detection according to the product image data of the production line to obtain third calibration point data; Calibration point matching is performed according to the first calibration point data, the second calibration point data and the third calibration point data to obtain calibration point matching data.

2. The method according to claim 1, characterized in that Step S1 is specifically as follows: Scan the production line through visual recognition equipment to obtain production line image data; Obtain standard product image data; The product visual area is extracted according to the standard product image data and the production line image data to obtain the production line product image data.

3. The method according to claim 1, characterized in that The sub-pixel precision calibration point detection is specifically as follows: Perform rough positioning processing of the calibration point area according to the product image data of the production line to obtain rough positioning data of the calibration point area; Perform sub-pixel edge detection based on the coarse positioning data of the calibration point area to obtain sub-pixel edge detection data; Perform sub-pixel centroid calculation according to sub-pixel edge detection data to obtain sub-pixel centroid data; Sub-pixel shape fitting is performed according to the sub-pixel centroid data to obtain first calibration point data.

4. The method according to claim 1, characterized in that The geometric invariant feature calibration point detection is specifically as follows: Extract geometric invariant features based on product image data of the production line to obtain geometric invariant feature data; Performing geometric invariant feature point matching on the geometric invariant feature data to obtain candidate calibration point matching pair data; According to the matching of candidate calibration points, the data is randomly sampled and consistent outliers are eliminated to obtain valid calibration point data; Estimating affine transformation parameters according to valid calibration point data to obtain affine transformation parameter data; The position of the calibration point is recalculated according to the affine transformation parameter data to obtain second calibration point data.

5. The method according to claim 1, characterized in that The self-calibration calibration point detection is specifically as follows: Perform edge detection and contour extraction based on the product image data of the production line to obtain the main contour data of the product; Perform corner point detection based on the main contour data of the product to obtain product corner point data; Perform repetitive feature analysis based on product corner point data to obtain product structural feature data; Perform symmetry detection and extract candidate calibration points based on product structure feature data to obtain candidate calibration point data; Performing geometric invariance analysis on candidate calibration point data to obtain geometric invariance data of candidate calibration points; The self-calibration feature point verification is performed according to the geometric indeformation data of the candidate calibration points to obtain the third calibration point data.

6. The method according to claim 1, characterized in that Step S3 is specifically as follows: Perform product morphology analysis according to product image calibration data to obtain product morphology analysis data; Perform surface quality inspection according to product image calibration data to obtain surface quality inspection data; Generate X-ray images through X-ray equipment to obtain product X-ray image data; Perform internal structure analysis based on product X-ray image data to obtain internal structure analysis data; Performing dimensional accuracy measurement on product image calibration data to obtain dimensional accuracy measurement data; Product morphology analysis data, surface quality inspection data, internal structure analysis data and dimensional accuracy measurement data are integrated to obtain product status data.

7. The method according to claim 1, characterized in that Step S4 is specifically as follows: Perform product status parameter conversion according to the product status data to obtain product status parameter data; Match the product status parameter data with the preset product quality inspection index library to obtain product status quality inspection data; The production line is adjusted based on the product status quality inspection data and the preset production line expert knowledge engine to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

8. A dynamic production line measurement and control system based on visual recognition, characterized in that: Used to execute the dynamic production line measurement and control method based on visual recognition as claimed in claim 1, the dynamic production line measurement and control system based on visual recognition comprises: The production line product data collection module is used to collect production line product data through visual recognition equipment to obtain production line product image data; A product image calibration module, used to perform dynamic calibration according to the product image data of the production line to obtain product image calibration data; A product status analysis module, used to perform product status analysis based on product image calibration data to obtain product status data; The production line adjustment processing module is used to perform production line adjustment processing according to product status data to obtain production line adjustment data for auxiliary operations of production line parameter adjustment.

Images