A machine vision-based real-time monitoring method and system for wire winding quality of a concrete pipe

Abstract

The application provides a kind of real-time monitoring method and system for concrete pipe wire winding quality based on machine vision, relating to machine vision technical field, wherein the method comprises: first, obtaining the pretreatment image of pipe wire winding, extracting the steel wire skeleton contour point set by improving Hough transform, and calculating the spacing and angle data of adjacent steel wires using the triangulation algorithm. Then compare the measured spacing and angle with the standard value respectively to generate deviation data, and input the contour point set into the convolution neural network and transformer hybrid model to identify the broken wire and overlapping wire defect position. Finally, the spacing deviation data, angle deviation data and defect position data are transmitted in real time to the wire winding machine controller to dynamically adjust the wire feeding speed and tension parameters, realizing closed-loop optimization control of the wire winding process. The application improves the stability and defect detection rate of wire winding quality.

Description

Technical Field

[0001] This application relates to the field of machine vision technology, and in particular to a method and system for real-time monitoring of the wire winding quality of concrete pipes based on machine vision. Background Technology

[0002] Concrete pipes require wire winding to enhance structural strength during manufacturing, and the stability of the winding quality directly affects the pipe's durability and safety. Traditional production processes necessitate precise monitoring of wire spacing, winding angle, and defects such as broken or overlapping wires to ensure uniform winding and process compliance. However, manual sampling is inefficient and prone to omissions, necessitating an automated, high-precision real-time monitoring method capable of dynamically adjusting process parameters to improve winding quality.

[0003] Currently, some advanced production lines employ laser scanning-based wire winding quality inspection technology. This technology uses laser sensors to scan the pipe surface, acquiring three-dimensional point cloud data of the steel wires. Geometric algorithms are then used to calculate the wire spacing and angles, and traditional image processing techniques are employed to identify broken and overlapping wire defects. The inspection data is transmitted via wired connection to the control system, where operators manually adjust the wire winding machine parameters based on the results.

[0004] Laser scanning solutions are prone to point cloud data distortion under strong light or reflective conditions on pipe surfaces, leading to increased measurement errors. Traditional image processing methods have poor adaptability to steel wire textures and a high false detection rate on rough concrete surfaces. Furthermore, this solution relies on manual intervention to adjust process parameters, resulting in a lag in response and making it difficult to meet the real-time control requirements of high-speed production lines. Summary of the Invention

[0005] This application provides a machine vision-based method and system for real-time monitoring of the quality of wire wrapping in concrete pipes, in order to solve the problems of low stability and low defect detection rate of wire wrapping quality in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a machine vision-based method for real-time monitoring of the wire wrapping quality of concrete pipes, comprising: Obtain a preprocessed image containing the wire wrapping of the concrete pipe; An improved Hough transform is applied to the preprocessed image to extract the skeleton contour point set of the winding wire. Based on the skeleton contour point set, the spacing and angle data of adjacent winding wires are calculated using a triangulation algorithm. Based on the spacing data and the angle data, the winding wire spacing deviation data and angle deviation data are generated respectively. The skeleton contour point set is input into the hybrid model built based on convolutional neural network and transformer, and the defect location data of winding wire is output. The wire winding spacing deviation data, the angle deviation data, and the defect location data are input into the wire winding machine controller. The wire feeding speed parameters and wire feeding tension parameters are adjusted by the wire winding machine controller to optimize the quality of wire winding in concrete pipes based on the adjusted wire feeding speed parameters and adjusted wire feeding tension parameters.

[0007] Optionally, based on the spacing data and the angle data, the step of generating winding spacing deviation data and angle deviation data respectively, inputting the skeleton contour point set into a hybrid model constructed based on a convolutional neural network and a transformer, and outputting defect location data of the winding wire includes: The difference between the spacing data and the preset standard spacing value is used as the winding spacing deviation data; The difference between the angle data and the value of the preset standard angle is used as the angle deviation data; The skeleton contour point set is input into the hybrid model, and the convolutional neural network in the hybrid model sorts the discrete points in the skeleton contour point set that belong to the same winding wire according to the spatial adjacency relationship to form an ordered point chain. The spatial distribution of the ordered point chain is analyzed by the transformer in the hybrid model, and the defect location data of the entangled wire are output.

[0008] Optionally, the analysis of the spatial distribution of the ordered point chain and the output of defect location data of the entangled wire includes: The ordered dot chain is divided into multiple dot chain segments according to a preset segment length; Spatial continuity detection is performed on each point chain segment to generate candidate regions for broken wires and candidate regions for stacked wires; Interpolate the endpoint coordinates of all candidate areas for wire breakage to obtain the start and end coordinates of the wire breakage defect; Cluster the boundary points of all candidate areas for wire stacking to obtain the coordinate set of the wire stacking defect area; The start and end coordinates of the broken wire defect are merged with the coordinate set of the overlapping wire defect area to generate defect location data.

[0009] Optionally, performing spatial continuity detection on each point chain segment to generate candidate regions for broken and stacked wires includes: For each point chain segment, calculate the straight-line distance between adjacent discrete points to form a set of straight-line distances for the point chain segment; The deviation between each straight-line distance value in the set of straight-line distances and the arithmetic mean of the straight-line distance values ​​is calculated. When the deviation exceeds a preset fracture threshold, the region corresponding to the point chain segment is marked as a candidate region for broken wires. For adjacent point chain segments, extract the closest boundary points from each of the two adjacent point chain segments to form a set of point pairs; Calculate the ratio of the sum of distances between all point pairs in the point pair set to the number of point pairs to obtain the average point spacing value; When the average point spacing value is less than the preset wire stacking threshold, the region corresponding to the adjacent point chain segment is marked as a wire stacking candidate region.

[0010] Optionally, the step of performing an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the wound wires, and calculating the spacing and angle data of adjacent wound wires using a triangulation algorithm based on the skeleton contour point set, includes: An improved Hough transform is applied to the preprocessed image to detect the edge pixels of all the wires wrapped around the concrete pipe surface in the transformed image. Adjacent edge pixels belonging to the same winding wire are connected by linear interpolation to generate multiple skeleton lines, each skeleton line corresponding to one winding wire; Discrete points are sampled from each skeleton line according to a preset fixed step size, and the discrete points of all skeleton lines are merged to form a skeleton contour point set; In the set of skeleton contour points, a triangular mesh is constructed for the corresponding discrete points of adjacent winding wires; Calculate the arithmetic mean of the side lengths of all triangles in the triangular mesh, and use the arithmetic mean as the spacing data between adjacent windings; Calculate the angle values ​​of the interior angles of the triangles in the triangular mesh, and use these angle values ​​as the angle data for winding the wire.

[0011] Optionally, the step of inputting the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller, and adjusting the wire feeding speed parameters and wire feeding tension parameters through the wire winding machine controller, includes: The winding spacing deviation data, the angle deviation data, and the defect location data are input into the winding machine controller. The real-time position information of the winding wire is spatiotemporally aligned with the defect location data through the alignment module of the winding machine controller to generate a winding quality heat map. The coordinates of the defect-dense area are determined by the coordinate recognition module of the winding machine controller based on the winding quality heat map; When the coordinates of the defect-dense area repeat within N consecutive rotation cycles, the historical data of production process parameters within N rotation cycles are retrieved through the retrieval module of the winding machine controller. The wire feeding speed and wire feeding tension parameters are adjusted by the adjustment module of the wire winding machine controller based on the wire winding spacing deviation data, the angle deviation data, and the historical data of the production process parameters.

[0012] Optionally, acquiring the preprocessed image containing the wire wrapped around the concrete pipe includes: Acquire data from a ring-shaped multispectral light source, multi-angle optical images of concrete pipes, and data from a light sensor. Based on the light sensor data, the light source intensity of the ring multispectral light source data and the exposure parameters of the multi-angle optical image are dynamically adjusted to generate a stable light image; The stable illumination image is subjected to weighted grayscale equalization and adaptive threshold segmentation in sequence to obtain a preprocessed image that suppresses the interference of concrete surface texture.

[0013] Secondly, this application provides a machine vision-based real-time monitoring system for the quality of wire wrapping in concrete pipes, comprising: The acquisition module is used to acquire preprocessed images containing the wire wrapping of concrete pipes; The extraction module is used to perform an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the winding wire, and to calculate the spacing and angle data of adjacent winding wires based on the skeleton contour point set using a triangulation algorithm. The generation module is used to generate winding wire spacing deviation data and angle deviation data based on the spacing data and the angle data, respectively, input the skeleton contour point set into the hybrid model built based on convolutional neural network and transformer, and output the defect location data of winding wire. The input module is used to input the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller. The wire winding machine controller adjusts the wire feeding speed parameters and the wire feeding tension parameters to optimize the quality of the wire winding of the concrete pipe based on the adjusted wire feeding speed parameters and the adjusted wire feeding tension parameters.

[0014] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the machine vision-based real-time monitoring method for the wire winding quality of concrete pipes as described in the first aspect above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the machine vision-based real-time monitoring method for the wire winding quality of concrete pipes as described in the first aspect above.

[0016] This application provides a machine vision-based method for real-time monitoring of the quality of wire winding in concrete pipes. The method includes: acquiring a preprocessed image containing the wire winding of a concrete pipe; performing an improved Hough transform on the preprocessed image to extract a skeleton contour point set of the wire winding; calculating the spacing and angle data of adjacent wire windings using a triangulation algorithm based on the skeleton contour point set; generating wire winding spacing deviation data and angle deviation data based on the spacing and angle data; inputting the skeleton contour point set into a hybrid model constructed based on a convolutional neural network and a transformer to output defect location data of the wire winding; inputting the wire winding spacing deviation data, the angle deviation data, and the defect location data into a wire winding machine controller; and adjusting the wire feeding speed and tension parameters through the wire winding machine controller to optimize the quality of the wire winding in the concrete pipe based on the adjusted wire feeding speed and tension parameters.

[0017] The technical solution provided in this application has the following beneficial effects: This application achieves high-quality image acquisition of the wire-wrapped surface of concrete pipes, providing a clear and reliable visual data foundation for subsequent analysis. It accurately identifies the spatial distribution characteristics of the steel wires, overcoming interference from concrete surface textures and ensuring the integrity of contour extraction. It precisely quantifies the geometric relationships between adjacent steel wires, providing quantifiable measurement basis for process evaluation. It establishes a quantitative comparison between actual process parameters and standard values, intuitively reflecting the degree of deviation in the wire-wrapping process. It comprehensively identifies complex defects such as broken wires and overlapping wires, improving the comprehensiveness and accuracy of defect detection. It achieves real-time closed-loop control of the wire-wrapping process, ensuring dynamic optimization of quality parameters during production.

[0018] Furthermore, this application also obtains deviation data by calculating the difference between the measured spacing and angle and the standard value; after inputting the skeleton point set into the hybrid model, the convolutional neural network first spatially sorts the discrete points to form an ordered point chain, and then the spatial distribution characteristics of the point chain are analyzed by the transformer to identify the defect location.

[0019] Furthermore, quantitative deviation data provides a precise basis for process adjustment; the construction of ordered point chains enhances the coherence of feature expression; spatial distribution analysis improves the ability to identify complex defects; and overall, it achieves a multi-dimensional and accurate assessment of winding quality.

[0020] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a machine vision-based real-time monitoring method for the wire wrapping quality of concrete pipes, provided as an embodiment of this application; Figure 2 A schematic diagram illustrating a specific implementation of a machine vision-based real-time monitoring method for the wire wrapping quality of concrete pipes, as provided in this application embodiment; Figure 3 This is a schematic diagram illustrating another specific implementation of a machine vision-based method for real-time monitoring of the wire wrapping quality of concrete pipes, as provided in this application. Figure 4 This is a schematic diagram of a machine vision-based real-time monitoring system for the wire winding quality of concrete pipes, provided in an embodiment of this application. Detailed Implementation

[0023] In the field of concrete pipe wire winding quality inspection, existing laser scanning solutions have significant shortcomings: scanning data is easily distorted under strong light, leading to inaccurate measurements of wire spacing and angle; traditional image processing methods have poor adaptability to the rough texture of concrete surfaces, resulting in a high false judgment rate; the entire inspection process relies on manual intervention to adjust equipment parameters, making it difficult to achieve real-time and precise control of the production line. These problems essentially stem from the insufficient resistance of the inspection methods to environmental interference and the lack of an intelligent closed-loop control mechanism.

[0024] To address these issues, this application proposes a machine vision-based method for real-time monitoring of the wire winding quality of concrete pipes. This method acquires high-resolution images of the pipe surface through multi-angle optical imaging, accurately extracts the distribution characteristics of the steel wire using advanced image analysis technology, and automatically identifies various defects using intelligent algorithms. The key innovation lies in feeding the detection results back to the control system in real time, automatically adjusting the operating parameters of the wire winding equipment. This method effectively overcomes interference from strong light and surface texture, achieving three major improvements: enhanced measurement accuracy, more reliable defect identification, and fully automated control of the production process. By establishing a closed-loop system of "detection-analysis-control," the problems of inaccurate measurement, high misjudgment rate, and lagging control in existing technologies are fundamentally solved, ensuring the stability and consistency of the wire winding quality.

[0025] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] The core of this application is to provide a machine vision-based method for real-time monitoring of the quality of wire wrapping in concrete pipes. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes: Step 101: Obtain a preprocessed image containing the wire wrapping of the concrete pipe.

[0027] In step 101, the preprocessed image refers to the image of the concrete pipe surface after preliminary processing, which eliminates interference such as uneven lighting and reflections, while retaining clear wire wrapping features. This image serves as the basis for subsequent analysis. The wire wrapping of the concrete pipe refers to the steel wire skeleton structure wrapped around the pipe surface to enhance pipe strength.

[0028] In this embodiment, a layer of neat cement slurry should be sprayed onto the pipe surface before each wire winding. The cement used in the slurry should be the same as that used in the core concrete. The water-cement ratio of the neat cement slurry should be 0.6~0.7, and the coating amount should be controlled at 0.4 L / m~0.5 L / m. During the wire winding process, if wire lap splicing is required, the tensile force that the wire joint can withstand should at least reach the standard value of the ultimate tensile strength of the wire, and tight winding is not allowed. The wire winding machine should be equipped with a stress display device or stress recording device that can continuously record the tensile stress of the wire. The fluctuation range of the tensile stress deviating from the average value during the wire winding process should not exceed ±10%. During the wire winding process, the system will identify and count the number of wire joints in real time. Once the total number of joints exceeds the preset allowable joint number threshold, the wire winding machine will automatically trigger an audible and visual warning or display screen warning to remind the operator to check and handle it in time to ensure that the number of joints meets the specifications and avoid affecting the overall quality and structural safety of the wire winding due to too many joints. Then, the original image of the pipe surface is acquired by an industrial camera, and then the image is processed by illumination compensation and noise filtering to enhance the contrast of the wire winding area. Next, an edge enhancement algorithm is used to highlight the outline of the steel wire, ultimately obtaining a preprocessed image that can be used for feature extraction. The entire process ensures that the entangled wire features in the image are clearly discernible, providing reliable input for subsequent steps.

[0029] For example, on a concrete pipe factory production line, an industrial camera of brand A is used to photograph the surface of pipes being wound with wire. Due to uneven lighting, the original image is first processed for brightness equalization, and then algorithm B is used to eliminate interference from the concrete surface texture, ultimately obtaining a clear pre-processed image of the wire-wound pipes. The brightness equalization process uses the formula Y=0.299R+0.587G+0.114B for grayscale conversion, where R, G, and B represent the red, green, and blue channel values, respectively, and Y represents the output brightness value.

[0030] Step 102: Perform an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the winding wire. Based on the skeleton contour point set, use a triangulation algorithm to calculate the spacing and angle data of adjacent winding wires.

[0031] In step 102, the improved Hough transform is an optimized straight line detection method that can accurately identify the centerline position of the winding wire in the image. The skeleton contour point set is a series of coordinate points representing the centerline position of the winding wire. The triangulation algorithm is used to construct a triangle network between the contour points of adjacent winding wires, facilitating geometric measurement. The spacing data refers to the average distance between two adjacent winding wires calculated by the triangulation algorithm, specifically expressed as the arithmetic mean of the side lengths of all triangles in the constructed triangle network, used to quantify the deviation between the actual spacing of the winding wires and the standard requirements. The angle data refers to the angle between the winding wire and the pipe axis, reflecting whether the actual angle of the wire winding meets the process specification requirements.

[0032] In this embodiment, an improved Hough transform is applied to the preprocessed image to detect the center line of the wound wire, obtaining preliminary contour points. These points are then sorted and connected according to the direction of the wound wire to form a complete skeleton line. Next, corresponding points are selected on the skeleton lines of adjacent wound wires to construct a triangle network. Spacing data is obtained by calculating the side lengths of the triangles, and angle data is obtained by measuring the interior angles of the triangles.

[0033] For example, following the previous example, an improved Hough transform was used to detect the skeleton lines of five winding wires in the preprocessed image. A triangle network was constructed by selecting 30 corresponding points from each of two adjacent winding wires, resulting in a total of 150 triangles. The average side length of all triangles was calculated as the spacing data, and the angle between the triangles and the pipe axis was measured as the angle data.

[0034] Step 103: Based on the spacing data and the angle data, generate winding spacing deviation data and angle deviation data respectively, input the skeleton contour point set into the hybrid model built based on convolutional neural network and transformer, and output the defect location data of winding wire.

[0035] In step 103, the spacing deviation data represents the difference between the measured spacing and the standard spacing. The angle deviation data represents the difference between the measured angle and the standard angle. The defect location data includes the specific location information of defects such as broken wires and overlapping wires.

[0036] In this embodiment, the measured spacing and angle are compared with preset standard values ​​to calculate the deviation data. Simultaneously, the skeleton point set is input into a hybrid model. The model first spatially sorts the discrete points to form an ordered point chain, then analyzes the distribution characteristics of the point chain to identify the defect location. The final output includes detection results containing coordinates of various defects.

[0037] For example, the measured average spacing was 5.2 mm, which is +0.2 mm deviation from the standard value of 5.0 mm. The measured angle was 53 degrees, which is -2 degrees deviation from the standard of 55 degrees. After analyzing the point chain using the hybrid model, a broken wire was found in the third winding, located 1.5 m from the end of the pipe.

[0038] Step 104: Input the wire winding spacing deviation data, the angle deviation data, and the defect location data into the wire winding machine controller. Adjust the wire feeding speed parameters and wire feeding tension parameters through the wire winding machine controller to optimize the quality of wire winding in concrete pipes based on the adjusted wire feeding speed parameters and adjusted wire feeding tension parameters.

[0039] In step 104, the wire winding machine controller is the core control unit of the wire winding machine, deployed within the machine body, and directly controls the machine's mechanical actuators. The wire feeding speed parameter controls the speed of wire delivery. The wire feeding tension parameter controls the tightness of wire winding. Both the wire feeding speed and tension parameters are real-time control command parameters from the wire winding machine controller, dynamically generated by the controller using a PID control algorithm based on received data on wire spacing deviation, angle deviation, and defect location.

[0040] In this embodiment, deviation data and defect information are transmitted to the controller via the OPC-UA protocol. The controller calculates the process parameters that need to be adjusted according to preset rules. When the spacing is too large, the wire feeding speed is increased; when the angle is too small, the tension is increased. When a defective area is detected, it is automatically marked and an alarm is triggered. The adjusted parameters are immediately applied to the winding equipment to achieve closed-loop control.

[0041] For example, for the aforementioned +0.2mm spacing deviation, the controller will increase the wire feeding speed by 5%. For the -2 degree angle deviation, the tension will be increased by 3%. After detecting a wire breakage defect, the system automatically records the location and triggers an audible and visual alarm, while simultaneously reducing the wire winding speed in that area.

[0042] This method achieves fully automated monitoring and control of the quality of wire wrapping in concrete pipes. It ensures measurement accuracy through precise image analysis, enhances defect identification capabilities through intelligent algorithms, and optimizes process parameters through real-time feedback control. Ultimately, it improves the stability and consistency of wire wrapping quality, reduces manual intervention, and lowers the defect rate.

[0043] To address the issue of insufficient defect identification accuracy in the quality monitoring of wire winding in concrete pipes, in some embodiments, step 103 involves generating wire winding spacing deviation data and angle deviation data based on the spacing data and angle data, inputting the skeleton contour point set into a hybrid model constructed based on a convolutional neural network and a transformer, and outputting defect location data for the wire winding, including: Step 201: The difference between the spacing data and the preset standard spacing is taken as the winding spacing deviation data.

[0044] In step 201, the preset standard spacing refers to the theoretical design distance between adjacent windings specified by the production process. This value is determined based on the structural strength requirements of the pipeline. For example, the standard spacing of a certain type of concrete pipeline is set at 5.0 mm, which is the optimal winding spacing obtained through mechanical calculations.

[0045] In this embodiment of the application, the actual spacing value calculated in step 102 is subtracted from the preset standard spacing, and the difference is the winding spacing deviation data, which is used for subsequent process parameter adjustment.

[0046] Step 202: The difference between the angle data and the preset standard angle is taken as the angle deviation data.

[0047] In step 202, the preset standard angle refers to the theoretical angle between the winding wire and the pipe axis as required by the process specifications. This angle directly affects the wrapping effect of the winding wire. For example, the standard winding angle of a certain production line is set at 55 degrees, which is the optimal angle determined through multiple process tests.

[0048] In this embodiment of the application, the angle measurement value obtained in step 102 is subtracted from the standard angle to obtain angle deviation data, which will be used for automatic adjustment of winding tension.

[0049] Step 203: Input the skeleton contour point set into the hybrid model, and sort the discrete points in the skeleton contour point set that belong to the same winding wire according to the spatial adjacency relationship through the convolutional neural network in the hybrid model to form an ordered point chain.

[0050] In step 203, spatial adjacency refers to the actual arrangement order of discrete points on the same winding wire in three-dimensional space. Specifically, it is represented by the positional relationship of continuous points along the extension direction of the winding wire, and the order is determined by calculating the relative positions between the points. An ordered point chain refers to a coherent sequence formed by rearranging the discrete contour points of a single winding wire according to the actual spatial distribution order. This structure preserves the spatial continuity characteristics of the winding wire, which is convenient for subsequent defect analysis.

[0051] In this embodiment, the convolutional neural network in the hybrid model first identifies discrete points belonging to the same winding wire, and then sorts them according to their actual distribution positions on the pipe surface to form a point chain structure with spatial coherence.

[0052] Step 204: Analyze the spatial distribution of the ordered point chain using the transformer in the hybrid model, and output the defect location data of the entangled wire.

[0053] In step 204, the spatial distribution of the ordered point chain refers to the geometric arrangement characteristics of the sorted entangled point set on the pipe surface, including the density of points, spacing changes, and other patterns. These characteristics are obtained by analyzing and calculating the relative positional relationship of the coordinates of each point in the ordered point chain.

[0054] In this embodiment, the converter module analyzes the distribution density and spacing variation of points in the ordered point chain. When a segment of the point chain exhibits an abnormally sparse or dense distribution, it is marked as a defect of the corresponding type and its position coordinates are recorded.

[0055] Here is a specific example: On the concrete pipe winding production line, based on the preprocessed images and measurement data obtained from previous steps, the system first compares the measured average spacing of 5.2 mm with the preset standard spacing of 5.0 mm, calculating the winding spacing deviation as +0.2 mm. This value is calculated using the formula Δd = d_measured - d_standard, where Δd represents the spacing deviation, d_measured represents the actual measured value of 5.2 mm, and d_standard represents the preset value of 5.0 mm. Simultaneously, the measured winding angle of 53 degrees is compared with the standard angle of 55 degrees, calculating the angle deviation as -2 degrees, using the same difference formula. Subsequently, the contour point set containing five winding skeleton lines is input into the hybrid model. The model first uses a convolutional neural network to spatially sort the discrete points of each winding line. For example, the point set of the third winding line is arranged in order from the pipe end to the middle as point 1 (at 1.45 meters), point 2 (at 1.48 meters), etc., forming an ordered point chain. When the converter module analyzed these point chains, it found that the third winding wire was abnormal in the range of 1.50 meters to 1.55 meters. The distance between three consecutive points in this segment of the point chain suddenly increased to twice the normal value. According to the judgment rules, it was confirmed as a wire breakage defect, and the start and end coordinates of the defect were recorded as 1.50 meters and 1.55 meters. At the same time, it was found that the average distance between the fourth and fifth winding wires at 2.10 meters was only 3.5 millimeters, which was less than the preset wire overlap judgment threshold of 3.8 millimeters, and was marked as a wire overlap defect area.

[0056] In the embodiments of this application, the method realizes multi-dimensional and accurate evaluation of wire winding quality, provides a basis for process adjustment by quantifying deviation data, enhances the accuracy of defect identification by the ordered point chain structure, and finally realizes intelligent monitoring and automatic control of wire winding quality.

[0057] To further improve the accuracy of defect detection, in some embodiments, step 204: analyzing the spatial distribution of the ordered point chain and outputting the defect location data of the tangled wire includes: Step 301: Divide the ordered point chain into multiple point chain segments according to a preset segment length.

[0058] In step 301, a point chain segment refers to a local segment of a complete ordered point chain divided into segments of fixed length. Each segment contains several discrete points arranged in a continuous manner and is used for local feature analysis.

[0059] In this embodiment of the application, the system automatically divides the ordered point chain of each winding wire into several segments of equal length to ensure that each segment can independently reflect the local winding wire state.

[0060] Step 302: Perform spatial continuity detection on each point chain segment to generate candidate regions for broken wires and candidate regions for stacked wires.

[0061] In step 302, the broken wire candidate area refers to the region within a dot chain segment where the distance between dots is abnormally increased. The overlapping wire candidate area refers to the region where the distance between dots between adjacent segments is abnormally decreased; both are areas to be confirmed as suspected defects.

[0062] In the embodiments of this application, the degree of fluctuation of the internal point distance is calculated for each point chain segment, and if the fluctuation exceeds the limit, it is marked as a candidate area for broken wire; at the same time, the boundary point distance of adjacent segments is compared, and if it is too small, it is marked as a candidate area for overlapping wire.

[0063] Step 303: Interpolate the endpoint coordinates of all candidate areas for broken wires to obtain the start and end coordinates of the broken wire defect.

[0064] In step 303, the start and end coordinates of the broken wire defect refer to the starting and ending positions of the fracture area determined by interpolation calculation, which accurately defines the defect range.

[0065] In this embodiment of the application, the coordinates of the first and last two points of the candidate area of ​​broken wire are taken, and several intermediate points are evenly inserted in between to form a complete coordinate description of the defect area.

[0066] Step 304: Perform boundary point clustering on all candidate areas of wire stacking to obtain the coordinate set of the wire stacking defect area.

[0067] In step 304, the coordinate set of the wire overlap defect area refers to the set of boundary points of the wire overlap area determined by cluster analysis, which reflects the spatial distribution of the overlap range.

[0068] In this embodiment, discrete points within the candidate region of the stacked wire are grouped according to spatial density, dense points are merged to form a closed region, and the region boundary coordinates are recorded.

[0069] Step 305: Merge the start and end coordinates of the broken wire defect with the coordinate set of the overlapping wire defect area to generate defect location data.

[0070] In this embodiment, the spatial distribution of the ordered point chain is represented by terms such as point chain segments, broken wire candidate areas, and overlapping wire candidate areas. Point chain segments characterize the spatial segmentation state of the point chain, broken wire candidate areas reflect the abnormal spatial continuity distribution within the point chain, and overlapping wire candidate areas reflect the spatial overlap distribution characteristics between adjacent point chain segments. The system merges and stores the start and end coordinates of all broken wires and the area coordinates of overlapping wires, forming structured data containing spatial location and defect type.

[0071] Here is a specific example: During the quality inspection of wire wrapping in concrete pipes, based on the ordered point chain data generated in the previous steps, the system first divides the point chain of the third wire wrapping into several segments every 0.5 meters. The point chain between 1.45 meters and 1.95 meters is further divided into continuous segments such as 1.45-1.50 meters and 1.50-1.55 meters. When inspecting each segment, the system calculates the fluctuation in the spacing between adjacent points within the 1.50-1.55 meter segment. The maximum spacing of 3.2 mm is twice the minimum spacing of 1.6 mm, exceeding the preset fluctuation threshold of 1.8 times. Therefore, this segment is marked as a candidate area for wire breakage. Simultaneously, the adjacent segments of the fourth and fifth wire wrappings at 2.10 meters are inspected. The average spacing between the points at the junction is calculated to be 3.5 mm, less than the wire overlap threshold of 3.8 mm, and is marked as a candidate area for wire overlap. For candidate areas of broken wires, the system inserts intermediate coordinate points such as 1.51 meters and 1.52 meters between the 1.50-meter starting point and the 1.55-meter ending point to form a complete description of the start and end coordinates of the broken wire. For candidate areas of overlapping wires, the system groups 35 boundary points near 2.10 meters that are too closely spaced according to density, and merges them to form a coordinate set of overlapping wire areas centered at 2.10 meters and covering a range of 2.08-2.12 meters. Finally, the system merges and stores the coordinates of the broken wires from 1.50 to 1.55 meters with the coordinates of the overlapping wire areas from 2.08 to 2.12 meters, generating a defect report containing specific location and range.

[0072] In the embodiments of this application, the method achieves precise defect localization through segmented detection and spatial analysis, and the identification range of broken wires and overlapping wires is more accurate and complete, providing a reliable basis for quality control and process optimization.

[0073] To further improve the accuracy of defect detection, in some embodiments, step 302: performing spatial continuity detection on each point chain segment to generate broken wire candidate regions and stacked wire candidate regions includes: Step 401: For each point chain segment, calculate the straight-line distance between adjacent discrete points to form a set of straight-line distances for the point chain segments.

[0074] In step 401, the straight-line distance set refers to the set of straight-line distance values ​​between all adjacent points within a single point chain segment, which is used to quantify the uniformity of point distribution within the segment.

[0075] In this embodiment, the system sequentially calculates the straight-line distance between every two adjacent discrete points in the point chain segment, stores all distance values ​​in order to form a distance set, and provides basic data for subsequent continuity analysis.

[0076] Step 402: Calculate the deviation between each straight-line distance value in the set of straight-line distances and the arithmetic mean of the straight-line distance values. When the deviation exceeds a preset fracture threshold, mark the region corresponding to the point chain segment as a candidate region for broken wires.

[0077] In step 402, the deviation measure refers to the degree of dispersion of each distance value in the distance set relative to the average distance, reflecting the uniformity of the point distribution. The larger the value, the more drastic the fluctuation of the point distance.

[0078] In this embodiment of the application, the system first calculates the average value of the distance set, then calculates the sum of squares of the deviations of each distance value from the average value, and finally divides by the number of distances to obtain the degree of deviation. When the value exceeds a preset threshold, it is determined to be abnormal.

[0079] Step 403: For adjacent point chain segments, extract the closest boundary points from each of the two adjacent point chain segments to form a set of point pairs.

[0080] In step 403, the closest boundary points represent the set of end boundary points of the first point chain segment (denoted as set A) and the set of start boundary points of the second point chain segment (denoted as set B). A point pair set refers to a set of pairs of points that are closest to each other in the boundary regions of two adjacent point chain segments, used to evaluate the relative positional relationship between segments.

[0081] In the embodiments of this application, the system extracts several points from the ends and beginnings of two adjacent segments respectively, and finds the nearest point in another segment for each point to form a one-to-one point pair combination.

[0082] Step 404: Calculate the ratio of the sum of the distances between all point pairs in the point pair set to the number of point pairs to obtain the average point spacing value.

[0083] In step 404, the average point spacing value refers to the arithmetic mean of the distances between all point pairs in the point pair set, reflecting the average interval distance between two segments.

[0084] In this embodiment of the application, the system adds up the distances of all point pairs and divides the sum by the number of point pairs to obtain an average value that characterizes the tightness between segments.

[0085] Step 405: When the average point spacing value is less than the preset wire stacking threshold, mark the region corresponding to the adjacent point chain segment as a wire stacking candidate region.

[0086] In this embodiment of the application, when the calculated average point spacing is less than the minimum safe distance allowed by the process, the system marks the adjacent segment area as a potential wire stacking defect area.

[0087] Here is a specific example: During the quality inspection of wire wrapping in concrete pipes, when the system performs detailed inspection on the pre-defined point chain segments, it first selects a segment of 1.50-1.55 meters to calculate the distance between adjacent points within it, obtaining distance sets of 1.6 mm, 1.7 mm, 3.2 mm, etc., with 3.2 mm being an outlier. The system calculates the average of these distances to be 2.1 mm, and then calculates the sum of squares of the deviation of each distance from the average, using the formula Σ(d_i-2.1)^2, where d_i represents each distance value. The calculated deviation reaches 0.25 square millimeters, exceeding the preset fracture threshold of 0.2 square millimeters, therefore this segment is marked as a candidate area for wire breakage. Simultaneously, the system detects two segments of the adjacent 4th and 5th winding wires at 2.10 meters. From each segment, 10 of the closest boundary points are selected to form 10 point pairs. The distances of these point pairs are calculated to be 3.4 mm, 3.5 mm, etc. After summing, the average point spacing is 3.5 mm, which is less than the 3.8 mm wire stacking threshold. Therefore, it is marked as a wire stacking candidate area.

[0088] In this embodiment of the application, the method achieves accurate prediction of wire breakage and wire overlap defects by quantitatively analyzing the spatial distribution characteristics within the point chain and between segments, providing reliable candidate regions for subsequent accurate coordinate calculation, and improving the efficiency and accuracy of defect identification.

[0089] To further improve the accuracy of measuring the geometric parameters of the wound wire, in some embodiments, step 102: performing an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the wound wire, and calculating the spacing and angle data of adjacent wound wires using a triangulation algorithm based on the skeleton contour point set, includes: Step 501: Perform an improved Hough transform on the preprocessed image to detect the edge pixels of all the wires wrapped around the concrete pipe surface from the transformed image.

[0090] In step 501, edge pixels refer to the pixel positions at the junction of the steel wire and the concrete background in the image, and these points together outline the edge contour of the wire.

[0091] In this embodiment of the application, the system uses an optimized line detection method to find all possible winding edge points from the preprocessed image. Although these points are scattered, they can reflect the general direction of the winding.

[0092] Step 502: Connect adjacent edge pixels belonging to the same winding wire through linear interpolation to generate multiple skeleton lines, each skeleton line corresponding to one winding wire.

[0093] In step 502, the skeleton line refers to a continuous line formed by connecting the edge points of the same winding wire, representing the center direction path of the winding wire.

[0094] In this embodiment of the application, the system first sorts the edge points belonging to the same winding wire according to their spatial positions, and then inserts a new intermediate point between two adjacent points to make them smoothly connected, and finally forms a continuous skeleton line.

[0095] Step 503: Sample discrete points from each skeleton line according to a preset fixed step size, merge the discrete points of all skeleton lines to form a skeleton contour point set.

[0096] In step 503, the preset fixed step size refers to the fixed interval distance used when uniformly selecting discrete points along the skeleton line. This distance is determined according to the detection accuracy requirements, for example, set to 1 / 4 of the wire diameter to ensure a moderate sampling density. Discrete points refer to representative position points selected from the skeleton line according to the fixed step size. These points retain the shape characteristics of the skeleton line while reducing the amount of data, and are used for subsequent triangulation and geometric parameter calculation.

[0097] In this embodiment, the system selects discrete points along each skeleton line with a fixed step size, which preserves the shape characteristics of the line while reducing the amount of data. All the sampling points of the winding wire are merged to form a complete point set.

[0098] Step 504: In the set of skeleton contour points, construct a triangular network for the corresponding discrete points of adjacent winding wires.

[0099] In step 504, the triangular network refers to the triangular network constructed between corresponding points of adjacent windings, which is used to establish a spatial relationship model between windings.

[0100] In this embodiment, the system selects points corresponding to the positions on two adjacent winding wires as vertices, constructs multiple interconnected triangles, and forms a triangular mesh covering the surface of the winding wires.

[0101] Step 505: Calculate the arithmetic mean of the side lengths of all triangles in the triangular mesh, and use the arithmetic mean as the spacing data of adjacent winding wires.

[0102] In this embodiment of the application, the system measures the side lengths of all triangles in the triangular mesh and calculates the average of these side lengths as the typical spacing between two winding wires.

[0103] Step 506: Calculate the angle values ​​of the interior angles of the triangles in the triangular mesh, and use the angle values ​​as the angle data for winding the wire.

[0104] In this embodiment of the application, the system measures the angle between the triangle in the triangular mesh and the pipe axis, and obtains the average winding angle of the wire by comprehensive analysis.

[0105] Here is a specific example: During the quality inspection of wire-wrapped concrete pipes, the system analyzes the pre-processed pipe surface image. First, an improved straight-line detection method is used to identify all edge points of the wire wrapping, detecting a total of 1260 edge pixels from 5 wires. After grouping these edge points according to their respective wires, the system inserts 3 equidistant points between adjacent edge points and connects them using linear interpolation to form 5 smooth skeleton lines. The skeleton line of the first wire wrapping contains 215 consecutive points arranged from the pipe end. A representative point is selected every 0.5 mm from each skeleton line, with this step size set to one-quarter of the wire diameter (2 mm), ultimately obtaining a skeleton contour point set containing 860 discrete points. From this point set, 30 corresponding points from each of two adjacent wire wrappings are selected to construct a triangular mesh composed of 150 triangles. The average side length of the triangles is calculated using the formula L_avg=Σ(a+b+c) / 3n, where a, b, and c represent the lengths of the three sides of the triangle, and n=150 is the total number of triangles. The calculated average spacing is 5.2 mm. At the same time, measure the included angle formed by the side of each triangle that is parallel to the pipe axis, and take the median of all included angles, 53 degrees, as the winding angle.

[0106] In the embodiments of this application, the method achieves high-precision measurement of the winding geometry parameters through precise point-to-line conversion and spatial grid construction, providing a reliable basis for spacing and angle data for quality assessment.

[0107] To further improve the accuracy of the wire winding process control, in some embodiments, step 104: inputting the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller, and adjusting the wire feeding speed parameters and wire feeding tension parameters through the wire winding machine controller, includes: Step 601: Input the winding spacing deviation data, the angle deviation data, and the defect location data into the winding machine controller. The real-time position information of the winding wire is spatiotemporally aligned with the defect location data through the alignment module of the winding machine controller to generate a winding quality heat map.

[0108] In step 601, the wire winding quality heat map is a visual distribution map that marks the defect location with the corresponding rotation position of the pipe, reflecting the spatial aggregation of defects on the pipe surface.

[0109] In this embodiment of the application, the system maps the detected defect location onto the circumferential coordinates of the pipe based on the pipe rotary encoder signal, and forms an intuitive quality distribution map by marking the defect density with different colors.

[0110] Step 602: Determine the coordinates of the defect-dense area based on the wire winding quality heat map using the coordinate recognition module of the wire winding machine controller.

[0111] In step 602, the coordinates of the defect-dense area refer to the pipe location range where defect points are concentrated in the heat map, which usually manifests as an abnormal area that repeats for multiple consecutive rotation cycles.

[0112] In this embodiment of the application, the system analyzes the distribution density of defect points in the heat map. When the number of defects per unit area in a certain region exceeds a set threshold, the system records the starting and ending angle coordinates of that region.

[0113] Step 603: When the coordinates of the defect-dense area repeat within N consecutive rotation cycles, the historical data of production process parameters within N rotation cycles are retrieved through the retrieval module of the winding machine controller.

[0114] In step 603, the historical data of production process parameters is retrieved from the process parameter database of the winding machine. This database continuously records the time sequence data of process parameters such as tension and speed during the winding process. The historical data of production process parameters refers to the set of records of the operating parameters of the winding machine, such as speed and tension, within the time period corresponding to the defect area.

[0115] In this embodiment of the application, the system extracts the change curves of parameters such as wire feeding speed and tension during the time period from the control database based on the time stamp of the defect area.

[0116] Step 604: Adjust the wire feeding speed parameter and wire feeding tension parameter by means of the adjustment module of the wire winding machine controller, based on the wire winding spacing deviation data, the angle deviation data and the historical data of the production process parameters.

[0117] In step 604, parameter adjustment refers to the control process of dynamically optimizing the operating parameters of the winding machine based on deviation analysis and comparison with historical data.

[0118] In this embodiment, the system integrates current deviation data and historical parameter trends, and uses an incremental adjustment strategy to gradually correct the wire feeding speed and tension until the quality indicators return to the normal range.

[0119] Here is a specific example: On the concrete pipe winding production line, the system inputs data on the detected +0.2 mm spacing deviation, -2 degree angle deviation, and wire breakage at 1.5 meters into the winding machine controller. The controller first obtains the real-time position of the pipe via a rotary encoder, converting the 1.5-meter wire breakage location into a circumferential angular coordinate of 120 degrees, generating a mass heatmap including the defect point. Analysis revealed anomalies in the 120-degree area over three consecutive rotation cycles, marking it as a defect-dense area. The system retrieved process parameter records for these three cycles, finding that the wire feeding speed fluctuated between 5.1 and 5.3 meters per minute, while the tension remained between 280 and 300 Newtons. Based on the current +0.2 mm spacing deviation, the system increased the baseline wire feeding speed from 5.0 to 5.1 meters per minute; based on the -2 degree angle deviation, it increased the baseline tension from 290 Newtons to 300 Newtons; and for the 120-degree defect-dense area, it further reduced the wire feeding speed in that area by 5% to 4.85 meters per minute. After adjustment, continuous system monitoring showed that no defects reappeared in the area during the subsequent two rotation cycles, and the heat map showed that the mass distribution tended to be uniform. The speed adjustment was calculated using the formula Vnew = Vbaseline × (1 + Δd / dstandard), where Δd = +0.2 mm is the spacing deviation and dstandard = 5.0 mm is the standard spacing, resulting in a speed adjustment coefficient of 1.04. However, considering the need to reduce speed in the defect area, a compromise parameter was ultimately adopted. The entire control process achieved precise location of quality anomalies and dynamic optimization of process parameters.

[0120] In this embodiment, the method uses a quality heatmap to visually locate problem areas and combines historical data and real-time deviations for precise control, thereby achieving dynamic optimization of winding process parameters and effectively improving product quality stability.

[0121] To further improve image acquisition quality, in some embodiments, step 101: acquiring the preprocessed image containing the wire wrapped around the concrete pipe includes: Step 701: Acquire data from the annular multispectral light source, multi-angle optical images of the concrete pipe, and light sensor data.

[0122] In step 701, the annular multispectral light source data refers to the combination of brightness parameters of multiple color light sources arranged around the pipe, used to provide a stable lighting environment. The annular multispectral light source data includes visible light sources and infrared light sources. The multi-angle optical images are photographs of the pipe surface taken from different directions at a 30-degree tilt angle by a high-frame-rate industrial camera, which can comprehensively capture the wire winding characteristics. The light sensor data is the real-time monitored ambient light intensity value.

[0123] In this embodiment of the application, the system uses special light sources and multiple industrial cameras arranged around the pipeline to simultaneously acquire images of the pipeline from different angles, while recording the current ambient light intensity, providing complete data for subsequent image processing.

[0124] Step 702: Based on the light sensor data, dynamically adjust the light source intensity of the annular multispectral light source data and the exposure parameters of the multi-angle optical image to generate a stable illumination image.

[0125] In step 702, the light source intensity of the annular multispectral light source data refers to the brightness levels of the various color light sources arranged around the pipe. This value is calculated using ambient brightness data monitored in real time by a light sensor. The light source intensity is automatically reduced when the ambient light is strong and increased when the ambient light is weak. The exposure parameters of the multi-angle optical image refer to the exposure time setting when the industrial camera is shooting. This parameter is dynamically adjusted according to changes in light source intensity and ambient light. The exposure time is shortened when the light source intensity is high and lengthened when the light source intensity is low. These two parameters work together to ensure consistent brightness in images taken from different angles. A stable illumination image refers to a photograph of the pipe surface with uniform illumination after brightness adjustment, eliminating the effects of uneven brightness.

[0126] In this embodiment, the system automatically adjusts the brightness of the light source and the camera exposure time based on real-time data measured by the light sensor, so that images taken from different angles maintain a consistent brightness level and avoid overexposure or underexposure.

[0127] Step 703: Perform weighted grayscale equalization and adaptive threshold segmentation on the stable illumination image in sequence to obtain a preprocessed image that suppresses interference from concrete surface texture.

[0128] In this embodiment, the system first performs weighted grayscale processing on the image to balance the contrast between light and dark in different areas, then performs intelligent threshold segmentation to highlight the outline of the wire, and finally obtains a clear image suitable for feature extraction.

[0129] Here is a specific example: On a concrete pipe production line, when the system initiates the image acquisition process, it first acquires the current light source intensity parameters through a ring light source array installed around the pipe. Simultaneously, six industrial cameras capture images of the pipe surface from different angles, and a light sensor monitors the ambient brightness in real time, displaying 350 lux. Based on a preset light-exposure correlation, when the ambient light exceeds 300 lux, the system reduces the ring light source intensity to 80% of the standard value, and adjusts the camera exposure time to 1 / 800 second to ensure balanced image brightness across all angles. The acquired images are then preprocessed. First, grayscale conversion is performed using the formula Y=0.299R+0.587G+0.114B, where R, G, and B represent the pixel values ​​of the red, green, and blue channels, respectively, and Y represents the output grayscale value, standardizing image brightness under different light source conditions. Next, adaptive threshold segmentation is performed, automatically adjusting the segmentation threshold based on the grayscale distribution of different regions of the image, focusing on preserving the wire outline features. In the final preprocessed image, the outlines of the five strands of wire are clearly visible, and interference from the concrete surface texture is effectively suppressed, providing quality assurance for the subsequent skeleton line extraction step.

[0130] In this embodiment of the application, the method effectively overcomes the interference of complex lighting conditions in industrial sites through intelligent illumination control and image processing, and obtains high-quality winding feature images, laying a reliable foundation for subsequent accurate measurement.

[0131] Figure 4 This application provides a schematic diagram of a specific implementation of a machine vision-based real-time monitoring system for the quality of wire wrapping in concrete pipes, as illustrated in the embodiments of this application. Figure 4 The system may include: The acquisition module 41 is used to acquire a preprocessed image containing the wire wrapping of the concrete pipe.

[0132] Extraction module 42 is used to perform improved Hough transform on the preprocessed image to extract the skeleton contour point set of the winding wire, and to calculate the spacing and angle data of adjacent winding wires based on the skeleton contour point set using a triangulation algorithm.

[0133] The generation module 43 is used to generate winding wire spacing deviation data and angle deviation data based on the spacing data and the angle data, respectively, input the skeleton contour point set into the hybrid model constructed based on convolutional neural network and transformer, and output the defect location data of winding wire.

[0134] The input module 44 is used to input the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller. The wire winding machine controller adjusts the wire feeding speed parameters and the wire feeding tension parameters to optimize the quality of the wire winding of the concrete pipe based on the adjusted wire feeding speed parameters and the adjusted wire feeding tension parameters.

[0135] The machine vision-based real-time monitoring system for the quality of wire winding in concrete pipes in this application is used to implement the aforementioned machine vision-based real-time monitoring method for the quality of wire winding in concrete pipes. Therefore, the specific implementation of the machine vision-based real-time monitoring system for the quality of wire winding in concrete pipes can be found in the embodiment section of the machine vision-based real-time monitoring method for the quality of wire winding in concrete pipes above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0136] This application also provides an electronic device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described machine vision-based real-time monitoring methods for the quality of wire wrapping in concrete pipes.

[0137] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described machine vision-based real-time monitoring methods for the quality of wire wrapping in concrete pipes.

[0138] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0139] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the machine vision-based real-time monitoring method for the quality of wire winding in concrete pipes.

[0140] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0141] The above provides a detailed description of the machine vision-based real-time monitoring method and system for the wire wrapping quality of concrete pipes. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for real-time monitoring of the wire wrapping quality of concrete pipes based on machine vision, characterized in that, include: Obtain a preprocessed image containing the wire wrapping of the concrete pipe; An improved Hough transform is applied to the preprocessed image to extract the skeleton contour point set of the winding wire. Based on the skeleton contour point set, the spacing and angle data of adjacent winding wires are calculated using a triangulation algorithm. Based on the spacing data and the angle data, the winding wire spacing deviation data and angle deviation data are generated respectively. The skeleton contour point set is input into the hybrid model built based on convolutional neural network and transformer, and the defect location data of winding wire is output. The wire winding spacing deviation data, the angle deviation data, and the defect location data are input into the wire winding machine controller. The wire feeding speed parameters and wire feeding tension parameters are adjusted by the wire winding machine controller to optimize the quality of wire winding in concrete pipes based on the adjusted wire feeding speed parameters and adjusted wire feeding tension parameters. Based on the spacing data and the angle data, the winding spacing deviation data and angle deviation data are generated respectively. The skeleton contour point set is input into a hybrid model built based on a convolutional neural network and a transformer, and the defect location data of the winding wire is output, including: The difference between the spacing data and the preset standard spacing is used as the winding spacing deviation data; The difference between the angle data and the value of the preset standard angle is used as the angle deviation data; The skeleton contour point set is input into the hybrid model, and the convolutional neural network in the hybrid model sorts the discrete points in the skeleton contour point set that belong to the same winding wire according to the spatial adjacency relationship to form an ordered point chain. The spatial distribution of the ordered point chain is analyzed by the transformer in the hybrid model, and the defect location data of the entangled wire are output.

2. The method according to claim 1, characterized in that, The analysis of the spatial distribution of the ordered point chain outputs defect location data of the entangled wire, including: The ordered dot chain is divided into multiple dot chain segments according to a preset segment length; Spatial continuity detection is performed on each point chain segment to generate candidate regions for broken wires and candidate regions for stacked wires; Interpolate the endpoint coordinates of all candidate areas for wire breakage to obtain the start and end coordinates of the wire breakage defect; Cluster the boundary points of all candidate areas for wire stacking to obtain the coordinate set of the wire stacking defect area; The start and end coordinates of the broken wire defect are merged with the coordinate set of the overlapping wire defect area to generate defect location data.

3. The method according to claim 2, characterized in that, The step of performing spatial continuity detection on each point chain segment to generate candidate regions for broken and stacked fibers includes: For each point chain segment, calculate the straight-line distance between adjacent discrete points to form a set of straight-line distances for the point chain segment; The deviation between each straight-line distance value in the set of straight-line distances and the arithmetic mean of the straight-line distance values ​​is calculated. When the deviation exceeds a preset fracture threshold, the region corresponding to the point chain segment is marked as a candidate region for broken wires. For adjacent point chain segments, the end boundary point of the first point chain segment and the start boundary point of the second point chain segment are combined to form a set of point pairs; Calculate the ratio of the sum of distances between all point pairs in the point pair set to the number of point pairs to obtain the average point spacing value; When the average point spacing value is less than the preset wire stacking threshold, the region corresponding to the adjacent point chain segment is marked as a wire stacking candidate region.

4. The method according to claim 1, characterized in that, The step of performing an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the wound wires, and then using a triangulation algorithm to calculate the spacing and angle data of adjacent wound wires based on the skeleton contour point set, includes: An improved Hough transform is applied to the preprocessed image to detect the edge pixels of all the wires wrapped around the concrete pipe surface in the transformed image. Adjacent edge pixels belonging to the same winding wire are connected by linear interpolation to generate multiple skeleton lines, each skeleton line corresponding to one winding wire; Discrete points are sampled from each skeleton line according to a preset fixed step size, and the discrete points of all skeleton lines are merged to form a skeleton contour point set; In the set of skeleton contour points, a triangular mesh is constructed for the corresponding discrete points of adjacent winding wires; Calculate the arithmetic mean of the side lengths of all triangles in the triangular mesh, and use the arithmetic mean as the spacing data between adjacent windings; Calculate the angle values ​​of the interior angles of the triangles in the triangular mesh, and use these angle values ​​as the angle data for winding the wire.

5. The method according to claim 1, characterized in that, The step of inputting the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller, and adjusting the wire feeding speed parameters and wire feeding tension parameters through the wire winding machine controller, includes: The winding spacing deviation data, the angle deviation data, and the defect location data are input into the winding machine controller. The real-time position information of the winding wire is spatiotemporally aligned with the defect location data through the alignment module of the winding machine controller to generate a winding quality heat map. The coordinates of the defect-dense area are determined by the coordinate recognition module of the winding machine controller based on the winding quality heat map; When the coordinates of the defect-dense area repeat within N consecutive rotation cycles, the historical data of production process parameters within N rotation cycles are retrieved through the retrieval module of the winding machine controller. The wire feeding speed and wire feeding tension parameters are adjusted by the adjustment module of the wire winding machine controller based on the wire winding spacing deviation data, the angle deviation data, and the historical data of the production process parameters.

6. The method according to claim 1, characterized in that, The process of acquiring a preprocessed image containing the wire wrapping of a concrete pipe includes: Acquire data from a ring-shaped multispectral light source, multi-angle optical images of concrete pipes, and data from a light sensor. Based on the light sensor data, the light source intensity of the ring multispectral light source data and the exposure parameters of the multi-angle optical image are dynamically adjusted to generate a stable light image; The stable illumination image is subjected to weighted grayscale equalization and adaptive threshold segmentation in sequence to obtain a preprocessed image that suppresses the interference of concrete surface texture.

7. A machine vision-based real-time monitoring system for the quality of wire wrapping in concrete pipes, characterized in that, include: The acquisition module is used to acquire preprocessed images containing the wire wrapping of concrete pipes; The extraction module is used to perform an improved Hough transform on the preprocessed image to extract the skeleton contour point set of the winding wire, and to calculate the spacing and angle data of adjacent winding wires based on the skeleton contour point set using a triangulation algorithm. The generation module is used to generate winding wire spacing deviation data and angle deviation data based on the spacing data and the angle data, respectively, input the skeleton contour point set into the hybrid model built based on convolutional neural network and transformer, and output the defect location data of winding wire. The input module is used to input the wire winding spacing deviation data, the angle deviation data, and the defect location data to the wire winding machine controller. The wire winding machine controller adjusts the wire feeding speed parameters and the wire feeding tension parameters to optimize the quality of the wire winding of the concrete pipe based on the adjusted wire feeding speed parameters and the adjusted wire feeding tension parameters. Based on the spacing data and the angle data, the winding spacing deviation data and angle deviation data are generated respectively. The skeleton contour point set is input into a hybrid model built based on a convolutional neural network and a transformer, and the defect location data of the winding wire is output, including: The difference between the spacing data and the preset standard spacing is used as the winding spacing deviation data; The difference between the angle data and the value of the preset standard angle is used as the angle deviation data; The skeleton contour point set is input into the hybrid model, and the convolutional neural network in the hybrid model sorts the discrete points in the skeleton contour point set that belong to the same winding wire according to the spatial adjacency relationship to form an ordered point chain. The spatial distribution of the ordered point chain is analyzed by the transformer in the hybrid model, and the defect location data of the entangled wire are output.

8. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the steps of the machine vision-based real-time monitoring method for the wire winding quality of concrete pipes as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the machine vision-based real-time monitoring method for the wire wrapping quality of concrete pipes as described in any one of claims 1 to 6.

Images