Image recognition-based rope net flaw online detection method and system

Abstract

This invention relates to the field of rope and net technology and discloses an online detection method and system for rope and net defects based on image recognition. The method includes, in the defect determination process, transforming a real-time image of the rope and net surface to be inspected based on the positional offset representing the rope and net's operational deviation to generate an aligned image and a deviation image; emitting probe lines from the real-time image of the rope and net surface to be inspected, and performing image correction processing by calculating color adjustment values ​​and fiber extension directions to identify suspected defect areas; and analyzing these suspected defect areas in conjunction with the deviation image to label them as defective or normal areas. This invention ensures the reliability of the labeling by combining the probe lines with the analysis of the deviation image, and by compensating for the physical displacement of the rope and net during production, it performs correction before image comparison, eliminating false alarms and missed alarms caused by positional mismatches, thereby improving the accuracy and reliability of defect detection.

Description

Technical Field

[0001] This invention belongs to the field of rope and net technology, specifically relating to an online detection method and system for rope and net defects based on image recognition. Background Technology

[0002] Rope nets are widely used in fishing, construction protection, sports facilities, and material transportation. The structural integrity and quality reliability of rope nets are directly related to the safety of production operations and facilities. With the improvement of industrial automation, automated defect detection technology using machine vision technology is used to inspect the quality of rope net products, thereby eliminating unqualified rope nets, ensuring the quality of rope net products, and improving inspection efficiency and objectivity.

[0003] However, existing technologies often employ relatively simple image processing algorithms or fixed judgment thresholds. If the goal is to achieve high sensitivity to detect minute defects, it is easy to misjudge normal texture fluctuations, uneven lighting, or slight shaking of the rope net itself as defects, resulting in false alarms and weakening the efficiency advantage of automated detection. If the judgment threshold is raised to reduce the false alarm rate, it may lead to missed detection of real defects that are highly concealed or atypical, potentially causing products with quality risks to enter the market.

[0004] Furthermore, due to the diverse materials, colors, weaving structures, and specifications of rope nets, the morphology of defects also varies. Existing detection technologies are typically trained for specific product types and lack the ability to adapt to diverse samples. When production lines switch product specifications or the working environment changes, the detection effect declines, making it difficult to meet the needs of dynamically changing product production. Moreover, due to the high production speed of rope nets, the delays caused by data transmission and centralized processing in high-speed online detection make it difficult to synchronize with the production rhythm when defects are detected, affecting the timeliness of defect location and subsequent processing.

[0005] In view of this, this application discloses an online detection method and system for rope and net defects based on image recognition. Summary of the Invention

[0006] The purpose of this invention is to provide an online detection method and system for rope and net defects based on image recognition, so as to solve the technical problem that the prior art cannot take into account both deviation risk and identification risk.

[0007] The specific technical solution adopted by this invention is as follows:

[0008] An online defect detection method for rope nets based on image recognition includes the following steps:

[0009] Export the correction area from the real-time image of the rope net surface to be processed;

[0010] Based on the real-time image of the rope net surface to be processed and the correction area, calculate the color adjustment value;

[0011] The real-time image of the rope net surface to be processed is corrected according to the color adjustment value to generate a corrected image as the real-time image of the rope net surface to be detected.

[0012] Acquire real-time images of the surface of the rope net to be inspected;

[0013] Perform defect detection processing on the real-time image of the rope net surface to be inspected;

[0014] The defect determination process for the real-time image of the rope net surface to be inspected includes: calculating the positional offset used to characterize the deviation of the rope net operation; transforming the real-time image of the rope net surface to be inspected based on the positional offset to generate an aligned image; calculating the difference between the aligned image and a preset reference image to generate a deviation image; and analyzing the suspected defective areas in the real-time image of the rope net surface to be inspected in conjunction with the deviation image to mark the suspected defective areas as defective areas or normal areas.

[0015] Preferably, the calculation of the color adjustment value based on the real-time image of the rope net surface to be processed and the correction area includes:

[0016] The evaluation path is extracted from multiple sampling nodes located in different grayscale gradient segments of the real-time image of the rope net surface to be processed; color adjustment values ​​are generated based on the evaluation path.

[0017] Preferably, the calculation of the positional offset used to characterize the deviation of the rope net operation includes:

[0018] A real-time image of the rope net surface that has been confirmed to be flawless and has caused the rope net operating equipment to trigger a stop signal is defined as an offset image; a preset feature is identified in the offset image; the displacement between the current position of the preset feature and the reference position defined based on the orientation direction is calculated to determine the position offset; when the position offset indicates that the operating state of the rope net is a deviation state, the position of the rope net is updated.

[0019] Preferably, the defect determination processing of the real-time image of the rope net surface to be inspected further includes:

[0020] Before analyzing suspected defect areas, the suspected defect areas are identified. The identification steps include: extracting color distribution areas from the real-time image of the rope net surface to be inspected; selecting reference points in the color distribution areas and emitting probe lines from the reference points in eight preset directions; determining the area to be calculated based on the probe lines and converting the area to be calculated into an interpolation area according to preset conversion rules, so that the interpolation area is defined as a suspected defect area; when the corner data of the interpolation area does not meet the preset sampling conditions, the reference point is reselected.

[0021] An online defect detection system for rope nets based on image recognition includes the following modules:

[0022] Image acquisition module, used to acquire real-time images of the rope net surface;

[0023] The image correction module is used to perform image correction processing on the real-time image of the rope net surface to generate a real-time image of the rope net surface to be inspected.

[0024] The offset calculation module is used to calculate the positional offset that characterizes the deviation of the rope net operation;

[0025] The defect detection module is used to perform defect detection processing on the real-time image of the rope net surface to be inspected.

[0026] The position control module is used to update the position of the rope net when the position offset indicates that the rope net is in a deviated state.

[0027] Preferably, the step of performing image correction processing on the real-time image of the rope net surface to generate a real-time image of the rope net surface to be detected includes:

[0028] The correction region is derived from the real-time image of the rope net surface to be processed; the color adjustment value is calculated based on the real-time image of the rope net surface to be processed and the correction region; the real-time image of the rope net surface to be processed is corrected according to the color adjustment value to generate a corrected image as the real-time image of the rope net surface to be detected.

[0029] Preferably, the calculation of the color adjustment value based on the real-time image of the rope net surface to be processed and the correction area includes:

[0030] The evaluation path is extracted from multiple sampling nodes located in different grayscale gradient segments of the real-time image of the rope net surface to be processed; color adjustment values ​​are generated based on the evaluation path.

[0031] Preferably, the calculation of the positional offset used to characterize the deviation of the rope net operation includes:

[0032] A real-time image of the rope net surface that has been confirmed to be flawless and has caused the rope net operating equipment to trigger a stop signal is defined as an offset image; a preset feature is identified in the offset image; the displacement between the current position of the preset feature and the reference position defined based on the orientation direction is calculated to determine the position offset.

[0033] Preferably, the defect determination processing of the real-time image of the rope net surface to be inspected includes:

[0034] Based on the calculated positional offset, the real-time image of the rope net surface to be inspected is transformed to generate an aligned image; the difference between the aligned image and the preset reference image is calculated to generate a deviation image; the deviation image is combined with the analysis of the suspected defect areas in the real-time image of the rope net surface to be inspected to mark the suspected defect areas as defect areas or normal areas.

[0035] Preferably, the defect determination processing of the real-time image of the rope net surface to be inspected further includes:

[0036] Before analyzing suspected defect areas, the suspected defect areas are identified. The identification steps include: extracting color distribution areas from the real-time image of the rope net surface to be inspected; selecting reference points in the color distribution areas and emitting probe lines from the reference points in eight preset directions; determining the area to be calculated based on the probe lines and converting the area to be calculated into an interpolation area according to preset conversion rules, so that the interpolation area is defined as a suspected defect area; when the corner data of the interpolation area does not meet the preset sampling conditions, the reference point is reselected.

[0037] Beneficial effects

[0038] In the defect detection process, this invention transforms the real-time image of the rope net surface to be detected based on the positional offset representing the deviation of the rope net operation to generate an aligned image. The difference between the aligned image and the preset reference image is calculated to generate a deviation image. The deviation image is then combined with the analysis of suspected defective areas to identify them as defective or normal areas. This compensates for the physical displacement of the rope net during production. Furthermore, by correcting before image comparison, false alarms and missed alarms caused by positional mismatch are eliminated, thereby improving the accuracy and reliability of defect detection.

[0039] This invention determines the operating status of the rope net based on the position offset. When the operating status is determined to be a deviation state, the position of the rope net is updated; when it is determined to be a normal state, the current position is maintained. This constructs a feedback mechanism that combines image processing and physical control, preventing the accumulation of rope net operating deviations and ensuring the long-term stability of the rope net detection system.

[0040] This invention identifies suspected defect areas by emitting probe lines from a real-time image of the surface of the rope net to be inspected, and analyzes the suspected defect areas in conjunction with deviation images to label them as defective or normal areas. By calculating color adjustment values ​​and fiber extension direction, image correction processing is performed. The use of probe lines makes the capture of local anomalies independent of the rope net's own texture, and the analysis of deviation images ensures the reliability of the labeling. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method of the present invention;

[0042] Figure 2 This is a flowchart of the defect identification method of the present invention;

[0043] Figure 3 This is a system module diagram of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely for explaining the invention and are not intended to limit the scope of protection of the invention.

[0045] Example 1

[0046] Please refer to Figures 1-2 This embodiment provides an online detection method for rope and net defects based on image recognition, applied to a high-speed rope and net production line, specifically including the following steps:

[0047] S1. Image acquisition: Real-time and continuous acquisition of images of the rope net surface through an image acquisition module deployed above the production line;

[0048] When the normal texture fluctuation or normal lateral displacement of the rope net accidentally triggers the stop signal of the rope net operating equipment, and after secondary verification by manual confirmation or backup sensor verification confirms that there are no actual defects, the current real-time image of the rope net surface is captured at this moment, and the real-time image of the rope net surface is defined as an offset image associated with the orientation direction of the rope net operation; wherein, the offset image defines the fault tolerance boundary, indicating the degree of image change that should not trigger a stop because it does not contain defects;

[0049] Further optimization of the method for acquiring real-time images of the rope net surface to enhance the detection dimensionality includes: acquiring images from multiple preset acquisition angles. In addition to the main camera perpendicular to the rope net surface, auxiliary cameras can be added on both sides of the rope net's running direction at specific angles such as 30 degrees or 45 degrees to the rope net surface; through a multi-angle acquisition strategy, three-dimensional defects such as yarn protrusions, knots, or depressions can be captured. These three-dimensional defects may be invisible and ignored from a single vertical viewpoint due to a lack of depth information; by fusing image information from multiple angles, more complete rope net surface data can be constructed, reducing the false negative rate.

[0050] S2. Calculate the position offset. During the production process, the rope net will shift laterally. To compensate for the lateral shift, identify the preset features from the acquired offset image. The preset features are stable and easily identifiable patterns in the texture of the rope net itself. Specifically, the preset features can be the interlacing points of yarns of a specific color or specific parts of the periodically appearing rope net pattern.

[0051] The current position of the preset feature in the current image frame is calculated through image processing processes such as template matching or feature point detection, and the displacement between it and the preset reference position in the ideal alignment state is calculated. The endpoint of the displacement is determined as the offset point. The preset reference position can be defined as the center line of the image acquisition field of view.

[0052] Based on the time series data of one or more offset points, the positional offset of the rope net is calculated by weighted averaging or linear regression analysis of the positions of multiple recent offset points; the positional offset is a quantitative value that serves as the basis for image alignment and physical position correction.

[0053] S3. Generate a reference image. When initializing or changing a batch of rope nets, obtain a rope net surface image that has been pre-calibrated to be flawless under ideal lighting and alignment conditions, and define it as a reference image.

[0054] Among them, the baseline image is the reference standard for all difference comparisons, representing the perfect and flawless surface state of the batch of rope nets; and the baseline image is a static reference standard, while the fine-tuning image and the recognition image are both dynamically acquired real-time images.

[0055] S4. Fine-tuning: Due to the slow changes in ambient light or the subtle differences between batches of rope dyeing, there may be global color and texture deviations between the real-time image and the reference image. To eliminate non-defect deviations, fine-tuning is performed.

[0056] The fine-tuning process specifically includes: acquiring a real-time image of the surface of the rope net to be processed and defining it as a fine-tuning image; deriving one or more correction regions from the fine-tuning image, where the correction region is a local area in the image that is presumed to be flawless, specifically the edge part of the image or the low-incidence area of ​​defects according to historical data statistics, as a sample for calculating the adjustment parameters;

[0057] Based on the fine-tuned image and the correction area, the color adjustment value and fiber extension direction are calculated; the evaluation path is extracted within the correction area. This evaluation path is a sampling path composed of multiple sampling nodes, which is used to track the micro-texture structure of the rope net; by calculating the average gray level and contrast on the sampling path, the distribution law of gray level values ​​on the sampling path is analyzed, and it is compared with the same parameters in the corresponding area of ​​the reference image to extract the color adjustment value used to compensate for illumination and color changes.

[0058] In this process, multiple sampling nodes are arranged in different gray-scale gradient segments of the fine-tuned image. Since the area where the gray-scale value changes usually corresponds to the edge of the fiber or yarn, the fiber direction can be fitted by connecting these sampling nodes to form a sampling path, thereby calculating the fiber extension direction.

[0059] The calculation of color adjustment values ​​and fiber extension direction can also be based on a preset mapping relationship between the fine-tuned image and the correction area. This preset mapping relationship can be a pre-established correspondence rule, which is applicable to rope net types with relatively fixed texture patterns. Specifically, the pre-established correspondence rule can be a lookup table that stores a large number of image statistical features such as average brightness, contrast, and texture direction histograms and their corresponding adjustment parameters, or a mathematical transformation rule used to calculate the output adjustment parameters based on the input statistical features.

[0060] After calculating the adjustment parameters, the color and brightness of the fine-tuned image are corrected according to the color adjustment value to generate the image to be corrected; the part of the image to be corrected that overlaps with the correction area is subjected to a preset update operation, which can be done by using local affine transformation to align the fiber texture and generate the corrected image.

[0061] The operating status of the rope net is determined by comparing the calculated position offset with a preset threshold, where the preset threshold is 1% of the rope net width. If the position offset is greater than the preset threshold, the operating status is determined to be a deviation state. A new correction area is generated based on the current correction image, and a control signal is output to adjust the tension or angle of the guide roller to return it to the center, thereby updating the physical position of the rope net. If the operating status is normal, the correction image is used as the output of this fine-tuning process for subsequent defect judgment processing, and the physical position of the rope net is not updated.

[0062] S5. Identify defects by acquiring a real-time image of the surface of the rope net to be inspected or a corrected image of the normal operating state, and define it as an identification image; in the identification image, extract the color distribution area whose color or texture differs from the surrounding background by a preset difference threshold by scanning.

[0063] Specifically, in a certain color distribution area, the point with the greatest difference between the color or grayscale value in that area and the average value of the surrounding pixels is selected as a reference point; with the reference point as the center, eight preset directions are emitted in the upward, downward, left, right and four diagonal directions, and the probe lines extend until the pixels on their path meet the preset background texture conditions; wherein, the preset background texture conditions are that the grayscale value, gradient or local texture features of the pixel are restored to the same level as the surrounding normal area;

[0064] The region enclosed by the termination points of all probe lines is defined as the region to be calculated. According to the preset transformation rules, specifically the convex hull construction method or morphological closure operation, the region to be calculated is transformed into a closed interpolation region with an internal filling. This interpolation region is initially defined as a suspected defect region. After defining the suspected defect region, in order to prevent image noise or accidentally attached small foreign objects from being misjudged as defects, a verification process is executed.

[0065] The verification process includes: extracting the corner data of the interpolation region, i.e., analyzing its geometric features such as perimeter, roundness, and number of corner points; determining whether the corner data meets the preset sampling conditions, which include requiring the area-to-perimeter ratio of the region to be greater than a certain threshold to exclude slender, broken, or noisy shapes;

[0066] If the corner data does not meet the preset sampling conditions, the initial reference point is considered to be an isolated noise point. The operation of reselecting the reference point will be performed to find a new and suboptimal reference point in the color distribution area and repeat the defect identification process.

[0067] After the reference point is verified, in order to make accurate comparison, the positional offset of the identified image is geometrically transformed to generate an aligned image that is spatially aligned with the reference image. The pixel-level difference between the aligned image and the reference image is calculated to generate a deviation image. The deviation image can highlight all the differences between the real-time image and the ideal flawless state.

[0068] The image data of the suspected defective area is compared and analyzed with the image data of the corresponding area in the deviation image. If the difference signal, such as the average pixel difference, displayed at the corresponding position of the suspected defective area in the deviation image is greater than the preset signal threshold, then the suspected defective area is finally marked as a defective area; otherwise, if the difference at the corresponding position in the deviation image is not greater than the preset signal threshold, then it is marked as a normal area.

[0069] After marking suspected defective areas as defective or normal areas, for all confirmed defective areas, the area of ​​the interpolated region is further extracted and analyzed to classify the defects. Specifically, different thresholds can be set according to the area size to classify defects into minor defects, general defects, and severe defects, and different responses can be triggered accordingly, such as logging only, issuing an audible and visual alarm to the operator, or immediately and automatically stopping the production line.

[0070] Example 2

[0071] Please refer to Figure 3 This embodiment provides an online defect detection system for rope nets based on image recognition, including the following modules:

[0072] The image acquisition module is used to acquire real-time images of the rope and net surface. In the specific production process, an industrial camera deployed above the rope and net production line continuously and uninterruptedly captures images of the high-speed running rope and net surface to generate a series of real-time images of the rope and net surface as raw data input.

[0073] The image correction module is used to perform image correction processing on the real-time image of the rope net surface acquired by the image acquisition module to generate a real-time image of the rope net surface to be detected. Due to uneven lighting or differences in the camera itself, the original image may have color or brightness deviations. The real-time image of the rope net surface is received as the real-time image of the rope net surface to be processed.

[0074] The correction region is derived from the real-time image of the rope net surface to be processed, which is, for example, a background area in the image that is considered to be standard and flawless; the color adjustment value is calculated based on the overall features of the real-time image of the rope net surface to be processed and the features of the correction region;

[0075] To calculate the color adjustment value, an evaluation path consisting of multiple sampling nodes located in different grayscale gradient segments of the real-time image of the rope net surface to be processed is extracted, and the color adjustment value is generated based on the evaluation path to ensure the global adaptability of the correction.

[0076] The real-time image of the rope net surface to be processed is corrected according to the color adjustment value to generate a corrected image with more uniform color and brightness. This corrected image is the real-time image of the rope net surface to be detected used by subsequent modules.

[0077] The offset calculation module is used to calculate the positional offset that characterizes the deviation of the rope net operation. Since the rope net may deviate laterally or longitudinally when it is running at high speed, the content of the acquired image will be translated.

[0078] The calculation of the position offset used to characterize the deviation of the rope net operation includes: defining the real-time image of the rope net surface that has been manually confirmed to be flawless but has historically triggered the rope net operation equipment to issue a stop signal due to its content as an offset image; calculating the displacement between the current position of the preset feature and the reference position such as the center line of the image defined based on the ideal orientation direction, and determining the displacement as the position offset;

[0079] Among them, the offset image represents a typical false alarm scenario caused by positional offset rather than actual defects. Pre-defined features such as unique texture points or periodic patterns on the rope network are identified in the offset image.

[0080] The defect detection module is used to perform defect detection processing, receive real-time images of the surface of the rope net to be inspected, and perform analysis operations.

[0081] Based on the position offset calculated by the offset calculation module, geometric transformations such as translation are performed on the real-time image of the rope net surface to be detected to generate an aligned image, thereby eliminating the influence of the rope net running deviation on the image position and aligning it with the standard position.

[0082] The difference between the aligned image and a preset reference image representing an ideal, flawless state is calculated to generate a deviation image, and areas that differ significantly from the reference image are highlighted or given specific values ​​in the deviation image.

[0083] Before making the final judgment, suspected defect areas are identified in the real-time image of the rope net surface to be inspected. Identifying suspected defect areas includes: extracting color distribution areas with abnormal color or texture in the image; selecting a reference point in the color distribution area, and emitting probe lines in eight preset directions (up, down, left, right, and four diagonals) centered on this reference point until the background area is encountered; determining the area to be calculated based on the endpoints of these probe lines, and converting the area to be calculated into an interpolation area according to preset transformation rules such as the convex hull algorithm or the minimum bounding rectangle algorithm, and defining the interpolation area as the suspected defect area.

[0084] If the corner data of the interpolation area does not meet the preset sampling conditions, the reference point is reselected and the defect judgment process is repeated to ensure the validity of the suspected defect area; the preset sampling conditions are that the area boundary exceeds the image range or is too irregular.

[0085] By combining the deviation image with the real-time image of the rope net surface to be detected, the suspected defect areas are comprehensively analyzed. Specifically, when only one suspected defect area corresponds spatially to the highlighted area in the deviation image, it is finally marked as a defect area; otherwise, it is marked as a normal area to eliminate false alarms caused by normal texture fluctuations.

[0086] The position control module receives the position offset generated by the offset calculation module to control the physical position of the rope net. When the position offset exceeds a preset threshold, indicating that the rope net is in a deviated state, it sends a control signal to the rope net operating equipment such as the servo motor of the correction device to adjust the rope net's transmission path, thereby updating the physical position of the rope net and returning it to the normal operating track.

[0087] Finally, it should be noted that the above examples are merely some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments and many variations are possible. All variations that can be directly derived or conceived by those skilled in the art from the disclosure of the present invention should be considered within the scope of protection of the present invention.

Claims

1. An online detection method for rope net defects based on image recognition, characterized in that, Includes the following steps: Export the correction area from the real-time image of the rope net surface to be processed; Based on the real-time image of the rope net surface to be processed and the correction area, calculate the color adjustment value; The real-time image of the rope net surface to be processed is corrected according to the color adjustment value to generate a corrected image as the real-time image of the rope net surface to be detected. Acquire real-time images of the surface of the rope net to be inspected; Perform defect detection processing on the real-time image of the rope net surface to be inspected; The defect determination process for the real-time image of the rope net surface to be inspected includes: calculating the positional offset used to characterize the deviation of the rope net operation; transforming the real-time image of the rope net surface to be inspected based on the positional offset to generate an aligned image; calculating the difference between the aligned image and a preset reference image to generate a deviation image; and analyzing the suspected defective areas in the real-time image of the rope net surface to be inspected in conjunction with the deviation image to mark the suspected defective areas as defective areas or normal areas.

2. The online detection method for rope net defects based on image recognition according to claim 1, characterized in that, The calculation of color adjustment values ​​based on the real-time image of the rope net surface to be processed and the correction area includes: The evaluation path is extracted from multiple sampling nodes located in different grayscale gradient segments of the real-time image of the rope net surface to be processed; color adjustment values ​​are generated based on the evaluation path.

3. The online detection method for rope and net defects based on image recognition according to claim 1, characterized in that, The calculation of the positional offset used to characterize the deviation of the rope net operation includes: A real-time image of the rope net surface that has been confirmed to be flawless and has caused the rope net operating equipment to trigger a stop signal is defined as an offset image; a preset feature is identified in the offset image; the displacement between the current position of the preset feature and the reference position defined based on the orientation direction is calculated to determine the position offset; when the position offset indicates that the operating state of the rope net is a deviation state, the position of the rope net is updated.

4. The online detection method for rope and net defects based on image recognition according to claim 1, characterized in that, The defect detection processing of the real-time image of the rope net surface to be inspected also includes: Before analyzing suspected defect areas, the suspected defect areas are identified. The identification steps include: extracting color distribution areas from the real-time image of the rope net surface to be inspected; selecting reference points in the color distribution areas and emitting probe lines from the reference points in eight preset directions; determining the area to be calculated based on the probe lines and converting the area to be calculated into an interpolation area according to preset conversion rules, so that the interpolation area is defined as a suspected defect area; when the corner data of the interpolation area does not meet the preset sampling conditions, the reference point is reselected.

5. An online defect detection system for rope nets based on image recognition, characterized in that, Includes the following modules: Image acquisition module, used to acquire real-time images of the rope net surface; The image correction module is used to perform image correction processing on the real-time image of the rope net surface to generate a real-time image of the rope net surface to be inspected. The offset calculation module is used to calculate the positional offset that characterizes the deviation of the rope net operation; The defect detection module is used to perform defect detection processing on the real-time image of the rope net surface to be inspected; The position control module is used to update the position of the rope net when the position offset indicates that the rope net is in a deviated state.

6. The online defect detection system for rope nets based on image recognition according to claim 5, characterized in that, The step of performing image correction processing on the real-time image of the rope net surface to generate a real-time image of the rope net surface to be detected includes: The correction region is derived from the real-time image of the rope net surface to be processed; the color adjustment value is calculated based on the real-time image of the rope net surface to be processed and the correction region; the real-time image of the rope net surface to be processed is corrected according to the color adjustment value to generate a corrected image as the real-time image of the rope net surface to be detected.

7. The online defect detection system for rope nets based on image recognition according to claim 6, characterized in that, The calculation of color adjustment values ​​based on the real-time image of the rope net surface to be processed and the correction area includes: The evaluation path is extracted from multiple sampling nodes located in different grayscale gradient segments of the real-time image of the rope net surface to be processed; color adjustment values ​​are generated based on the evaluation path.

8. The online defect detection system for rope nets based on image recognition according to claim 5, characterized in that, The calculation of the positional offset used to characterize the deviation of the rope net operation includes: A real-time image of the rope net surface that has been confirmed to be flawless and has caused the rope net operating equipment to trigger a stop signal is defined as an offset image; a preset feature is identified in the offset image; the displacement between the current position of the preset feature and the reference position defined based on the orientation direction is calculated to determine the position offset.

9. The online defect detection system for rope nets based on image recognition according to claim 8, characterized in that, The defect detection processing of the real-time image of the rope net surface to be inspected includes: Based on the calculated positional offset, the real-time image of the rope net surface to be inspected is transformed to generate an aligned image; the difference between the aligned image and the preset reference image is calculated to generate a deviation image; the deviation image is combined with the analysis of the suspected defect areas in the real-time image of the rope net surface to be inspected to mark the suspected defect areas as defect areas or normal areas.

10. The online defect detection system for rope nets based on image recognition according to claim 9, characterized in that, The defect detection processing of the real-time image of the rope net surface to be inspected also includes: Before analyzing suspected defect areas, the suspected defect areas are identified. The identification steps include: extracting color distribution areas from the real-time image of the rope net surface to be inspected; selecting reference points in the color distribution areas and emitting probe lines from the reference points in eight preset directions; determining the area to be calculated based on the probe lines and converting the area to be calculated into an interpolation area according to preset conversion rules, so that the interpolation area is defined as a suspected defect area; when the corner data of the interpolation area does not meet the preset sampling conditions, the reference point is reselected.

Images