Automobile seat defect detection method based on image recognition

Abstract

The invention relates to the technical field of automobile seat surface defect detection, in particular to an automobile seat defect detection method based on image recognition. The method comprises the following steps: S1, image acquisition and modeling; s2, texture continuous detection; s3, the swelling degree is judged, and the swelling abnormal reason is judged; s4, performing suture correlation analysis; s5, comprehensively judging the temperature; according to the invention, continuous detection is carried out on the texture by constructing the point cloud model, and material filling judgment or suture correlation analysis is carried out in a classified manner during swelling; and temperature comprehensive judgment is carried out in the suture extension area so as to identify specific abnormal reasons. According to the invention, based on three-dimensional point cloud and texture continuity analysis, automatic identification and classification judgment of defects such as surface damage and swelling of the seat are realized; fusing suture process characteristics and temperature data to effectively distinguish material abnormity and structure abnormity; the method is suitable for high-precision detection of a multifunctional area, and manual intervention and misjudgment are effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of seat detection data processing, and in particular to an automobile seat defect detection method based on image recognition. Background Art

[0002] With the continuous development of the automobile manufacturing industry, automobile seat defect detection technology has also evolved from traditional manual methods to modern intelligent systems.

[0003] In the early days, seat defect detection relied primarily on visual inspection and manual measurement, testing items including appearance quality, dimensions, compression resilience, wear resistance, aging resistance, and flame retardancy. However, these methods suffered from low efficiency, high subjectivity, and difficulty in standardization, making them unable to meet the high efficiency and high precision requirements of modern automotive manufacturing.

[0004] To address these issues, machine vision technology has been introduced to seat defect detection. Using industrial cameras and image processing software, it can automatically identify surface defects such as damage, stains, and color variations, improving the automation and standardization of detection. With the development of artificial intelligence (AI), deep learning and machine learning algorithms are being applied to seat defect detection. By training a multi-feature fusion classifier, it can identify seat materials of varying textures and colors, and detect defects such as damage and stains, improving detection accuracy and adaptability.

[0005] In addition, the application of 3D vision technology and point cloud processing algorithms has expanded seat defect detection from two-dimensional images to three-dimensional space, and can detect structural defects of seats such as bulging and collapse, thereby improving the comprehensiveness and accuracy of detection.

[0006] In summary, automobile seat defect detection technology is developing towards automation, intelligence and high precision, gradually meeting the higher requirements of modern automobile manufacturing for product quality and production efficiency.

[0007] Chinese Patent Publication No. CN118864448A discloses a method, device, and medium for detecting automotive seat defects based on image recognition. Existing automotive seat defect detection methods, therefore, lack the ability to establish a correlation between defect images and specific defect types and causes. The large amount of computation required to process complex and atypical defects leads to slow recognition speeds, and the method lacks an adaptive adjustment mechanism and the ability to adapt to diverse defect scenarios. Summary of the Invention

[0008] To this end, the present invention provides a car seat defect detection method based on image recognition to overcome the problems in the prior art of weak defect association establishment and lack of adaptive adjustment mechanism, which results in single feedback information and increased risk of misjudgment and missed judgment.

[0009] To achieve the above object, the present invention provides a method for detecting defects in automobile seats based on image recognition, comprising: Delineating each functional area of ​​the automobile seat according to the functional area classification standard, and obtaining the factory requirement standards of each factory inspection parameter corresponding to each functional area; Power up the car seat; Continuously collecting visual images of the powered car seat and various identification information in the visual images within a standard collection period, and matching the various identification information with corresponding points to construct a point cloud structure model of the powered car seat; Continuously detect the texture of the car seat surface to obtain continuous detection results and corresponding continuous areas; If the second continuous detection result is obtained, the bulging degree is judged, and if the bulging degree is abnormal, the corresponding bulging abnormality cause is judged; The abnormal result of the bulging degree includes the first bulging degree determination result and the third bulging degree determination result; When the first result of the bulging degree judgment is obtained, a material filling comprehensive judgment is performed, or when the third result of the bulging degree judgment is obtained, a seam correlation analysis is performed; If the seam correlation analysis is performed to obtain the seam extension area result, the temperature data of the powered car seat is analyzed based on the point cloud structure model, a comprehensive temperature judgment is performed and the category to which the comprehensive temperature judgment result belongs is determined, and when the comprehensive temperature judgment result is an equipment imbalance type inflation abnormality category, an identification self-inspection step is performed to determine the cause of the inflation abnormality.

[0010] Furthermore, the constructing of the point cloud structure model of the powered car seat includes: Continuously acquire visual images of powered car seats within a standard acquisition cycle; Wherein, the visual image includes a color image and a depth image, and the color image and the depth image are converted into a three-dimensional point cloud; Acquire each identification information in the visual image and convert it into each identification field, and superimpose the identification field of each point and store it in each corresponding point cloud structure; The point cloud is divided into functional areas based on the functional area division standard, and the functional areas corresponding to each point cloud structure are superimposed and stored in the corresponding point cloud structure in the form of marks.

[0011] Furthermore, the continuous detection of the texture of the car seat surface includes: Acquire a reflection image of a surface of a car seat illuminated by a movable light source, and divide the reflection image into a plurality of reflection image blocks based on a reflection image block division standard; Traversing each reflected image block, obtaining the actual texture angle in the adjacent reflected image block having a common edge relationship with the reflected image block; Obtaining an actual texture angle change difference between the reflection image block and an adjacent reflection image block and comparing it with a preset texture angle change threshold to obtain continuous detection results; If the first continuous detection result is obtained, there is a damage defect in the reflected image block, triggering the next damage reminder; If the second continuous detection result is obtained, there is no damage defect in the reflected image block, and the bulge degree determination is performed.

[0012] Furthermore, the execution of the bulge degree determination includes: Acquire the height information of the sampling points in each functional area based on the point cloud data, and compare the height information of the sampling points with the standard height interval of the functional area where the sampling points are located to obtain a height comparison result; Obtaining the swelling degree judgment result corresponding to each height comparison result, and the category to which the swelling degree judgment result belongs; If the first result of the swelling degree judgment is obtained, the swelling area is superimposed and stored in the corresponding point cloud structure in the form of a mark, and a comprehensive judgment of the material filling is performed; If the third result of the bulging degree determination is obtained, the suture correlation analysis is performed.

[0013] Furthermore, the performing of comprehensive material filling judgment includes: The comprehensive judgment of material filling includes material detection steps and filling adjustment steps; The material detection step is used to determine whether the material of the functional area corresponding to the sampling point is in a performance fatigue state; If the material test result is a material performance fatigue result, a reminder to change the material is triggered; If the material test result shows that the material performance is stable, perform the filling adjustment step; Obtain the filling material quality within the factory inspection parameters corresponding to the functional area, and trigger a reminder to adjust the filling material quality.

[0014] Furthermore, the suture correlation analysis includes: Obtain the swelling area surrounded by the sampling points of the swelling mark, and extract the acquisition point at the geometric center of the swelling area and the coordinates of each sampling point on the outline of the swelling area; Obtain the suture path segment of the functional area where the collection point at the geometric center of the bulging area is located and the corresponding suture path segment constraint area; Starting from the collection point at the geometric center, draw a perpendicular line perpendicular to the suture path segment with the shortest distance to the segment constraint area; The intersection of the bulging area outline and the vertical line is obtained as a collection point, and the determined collection point is superimposed and stored in the point cloud structure corresponding to the collection point in the form of a mark; Determining whether the determination acquisition point is within the suture path segment constraint area; If the collection point is determined to be within the suture path segment constraint area, the bulging area is the suture influence area, and the suture process and support judgment is performed; If it is determined that the collection point is not within the suture path segment constraint area, the bulging area is the suture extension area, and a comprehensive temperature judgment is performed.

[0015] Furthermore, the suture process and support judgment includes suture process identification, wherein: Obtain the suture depth and suture shape of the suture path segment detected at the factory; Obtaining the overlapping portion of the suture influence area and the suture path segment constraint area, and superimposing and storing the overlapping area in the form of a mark into the point cloud structure corresponding to each acquisition point of the overlapping portion; Obtaining the curvature of each acquisition point in the overlapping area, traversing the curvature of each acquisition point, and obtaining the curvature of the acquisition point that is adjacent to the acquisition point; Obtaining an actual curvature change difference between the acquisition point and an adjacent acquisition point and comparing it with a preset curvature change threshold, and obtaining a corresponding suture process recognition result based on the curvature comparison result; If the first curvature comparison result is obtained, corresponding to the stretch suture process identification result, the suture support judgment is performed; If the second curvature comparison result is obtained, it corresponds to the stacking stitching process identification result, triggering the material stacking reminder; Furthermore, the suture process and support judgment also includes suture support judgment, wherein, Obtaining the endpoints of the suture path segments, and extracting the three-dimensional coordinates of the sampling points at the endpoints in the point cloud structure and their corresponding actual depth values; Extracting the standard suture depth value corresponding to the functional area where the endpoint is located, calculating the actual suture height difference value and comparing the actual suture height difference value with the standard suture height difference threshold value; Among them, the actual suture height value is the absolute value of the difference between the standard suture depth value and the actual depth value; If the first seam height difference comparison result is obtained, a lockstitch connection failure reminder is triggered; If the second suture height difference comparison result is obtained, a reminder to adjust the filling is triggered; Furthermore, the comprehensive temperature judgment includes eliminating interference from the real-time working environment temperature, wherein: Determine whether the real-time working environment temperature and the preset standard working environment temperature are both within the linear change range of the temperature-expansion curve; If both ambient temperatures are within the linear range of the temperature-expansion curve, a linear compensation calculation method is used to eliminate the expansion effect caused by the real-time working ambient temperature, and after elimination, an expansion temperature correlation judgment is performed; If any ambient temperature is not within the linear change range of the temperature-expansion curve, an ambient temperature interference reminder is triggered, and the expansion temperature association judgment is performed after the real-time working ambient temperature returns to the standard working ambient temperature.

[0016] Furthermore, the comprehensive temperature judgment also includes a bulge temperature correlation judgment, wherein: Acquire each collection point in the area directly affected by the heating wire corresponding to each functional area, and store the area directly affected by the heating wire in the form of a mark in the point cloud structure corresponding to each collection point; Traverse each collection point in the area directly affected by the heating wire to determine whether there is a mark for the bulging area; If there is a mark of a swelling area, obtain the maximum height information of each collection point with the swelling area mark and compare it with the standard heating height threshold; If the maximum height information is greater than or equal to the standard heating height threshold, an abnormal heating wire position reminder is triggered; If the maximum altitude information is less than the standard heated altitude threshold, there is no abnormality with the powered seat; If there is no mark for the bulging area, a reminder to adjust the filling quality is triggered.

[0017] Compared with existing technologies, the present invention offers the following advantages: it precisely defines the functional areas of automotive seats through functional area classification standards, enables continuous acquisition of visual images of powered automotive seats and construction of a point cloud structural model, and supports dynamic detection and defect identification of seat surface textures. Furthermore, it addresses bulging anomalies by subdividing them into two categories: collapse and bulge. Material filling comprehensive judgment and seam correlation analysis are used, respectively. Temperature data and comprehensive temperature judgment are combined to distinguish between equipment imbalance and heat source interference-induced bulging anomalies. Furthermore, a self-check step is introduced to ensure accurate diagnosis. This overall method is highly systematic and comprehensive, capable of precisely pinpointing the cause of defects, improving detection accuracy and intelligence, and effectively ensuring the quality and safety of automotive seats.

[0018] Furthermore, by continuously collecting color images and depth images and converting them into three-dimensional point clouds, the three-dimensional data of the car seat can be reconstructed; at the same time, the identification fields and functional area marks are superimposed to form a point cloud model with clear structure and rich information, which provides an accurate spatial basis and functional area division support for subsequent defect detection and analysis, and improves the accuracy and efficiency of detection.

[0019] Furthermore, the reflected image of a movable light source is divided into several reflected image blocks, and damage defects are judged based on the difference in texture angle changes, thereby achieving highly sensitive detection of continuous textures on the seat surface; damage defects and normal textures are strictly distinguished through thresholds, effectively ensuring the accuracy of the detection results and improving the ability to identify tiny surface defects.

[0020] Furthermore, the height information of the sampling points is obtained based on the point cloud data and compared with the standard height range to achieve quantitative judgment of the swelling degree; different processing strategies are adopted for the swelling degrees of collapse and bulge, respectively, combined with the marking and storage of functional area defects, to improve the accuracy and refinement of defect positioning, which is convenient for subsequent targeted analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Schematic diagram of the process of a method for detecting defects in automobile seats based on image recognition according to an embodiment of the present invention; Figure 2 This is a logic decision diagram for continuously detecting the texture of the surface of a car seat according to an embodiment of the present invention; Figure 3 This is a logic decision diagram for suture correlation analysis according to an embodiment of the present invention; Figure 4 This is a logic decision diagram for eliminating real-time working environment temperature interference according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0023] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0025] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0026] See also Figure 1 As shown, it is a method for detecting defects in automobile seats based on image recognition according to an embodiment of the present invention, which is characterized by including: Delineating each functional area of ​​the automobile seat according to the functional area classification standard, and obtaining the factory requirement standards of each factory inspection parameter corresponding to each functional area; Power up the car seat; Continuously collecting visual images of the powered car seat and various identification information in the visual images within a standard collection period, and matching the various identification information with corresponding points to construct a point cloud structure model of the powered car seat; Continuously detect the texture of the car seat surface to obtain continuous detection results and corresponding continuous areas; If the second continuous detection result is obtained, the bulging degree is judged, and if the bulging degree is abnormal, the corresponding bulging abnormality cause is judged; The abnormal result of the bulging degree includes the first bulging degree determination result and the third bulging degree determination result; When the first result of the bulging degree judgment is obtained, a material filling comprehensive judgment is performed, or when the third result of the bulging degree judgment is obtained, a seam correlation analysis is performed; If the seam correlation analysis is performed to obtain the seam extension area result, the temperature data of the powered car seat is analyzed based on the point cloud structure model, a comprehensive temperature judgment is performed and the category to which the comprehensive temperature judgment result belongs is determined, and when the comprehensive temperature judgment result is an equipment imbalance type inflation abnormality category, an identification self-inspection step is performed to determine the cause of the inflation abnormality.

[0027] The system precisely defines the functional areas of automotive seats using functional area classification standards, enabling continuous capture of visual images of powered automotive seats and construction of point cloud structural models. This supports dynamic detection and defect identification of seat surface textures. For abnormal bulging, the system further differentiates between collapse and bulging, using a comprehensive material filling assessment and seam correlation analysis. Temperature data and comprehensive temperature assessment differentiate between equipment imbalance and heat source interference-induced bulging, and further introduces self-test procedures to ensure accurate diagnosis. This highly systematic and comprehensive approach precisely locates the cause of defects, improves detection accuracy and intelligence, and effectively ensures the quality and safety of automotive seats.

[0028] Specifically, constructing the point cloud structure model of the powered car seat includes: Capturing visual images of powered car seats; Wherein, the visual image includes a color image and a height image, and the color image and the height image are converted into a three-dimensional point cloud; Obtaining color and height identification information corresponding to each point in the visual image, converting them into corresponding identification values, and superimposing and storing each identification value in the form of a mark into a corresponding point cloud structure; Wherein, the height identification information is the actual height value of each collection point; The point cloud structure is divided into functional areas based on the functional area division standard, and the functional areas corresponding to each point cloud structure are superimposed and stored in the corresponding point cloud structure in the form of marks.

[0029] In this embodiment, under natural indoor lighting conditions, with a 60-second overall inspection cycle, an industrial camera mounted above the inspection station acquires color images and depth images obtained by structured light or binocular computation. The images are continuously captured at a frame rate of 15 frames per second, with an image resolution of 1920×1080. The image acquisition time is uniformly set to 5 seconds after the seat is powered on. The color image and depth image are jointly processed using the OpenCV and PCL libraries. Each pixel in the depth image is converted into three-dimensional space coordinates (X, Y, Z) through perspective projection and bound to the RGB value of the corresponding pixel in the color image to form a preliminary point cloud. Using an XYZRGB field structure, the number of point clouds in a single frame is approximately 400,000. Use the RANSAC algorithm and the corresponding factory requirements to fit the reference plane of each functional area from the point cloud; The reference plane is the surface of the unfilled seat structure when the bare chair is covered with leather and has no filling material; For each sampling point, calculate its normal distance to the fitting plane as the actual height value of the point; At the same time, the color value of the point is extracted and converted into HSV space value as the material identification parameter; Store the above two values ​​in the point cloud structure; In this embodiment, the seat is divided into functional areas including a seat cushion area, a lumbar support area, a neck support area, a side support area and a leg support area.

[0030] By continuously collecting color images and depth images and converting them into three-dimensional point clouds, the three-dimensional data of the car seat can be reconstructed. At the same time, identification fields and functional area markers are superimposed to form a point cloud model with a clear structure and rich information. This provides an accurate spatial basis and functional area division support for subsequent defect detection and analysis, improving the accuracy and efficiency of detection.

[0031] See Figure 2 As shown, it is a logic decision diagram for continuously detecting the texture of the surface of a car seat according to an embodiment of the present invention; Specifically, continuous detection of the texture of the car seat surface includes: Acquire a reflection image of a surface of a car seat illuminated by a movable light source, and divide the reflection image into a plurality of reflection image blocks based on a reflection image block division standard; Traversing each reflected image block, obtaining the actual texture angle in the adjacent reflected image block having a common edge relationship with the reflected image block; Obtaining an actual texture angle change difference between the reflection image block and an adjacent reflection image block and comparing it with a standard texture angle change threshold to obtain continuous detection results; If the first continuous detection result is obtained, there is a damage defect in the reflected image block, triggering a damage reminder; If the second continuous detection result is obtained, there is no damage defect in the reflected image block, and the bulge degree determination is performed.

[0032] In this embodiment, after the point cloud model is constructed, the surface of the powered car seat is imaged under the illumination of the configured movable light source, using a high-resolution industrial camera or a structured light scanner to obtain a surface reflection image. The movable light source illumination time is 5 seconds. The reflected image is a grayscale image; Based on the reflection image block division standard, the reflection image is divided into several reflection image blocks with each block being 32×32 pixels; Apply image processing algorithm to each reflected image block to obtain the actual texture angle θ of the area; The actual texture angle unit is degree, range [0,180]; Traverse each reflected image block, compare it with the adjacent image blocks that have a common edge relationship with it, calculate the actual texture angle change difference, and compare it with the standard texture angle change threshold; If the actual texture angle change difference is greater than the standard texture angle change threshold, the texture has undergone a sudden change, and there may be a crack or damage, and a first continuous detection result is obtained; If the actual texture angle change difference is less than or equal to the standard texture angle change threshold, the texture change is continuous, and a second continuous detection result is obtained; Among them, the standard texture angle change threshold is determined to be 25 degrees based on historical factory recognition data; The reflected image of a movable light source is divided into several reflected image blocks, and damage defects are judged based on the difference in texture angle changes, achieving highly sensitive detection of continuous texture on the seat surface. Damage defects and normal textures are strictly distinguished through thresholds, effectively ensuring the accuracy of the test results and improving the ability to identify tiny surface defects.

[0033] Specifically, the execution of the bulge degree judgment includes: Acquire the actual height value corresponding to the actual height information of any sampling point in each functional area based on the point cloud data, and compare the actual height value with the standard actual height range of the functional area where the sampling point is located to obtain a height comparison result; Obtaining the swelling degree judgment result corresponding to each height comparison result, and the category to which the swelling degree judgment result belongs; If the first result of the swelling degree judgment is obtained, the swelling area is superimposed and stored in the corresponding point cloud structure in the form of a mark, and a comprehensive judgment of the material filling is performed; If the third result of the bulging degree determination is obtained, the suture correlation analysis is performed.

[0034] In this embodiment, under the condition that artificial leather material is used, the standard actual height ranges of each area are: The standard actual height range for the seat cushion area is [0, 2], the standard actual height range for the lumbar support area is [1.5, 3], the standard actual height range for the neck support area is [0.5, 2.5], the standard actual height range for the side support area is [1, 1.5], and the standard actual height range for the leg support area is [0, 3]. The unit of each range is mm. The unit of each interval is millimeter, and the interval floating value is the floating value of the sampling point on the normal vector of the reference plane corresponding to the corresponding functional area, and the direction outside the seat is defined as positive; Compare the actual height value with the standard actual height range of the functional area where the sampling point is located; If the actual height value is less than the minimum value of the standard actual height range, the first result of the bulging degree judgment is obtained, and the actual bulging degree is less than the normal range, and the material filling comprehensive judgment is performed; If the actual height value is within the standard actual height range, the second result of the bulge degree judgment is obtained, and the actual bulge degree is within the normal range; If the actual height value is greater than the maximum value of the standard actual height interval, the third result of the bulging degree judgment is obtained, and the actual bulging degree is greater than the normal interval, and the suture correlation analysis is performed.

[0035] Based on point cloud data, the height information of the sampling points is obtained and compared with the standard height range to achieve quantitative judgment of the swelling degree. Different processing strategies are adopted for the swelling degrees of collapse and bulge, respectively. Combined with the marking and storage of functional area defects, the accuracy and refinement of defect positioning are improved, which facilitates subsequent targeted analysis.

[0036] Specifically, performing comprehensive judgment of material filling includes: The comprehensive judgment of material filling includes material detection steps and filling adjustment steps; The material detection step is used to determine whether the material of the functional area corresponding to the sampling point is in a performance fatigue state; If the material detection result is a material performance fatigue result, a reminder to adjust the material is triggered; If the material test result shows that the material performance is stable, perform the filling adjustment step; The filling material quality within the factory inspection parameters corresponding to the functional area is obtained through the filling adjustment step, triggering a reminder to adjust the filling material quality.

[0037] In this embodiment, under the illumination of a configured movable light source, surface imaging of the powered car seat is performed using a high-resolution industrial camera or a structured light scanner to obtain a surface reflection image, and the gloss reflection intensity of each acquisition point in the surface reflection image is obtained; the movable light source illumination time is 5 seconds; Obtain the gloss reflection intensity of all collection points in each functional area, calculate the average gloss reflection value of each functional area, and compare it with the standard reflection intensity; The calculation formula for the average value of glossy reflection is:

[0038] in, is the average value of glossy reflection; N is the number of sampling points in the functional area; is the reflection intensity of a single sampling point; Define the gloss falloff rate:

[0039] in, is the standard reflection intensity; Obtain the standard reflection intensity under the factory requirement standard. In this embodiment, the standard reflection intensity under artificial leather fabric is 0.25; Compare the average glossy reflection with the standard reflection intensity; If the average glossy reflection intensity is greater than the standard reflection intensity, the reflection decreases significantly, triggering a material fatigue reminder; If the average value of the gloss reflection is less than or equal to the standard reflection intensity, the gloss is judged to be normal, that is, the filler quality under the factory requirement standard is small, triggering a filler quality adjustment reminder.

[0040] Through a two-step comprehensive assessment of material testing and filling adjustment, material fatigue is promptly detected and a replacement reminder is triggered, ensuring stable seat material performance. The filling adjustment step guides the adjustment of filling material quality based on factory test parameters, improving the scientific and rational filling effect, extending the seat lifespan, and enhancing ride comfort.

[0041] See Figure 3 As shown, it is a logic decision diagram of the suture correlation analysis according to an embodiment of the present invention; Specifically, the suture correlation analysis includes: Obtaining a swelling area formed by sampling points with swelling marks, and extracting the acquisition point at the geometric center of the swelling area and the coordinates of each sampling point on the outline of the swelling area; Obtain the suture path segment of the functional area where the collection point at the geometric center of the bulging area is located and the corresponding suture path segment constraint area; Starting from the acquisition point at the geometric center, draw several candidate perpendicular lines perpendicular to the suture path segment, and obtain the target perpendicular line with the shortest distance to the segment constraint area; Obtaining the intersection of the bulging area outline and the target vertical line as a judgment collection point, and superimposing and storing the judgment collection point in the form of a mark into the point cloud structure corresponding to the collection point; Determining whether the determination acquisition point is within the suture path segment constraint area; If the collection point is determined to be within the suture path segment constraint area, the bulging area is the suture influence area, and the suture process and support judgment is performed; If it is determined that the collection point is not within the suture path segment constraint area, the bulging area is the suture extension area, and a comprehensive temperature judgment is performed.

[0042] In this embodiment, the suture path segment constraint area is the area where the suture is deformed due to the actual suture trajectory and tension, restraint, and sewing traction on both sides thereof; The suture extension area is the non-suture path segment constraint area.

[0043] A method combining geometric center and contour sampling points is used to determine the relationship between the bulging area and the suture path, distinguish the suture-affected area from the extension area, and perform suture process judgment or temperature comprehensive judgment on different areas to accurately locate the cause of abnormal bulging, thereby improving the pertinence and scientific nature of defect analysis.

[0044] Specifically, the suture process and support judgment includes suture process identification, wherein: Obtain the suture height and suture shape of the suture path segment detected at the factory; Obtaining the overlapping portion of the suture influence area and the suture path segment constraint area, and superimposing and storing the overlapping area in the form of a mark into the point cloud structure corresponding to each acquisition point of the overlapping portion; Obtaining the curvature of each collection point in the overlapping area, traversing the curvature of each collection point, and obtaining the curvature of the collection points that are adjacent to the collection point; Obtaining an actual curvature change difference between the acquisition point and an adjacent acquisition point and comparing it with a preset curvature change threshold, and obtaining a corresponding suture process recognition result based on the curvature comparison result; If the first curvature comparison result is obtained, corresponding to the stretch suture process identification result, the suture support judgment is performed; If the second curvature comparison result is obtained, corresponding to the stacking stitching process identification result, a material stacking reminder is triggered.

[0045] In this embodiment, each sampling point in the overlapping area is obtained Curvature , and get its adjacent points and Curvature and ; The formula for calculating the actual curvature change difference is:

[0046] And compare it with the preset curvature change threshold; The preset curvature change thresholds for each functional area are: 0.15 for the seat cushion area, 0.2 for the lumbar support area, 0.2 for the neck support area, 0.1 for the side support area, and 0.15 for the leg support area. If the actual curvature change difference is less than or equal to the preset curvature change threshold, a first curvature comparison result is obtained, corresponding to the stretch suture process identification result, and suture support judgment is performed; If the actual curvature change difference is greater than the preset curvature change threshold, a second curvature comparison result is obtained, which corresponds to the stacking stitching process identification result and triggers a material stacking reminder.

[0047] Based on factory inspection standards for stitch depth and shape, combined with curvature variation comparison, stitching process types are identified, distinguishing between stretch stitching and stacked stitching. Stacked stitching triggers material stacking reminders, scientifically identifying stitching defects, supporting efficient and accurate process quality control, and ensuring the stability of the seat manufacturing process.

[0048] Specifically, the suture process and support judgment also includes suture support judgment, wherein, Obtaining the endpoints of the suture path segments, and extracting the three-dimensional coordinates of the sampling points at the endpoints in the point cloud structure and their corresponding actual height values; Extracting the standard suture height value corresponding to the factory requirement standard of the functional area where the endpoint is located, calculating the actual suture height difference value and comparing the actual suture height difference value with the standard suture height difference threshold value; Among them, the actual suture height value is the absolute value of the difference between the standard suture height value and the actual height value; If the first seam height difference comparison result is obtained, a lockstitch connection failure reminder is triggered; If the second suture height difference comparison result is obtained, a reminder to adjust the filling material quality is triggered.

[0049] In this embodiment, the endpoints of the suture path segments are visually identified; The endpoints include the starting point and the end point of the suture path segment; Calculate the actual seam height difference ; in, is the standard suture height value; is the actual height value; In this embodiment, the standard suture height value under the factory requirement standard is 1.8 mm; the standard suture height difference threshold is 0.3 mm; taking an actual suture height difference value and comparing the actual suture height difference value with a standard suture height difference threshold value; If the actual stitch height difference is greater than the standard stitch height difference threshold, a first stitch height difference comparison result is obtained, indicating a lockstitch connection failure. If the actual suture height difference is less than or equal to the standard suture height difference threshold, a second suture height difference comparison result is obtained, indicating that the filling material has a small mass; Among them, the lockstitch connection failure is that the upper thread and the lower thread of the sewing thread do not form a buckling structure, that is, they are not cross-entangled in the middle of the material, resulting in a loose or invalid structure at the seam.

[0050] By comparing the three-dimensional coordinates of the stitch path endpoints and the actual depth values ​​with the standard stitch depth values, the stitch height differences are quantified, and intelligent reminders of lockstitch connection defects and filler adjustments are implemented. This method refines the judgment of stitch support quality, helps to promptly discover key structural defects in the manufacturing process, and improves the overall durability of the seat.

[0051] See Figure 4 As shown, it is a logic decision diagram for eliminating real-time working environment temperature interference according to an embodiment of the present invention; Specifically, the comprehensive temperature judgment includes eliminating the interference of the real-time working environment temperature, wherein: Determine whether the real-time working environment temperature and the preset standard working environment temperature are both within the linear change range of the temperature-expansion curve; If both ambient temperatures are within the linear range of the temperature-expansion curve, a linear compensation calculation method is used to eliminate the expansion effect caused by the real-time working ambient temperature, and after elimination, an expansion temperature correlation judgment is performed; If any ambient temperature is not within the linear change range of the temperature-expansion curve, an ambient temperature interference reminder is triggered, and the expansion temperature association judgment is performed after the real-time working ambient temperature returns to the standard working ambient temperature.

[0052] In this embodiment, the temperature-expansion curve refers to a trend diagram of the change in the thermal expansion of the structure of a material under different temperature conditions. The curve usually has a two-segment structure, namely a linear change interval and a nonlinear change interval; The linear compensation calculation method is to determine whether the real-time working environment temperature is greater than the preset standard working environment temperature, and if so, subtract the inflation value caused by the difference between the real-time working environment temperature and the preset standard working environment temperature.

[0053] Eliminate the interference of the real-time working environment temperature and ensure the accuracy of the expansion judgment through temperature difference and expansion trend analysis; this step avoids the misleading of the test results by ambient temperature fluctuations and improves the robustness and practicality of the detection method.

[0054] Specifically, the comprehensive temperature judgment also includes a bulging temperature correlation judgment, wherein: Acquire each collection point in the area directly affected by the heating wire corresponding to each functional area, and store the area directly affected by the heating wire in the form of a mark in the point cloud structure corresponding to each collection point; Traverse each collection point in the area directly affected by the heating wire to determine whether there is a mark for the bulging area; If there is a mark of a swelling area, obtain the maximum height information of each collection point with the swelling area mark and compare it with the standard heating height threshold; If the maximum height information is greater than or equal to the standard heating height threshold, a reminder to adjust the heating wire position is triggered; If the maximum altitude information is less than the standard heated altitude threshold, there is no abnormality with the powered seat; If there is no mark for the bulging area, a reminder to adjust the filling quality is triggered.

[0055] In this embodiment, the center line of the heating wire in the area is obtained, and the area directly affected by the heating wire is constructed with a radius of 1.5 cm; Obtain the maximum height information of each collection point in the swelling area mark and compare it with the standard heating height threshold; The standard heating height thresholds for each functional area are: 4mm for the seat cushion area, 6mm for the lumbar support area, 5mm for the neck support area, 3mm for the side support area, and 6mm for the leg support area. If the maximum height information is greater than the standard heating height threshold, the local heat accumulation of the heating wire is too high, triggering a reminder to adjust the heating wire position; If the maximum height information is less than or equal to the standard heating height threshold and the heating wire has no effect, the system will trigger a reminder to adjust the filling quality to check the filling quality.

[0056] Combining point cloud markers of the area directly affected by the heating wire with markers of the swelling area enables quantitative judgment of swelling temperature correlation. Abnormal height triggers a position anomaly alert, while normal seat function is determined when no abnormality occurs. This improves the accuracy and real-time response capabilities of temperature-related defect detection, ensuring safe and stable operation of the seat.

[0057] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0058] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for detecting defects in automobile seats based on image recognition, characterized in that: include, Delineating each functional area of ​​the automobile seat according to the functional area classification standard, and obtaining the factory requirement standards of each factory inspection parameter corresponding to each functional area; Power up the car seat; continuously collecting visual images of the powered vehicle seat and various identification information within the visual images during a current detection cycle, and matching the various identification information with corresponding points to construct a point cloud structure model of the powered vehicle seat; Continuously detect the texture of the car seat surface to obtain continuous detection results and corresponding continuous areas; If the second continuous detection result is obtained, the bulging degree is judged, and if the bulging degree is abnormal, the corresponding bulging abnormality cause is judged; The abnormal result of the bulging degree includes the first bulging degree determination result and the third bulging degree determination result; When the first result of the bulging degree determination is obtained, a material filling comprehensive determination is performed, or when the third result of the bulging degree determination is obtained, a seam correlation analysis is performed; If the seam correlation analysis is performed to obtain the seam extension area result, the temperature data of the powered car seat is analyzed based on the point cloud structure model, a comprehensive temperature judgment is performed and the category to which the comprehensive temperature judgment result belongs is determined, and when the comprehensive temperature judgment result is an equipment imbalance type inflation abnormality category, an identification self-inspection step is performed to determine the cause of the inflation abnormality.

2. The method for detecting defects in automobile seats based on image recognition according to claim 1, characterized in that: The step of constructing the point cloud structure model of the powered car seat includes: Capturing visual images of powered car seats; Wherein, the visual image includes a color image and a height image, and the color image and the height image are converted into a three-dimensional point cloud; Obtaining color and height identification information corresponding to each point in the visual image, converting them into corresponding identification values, and superimposing and storing each identification value in the form of a mark into a corresponding point cloud structure; Wherein, the height identification information is the actual height value of each collection point; The point cloud structure is divided into functional areas based on the functional area division standard, and the functional areas corresponding to each point cloud structure are superimposed and stored in the corresponding point cloud structure in the form of marks.

3. The method for detecting defects in automobile seats based on image recognition according to claim 1, characterized in that: The continuous detection of the texture of the car seat surface includes: Acquire a reflection image of a surface of a car seat illuminated by a movable light source, and divide the reflection image into a plurality of reflection image blocks based on a reflection image block division standard; Traversing each reflected image block, obtaining the actual texture angle in the adjacent reflected image block having a common edge relationship with the reflected image block; Obtaining an actual texture angle change difference between the reflection image block and an adjacent reflection image block and comparing it with a standard texture angle change threshold to obtain continuous detection results; If the first continuous detection result is obtained, there is a damage defect in the reflected image block, triggering a damage reminder; If the second continuous detection result is obtained, there is no damage defect in the reflected image block, and the bulge degree determination is performed.

4. The method for detecting defects in automobile seats based on image recognition according to claim 3, characterized in that: The execution of the bulge degree determination includes: Acquire the actual height value corresponding to the actual height information of any sampling point in each functional area based on the point cloud data, and compare the actual height value with the standard actual height range of the functional area where the sampling point is located to obtain a height comparison result; Obtaining the swelling degree judgment result corresponding to each height comparison result, and the category to which the swelling degree judgment result belongs; If the first result of the swelling degree judgment is obtained, the swelling area is superimposed and stored in the corresponding point cloud structure in the form of a mark, and a comprehensive judgment of the material filling is performed; If the third result of the bulging degree determination is obtained, the suture correlation analysis is performed.

5. The method for detecting defects in automobile seats based on image recognition according to claim 4, characterized in that: The execution of comprehensive material filling judgment includes: The comprehensive judgment of material filling includes material detection steps and filling adjustment steps; The material detection step is used to determine whether the material of the functional area corresponding to the sampling point is in a performance fatigue state; If the material detection result is a material performance fatigue result, a reminder to adjust the material is triggered; If the material test result shows that the material performance is stable, perform the filling adjustment step; The filling material quality within the factory inspection parameters corresponding to the functional area is obtained through the filling adjustment step, triggering a reminder to adjust the filling material quality.

6. The method for detecting defects in automobile seats based on image recognition according to claim 4, characterized in that: The suture correlation analysis includes: Obtaining a swelling area formed by sampling points with swelling marks, and extracting the acquisition point at the geometric center of the swelling area and the coordinates of each sampling point on the outline of the swelling area; Obtain the suture path segment of the functional area where the collection point at the geometric center of the bulging area is located and the corresponding suture path segment constraint area; Starting from the acquisition point at the geometric center, draw several candidate perpendicular lines perpendicular to the suture path segment, and obtain the target perpendicular line with the shortest distance to the segment constraint area; Obtaining the intersection of the bulging area outline and the target vertical line as a judgment collection point, and superimposing and storing the judgment collection point in the form of a mark into the point cloud structure corresponding to the collection point; Determining whether the determination acquisition point is within the suture path segment constraint area; If the collection point is determined to be within the suture path segment constraint area, the bulging area is the suture influence area, and the suture process and support judgment is performed; If it is determined that the collection point is not within the suture path segment constraint area, the bulging area is the suture extension area, and a comprehensive temperature judgment is performed.

7. The method for detecting defects in automobile seats based on image recognition according to claim 6, characterized in that: The suture process and support judgment includes suture process identification, wherein: Obtain the suture height and suture shape of the suture path segment detected at the factory; Obtaining the overlapping portion of the suture influence area and the suture path segment constraint area, and superimposing and storing the overlapping area in the form of a mark into the point cloud structure corresponding to each acquisition point of the overlapping portion; Obtaining the curvature of each acquisition point in the overlapping area, traversing the curvature of each acquisition point, and obtaining the curvature of the acquisition point that is adjacent to the acquisition point; Obtaining an actual curvature change difference between the acquisition point and an adjacent acquisition point and comparing it with a preset curvature change threshold, and obtaining a corresponding suture process recognition result based on the curvature comparison result; If the first curvature comparison result is obtained, corresponding to the stretch suture process identification result, the suture support judgment is performed; If the second curvature comparison result is obtained, corresponding to the stacking stitching process identification result, a material stacking reminder is triggered.

8. The method for detecting defects in automobile seats based on image recognition according to claim 6, characterized in that: The suture process and support judgment also includes suture support judgment, wherein, Obtaining the endpoints of the suture path segments, and extracting the three-dimensional coordinates of the sampling points at the endpoints in the point cloud structure and their corresponding actual height values; Extracting the standard suture height value corresponding to the factory requirement standard of the functional area where the endpoint is located, calculating the actual suture height difference value and comparing the actual suture height difference value with the standard suture height difference threshold value; Among them, the actual suture height value is the absolute value of the difference between the standard suture height value and the actual height value; If the first seam height difference comparison result is obtained, a lockstitch connection failure reminder is triggered; If the second suture height difference comparison result is obtained, a reminder to adjust the filling is triggered.

9. The method for detecting defects in automobile seats based on image recognition according to claim 6, characterized in that: The comprehensive temperature judgment includes eliminating the interference of the real-time working environment temperature, wherein: Determine whether the real-time working environment temperature and the preset standard working environment temperature are both within the linear change range of the temperature-expansion curve; If both ambient temperatures are within the linear range of the temperature-expansion curve, a linear compensation calculation method is used to eliminate the expansion effect caused by the real-time working ambient temperature, and after elimination, an expansion temperature correlation judgment is performed; If any ambient temperature is not within the linear change range of the temperature-expansion curve, an ambient temperature interference reminder is triggered, and the expansion temperature association judgment is performed after the real-time working ambient temperature returns to the standard working ambient temperature.

10. The method for detecting defects in automobile seats based on image recognition according to claim 9, characterized in that: The comprehensive temperature judgment also includes a bulging temperature correlation judgment, wherein: Acquire each collection point in the area directly affected by the heating wire corresponding to each functional area, and store the area directly affected by the heating wire in the form of a mark in the point cloud structure corresponding to each collection point; Traverse each collection point in the area directly affected by the heating wire to determine whether there is a mark for the bulging area; If there is a mark of a swelling area, obtain the maximum height information of each collection point with the swelling area mark and compare it with the standard heating height threshold; If the maximum height information is greater than or equal to the standard heating height threshold, a reminder to adjust the heating wire position is triggered; If the maximum altitude information is less than the standard heated altitude threshold, there is no abnormality with the powered seat; If there is no mark for the bulging area, a reminder to adjust the filling quality is triggered.

Images