Paint surface orange peel defect detection method based on visual detection equipment

Abstract

The invention discloses a paint surface orange peel defect detection method based on visual detection equipment. The visual detection equipment comprises a light source and a camera; the light source continuously generates uniform light strips on the surface of the to-be-detected object and moves relative to the to-be-detected object in the detection station, and the camera collects light strip images; extracting edge lines on two sides of a light strip for each frame of image; recording the extending direction of the light source as a direction A and a direction perpendicular to the direction A as a direction B; starting from the initial position of the single-frame image, searching the number of edge points in the B direction pixel by pixel along the A direction on the single-side edge line, and if the number is greater than 1, marking each currently searched edge point as a defect point; sequentially splicing all the frames of images to form an overall image of the surface of the object to be detected, and marking all the defect points on the overall image to complete the detection of the orange peel defects; according to the method, the orange peel defect position is determined based on the overlapping condition of the light strip edge in the specific direction; the method has the characteristics of low false detection rate, high detection speed and strong anti-interference capability.

Description

Technical Field

[0001] This invention relates to the field of paint surface quality inspection, specifically to a method for detecting orange peel defects in paint surfaces based on visual inspection equipment. Background Technology

[0002] Paint is mainly used for the protection and decoration of surfaces such as industrial equipment, steel structures, and sheet metal parts. It provides functions such as corrosion prevention, rust prevention, insulation, and aesthetic enhancement, extending the service life of equipment. For example, the automotive manufacturing process includes painting, used to protect sheet metal parts and enhance the vehicle's appearance. After a workpiece is painted, the paint surface needs quality inspection. Orange peel defects are a typical type of defect, characterized by an uneven texture resembling orange peel. The causes involve various factors including paint materials, painting techniques, environmental factors, and substrate treatment.

[0003] Existing paint orange peel detection technologies are mainly divided into two categories: manual detection and automated detection.

[0004] Manual inspection relies on the experience of inspectors, who judge defects by visually observing the distribution of reflected light on the paint surface or by touch. This method is inefficient, subjective, and prone to false positives, making it difficult to meet the needs of large-scale industrial production.

[0005] Automated inspection technology uses machine vision and image processing algorithms to identify defects, and mainly includes the following methods:

[0006] 1) Methods based on grayscale feature analysis analyze the uniformity of surface texture by calculating the grayscale histogram, mean, or variance of the paint surface image, thereby determining whether orange peel defects exist. This method is susceptible to the influence of lighting conditions; when the lighting is uneven, the grayscale features will fluctuate significantly, leading to a decrease in detection accuracy.

[0007] 2) Texture feature extraction-based methods: These methods employ algorithms such as wavelet transform and Gabor filtering to extract texture features from the painted surface, identifying defects through differences in feature vectors. However, these methods require complex feature engineering, involve high computational costs, and have poor real-time performance, making them difficult to apply to online inspection scenarios on production lines. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention provides a method for detecting orange peel defects on paint surfaces based on a visual inspection device. This method utilizes the principle of high reflectivity of the paint surface, employing a long strip light source to project light strips onto the paint surface (the surface of the object to be tested), acquiring images of the light strips through a camera, and then analyzing the overlap of the light strip edges in a specific direction to determine the location of the orange peel defect. It features a low false detection rate, fast detection speed, and strong anti-interference capability.

[0009] The technical solution is as follows:

[0010] A method for detecting orange peel defects on paint surfaces based on a visual inspection device, wherein the visual inspection device is set in an inspection station and includes a light source and a camera; the inspection station is isolated from natural light and is in a dark state;

[0011] The light source is elongated and used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera.

[0012] The camera is fixedly installed facing the object to be tested; the light source and the object to be tested move relative to each other at the detection station. At the same time, the camera acquires light strip images at a preset frame rate, and multiple frames of light strip images are stored sequentially in the detection image set.

[0013] Detect orange peel defects using the following methods:

[0014] ① Extract the edge lines on both sides of the light stripe for each frame of the detection image set;

[0015] ② Perform the following processing on each frame of the image:

[0016] Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B;

[0017] Starting from the beginning of a single frame image, along the edge line on one side, find the number of edge points in the B direction pixel by pixel along the A direction. If the number is greater than 1, mark each edge point found as a defect point.

[0018] Perform the same search on the other side of the edge line and mark the defect points on that edge line;

[0019] ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image to complete the detection of orange peel defects.

[0020] The present invention also relates to another method for detecting orange peel defects on paint surfaces based on a visual inspection device, wherein the visual inspection device is set in an inspection station and includes a light source and a camera; the inspection station is isolated from natural light and is in a dark state;

[0021] The camera is fixedly installed facing the object being tested.

[0022] The light source is elongated and used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera.

[0023] The light source can rotate sequentially to the specified installation angle according to the preset angle step size;

[0024] Each time the installation angle of the light source is rotated, the light source and the object under test are made to move relative to each other again in the detection station. At the same time, the camera acquires light strip images at a preset frame rate and stores multiple frames of light strip images into the detection image set in sequence.

[0025] Multiple detection image sets were obtained under different light source installation angles;

[0026] Detect orange peel defects using the following methods:

[0027] 1) Perform steps ①~③ on multiple detection image sets to obtain multiple overall images:

[0028] ① Extract the edge lines on both sides of the light stripe for each frame of the detection image set;

[0029] ② Perform the following processing on each frame of the image:

[0030] Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B;

[0031] Starting from the beginning of a single frame image, along the edge line on one side, find the number of edge points in the B direction pixel by pixel along the A direction. If the number is greater than 1, mark each edge point found as a defect point.

[0032] Perform the same search on the other side of the edge line and mark the defect points on that edge line;

[0033] ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image;

[0034] 2) Summarize the locations of defect points in different overall images, mark all defect points in one overall image, and complete the detection of orange peel defects.

[0035] Preferably, the angle step size is set in the range of 10° to 90°.

[0036] Furthermore, starting from the image's initial position, the method for finding the number of edge points in direction B pixel by pixel along direction A on a single-sided edge line includes the following two methods:

[0037] Method I: Starting from the beginning position of the image, draw auxiliary lines in direction B pixel by pixel along direction A, find the intersection points of the auxiliary lines and the single-sided edge lines. If there is more than one intersection point, mark each current intersection point as a defect point.

[0038] Method II: Determine whether the installation angle of the light source is horizontal / vertical;

[0039] If not, first calculate the angle required for the light source to rotate to horizontal / vertical, and based on the angle value, rotate all edge points in the image to make the light strip horizontal / vertical; then proceed to step I.

[0040] If so, proceed directly to step I;

[0041] Step I is as follows: starting from the starting position of the image, obtain the coordinate components of each edge point in the horizontal light strip in the vertical direction, or obtain the coordinate components of the vertical light strip in the horizontal direction.

[0042] If there are identical coordinate components, the edge points corresponding to the identical coordinate components are marked as defect points.

[0043] Compared with the prior art, the present invention has the following characteristics:

[0044] 1. Reduced false positive rate:

[0045] Because the external manifestation of orange peel defects is "roundabout" (such as...) Figure 1a This is reflected in the light stripe image as "wrinkles" at the edges of the light stripes (e.g., Figure 1c (The wrinkled lines are marked in green within the dashed box). Because the wrinkled features of orange peel defects are quite subtle, it is difficult to accurately identify orange peel by analyzing the continuity or curvature of the light stripe edge lines alone, often resulting in missed or false detections (such as misjudging slight surface scratches as orange peel).

[0046] This invention directly utilizes the spatial distribution characteristics of edges to analyze the overlap of light stripe edges in a specific direction (perpendicular to the light stripe direction). Overlapping edge points are used to characterize "wrinkles," indicating that the edges exhibit reciprocating curvature in a specific direction. Normal edges, on the other hand, are smooth and continuous with no overlapping edge points. This method significantly improves the intuitiveness and accuracy of orange peel recognition and reduces the false detection rate.

[0047] 2. Improved detection efficiency:

[0048] This method identifies the defect location based on the geometric features of the orange peel defect contour. It eliminates the need for complex feature recognition and parameter settings, significantly reducing computational complexity. Traditional methods take an average of 30ms to process one image, while this invention reduces the processing time to less than 10ms, improving detection efficiency and meeting the real-time requirements of industrial online inspection. Attached Figure Description

[0049] Figure 1a This is a schematic diagram of the outline of a unilateral light stripe extracted from a single frame image in Example 1; Figure 1b For Example 1 Figure 1aA schematic diagram showing the marked edge lines on one side of the middle section;

[0050] Figure 1c This is a schematic diagram of marking defect points (green pixels in the dashed box) on one side of the edge line in Example 1;

[0051] Figure 2 This is a schematic diagram of a single frame image in Example 1;

[0052] Figure 3 This is a schematic diagram of a single light stripe image when the light source is installed at a vertical (90°) angle in Example 2;

[0053] Figure 4 This is a schematic diagram of a single light stripe image when the installation angle of the light source is 135° in Example 2. Detailed Implementation

[0054] The technical solution of the present invention will be described in detail below with reference to the embodiments.

[0055] Example 1:

[0056] A method for detecting orange peel defects in paint surfaces based on a visual inspection device, wherein the visual inspection device is set in an inspection station and includes a light source and a camera; the inspection station is isolated from natural light and is in a dark state;

[0057] The light source is long and narrow, used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera.

[0058] The camera is fixedly installed facing the object to be tested; the light source and the object to be tested move relative to each other at the detection station. At the same time, the camera acquires light strip images at a preset frame rate, and multiple frames of light strip images are stored sequentially in the detection image set.

[0059] Explanatoryly, during the test, the light source and the object under test move relative to each other to ensure that the light strip is projected onto different positions of the object under test. While the relative movement is continuous, the camera captures multiple frames of images, and the stitched images can cover the area under test.

[0060] Detect orange peel defects using the following methods:

[0061] ① For each frame of the detection atlas (e.g.) Figure 2 Extract the edge lines on both sides of the light stripe (e.g., using the Canny edge detection algorithm).

[0062] ② Perform the following processing on each frame of the image:

[0063] Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B;

[0064] Starting from the beginning position of a single frame image, along the edge line on one side ( Figure 1b The algorithm searches for the number of edge points along direction A in direction B pixel by pixel. If the number is greater than 1, then all the edge points found are marked as defect points. Figure 1c );

[0065] Perform the same search on the other side of the edge line and mark the defect points on that edge line;

[0066] ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image to complete the detection of orange peel defects.

[0067] Specifically, the relative movement between the light source and the object under test at the testing station includes the following two methods:

[0068] Method 1: The light source is fixedly installed. After the object to be tested enters the testing station, it is moved at a constant speed (e.g., placed on a conveyor belt). At this time, the pixel distance of the object to be tested between two adjacent frames is pre-calibrated. The pixel distance is compensated to each frame and then the frames are stitched together in sequence. The gray value of the pixels in the overlapping area is averaged to form an overall image of the surface of the object to be tested.

[0069] Method 2: The object to be tested is fixed in the detection station, and the light source moves at a constant speed (e.g., installed at the end of a robot to scan the area to be tested). At this time, the images of each frame are stitched together pixel by pixel, and the grayscale of the pixels in the overlapping area is averaged to form an overall image of the surface of the object to be tested.

[0070] In practice, starting from the image's initial position, the method of counting edge points along direction B pixel by pixel along direction A on a single edge line includes the following two approaches:

[0071] Method I: Starting from the beginning position of the image, draw auxiliary lines in direction B pixel by pixel along direction A, find the intersection points of the auxiliary lines and the single-sided edge lines. If there is more than one intersection point, mark each current intersection point as a defect point.

[0072] Method II: Determine whether the installation angle of the light source is horizontal / vertical. If not, first calculate the angle required to rotate the light source to horizontal / vertical. Based on the angle value, rotate all edge points in the image to make the light strip horizontal / vertical; then proceed to step I.

[0073] If so, proceed directly to step I;

[0074] Step I is as follows: starting from the starting position of the image, obtain the coordinate components of each edge point in the horizontal light strip in the vertical direction, or obtain the coordinate components of the vertical light strip in the horizontal direction.

[0075] If there are identical coordinate components, the edge points corresponding to the identical coordinate components are marked as defect points.

[0076] Example 2

[0077] Since the outline of the orange peel defect cannot be reflected on the edge of the light strip when it is parallel to the direction of the light source extension (direction A), it may lead to missed detection. In order to reduce the false detection rate, this embodiment rotates the light source to obtain different installation angles to avoid the orange peel defect outline being parallel to a certain light source installation direction, which may lead to missed detection.

[0078] The specific process is as follows:

[0079] A method for detecting orange peel defects in paint surfaces based on a visual inspection device, wherein the visual inspection device is set in an inspection station and includes a light source and a camera; the inspection station is isolated from natural light and is in a dark state;

[0080] The camera is fixedly mounted facing the object being measured.

[0081] The light source is long and narrow, used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera.

[0082] In this embodiment, the installation angle of the light source is adjustable (e.g., it is installed on a rotatable bracket or at the end of a robot), and it can rotate to a specified angle sequentially according to a preset angle step (set range 10°~90°).

[0083] Each time the installation angle of the light source is rotated, the light source and the object under test are re-moved relative to each other in the detection station. At the same time, the camera acquires light strip images at a preset frame rate and stores multiple frames of light strip images into the detection image set in sequence.

[0084] Multiple detection image sets were obtained under different light source installation angles;

[0085] Detect orange peel defects using the following methods:

[0086] 1) Perform steps ①~③ on multiple detection image sets to obtain multiple overall images:

[0087] ① Extract the edge lines on both sides of the light stripe for each frame of the detection image set;

[0088] ② Perform the following processing on each frame of the image:

[0089] Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B;

[0090] Starting from the beginning of a single frame image, along the edge line on one side, find the number of edge points in the B direction pixel by pixel along the A direction. If the number is greater than 1, mark each edge point found as a defect point.

[0091] Perform the same search on the other side of the edge line and mark the defect points on that edge line;

[0092] ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image;

[0093] 2) Summarize the locations of defect points in different overall images, mark all defect points in one overall image, and complete the detection of orange peel defects.

[0094] Taking an angle step of 45° as an example, the following is an illustrative example:

[0095] The installation angles of the light source are horizontal (0°), rotated 45°, vertical (90°), and rotated 135°.

[0096] After each adjustment of the installation angle of the light source, the light source and the object under test are made to move relative to each other again in the detection station. The camera acquires light strip images at a preset frame rate and stores multiple frames of light strip images into the detection image set in sequence.

[0097] Finally, four sets of detection images were obtained under four different installation angles; among them, such as Figure 3 This is a single-frame light stripe image captured when the image is vertical (90°). Figure 4 This is a single-frame light stripe image captured when rotated 135°;

[0098] Each acquired detection image set is stored in relation to the corresponding light source installation angle to obtain a set of detection images. Steps ① to ③ are performed on each set of detection images to obtain four overall images.

[0099] Next, the locations of the defects in the four overall images are summarized, and all defect points are marked on the overall images to complete the detection of orange peel defects.

[0100] In practice, starting from the image's initial position, the method of counting edge points along direction B pixel by pixel along direction A on a single edge line includes the following two approaches:

[0101] Method I: Starting from the initial position of the image, draw auxiliary lines in direction B pixel by pixel along direction A (e.g., ... Figure 3 , 4 Find the intersection of the auxiliary line and the single-sided edge line. If there is more than one intersection, mark each intersection as a defect point.

[0102] Method II: Determine whether the installation angle of the light source is horizontal / vertical. If not, first calculate the angle required to rotate the light source to horizontal / vertical. Based on the angle value, rotate all edge points in the image to make the light strip horizontal / vertical; then proceed to step I.

[0103] If so, proceed directly to step I;

[0104] Step I is as follows: starting from the starting position of the image, obtain the coordinate components of each edge point in the horizontal light strip in the vertical direction, or obtain the coordinate components of the vertical light strip in the horizontal direction.

[0105] If there are identical coordinate components, the edge points corresponding to the identical coordinate components are marked as defect points.

[0106] like: Figure 4 This is a single-frame light stripe image acquired when rotated 135°; the steps for Method II are as follows:

[0107] First, calculate the angle of 45° required for the light source to rotate to horizontal / vertical. Based on the angle value, rotate all edge points in the image by 45 degrees to the right to make the light strip vertical; then proceed to step I.

[0108] Step I is as follows: Starting from the starting position of the image, obtain the coordinate components of each edge point in the vertical direction. If there are the same coordinate components, mark the pixel points corresponding to the same coordinate components as defect points.

[0109] This invention directly utilizes the spatial distribution characteristics of edges to analyze the overlap of light stripe edges in a specific direction (perpendicular to the light stripe direction), and uses the overlapping edge points to represent "wrinkles." This significantly improves the intuitiveness and accuracy of orange peel recognition and reduces the false detection rate.

[0110] Based on the geometric features of the orange peel defect contour, the defect location can be identified without the need for complex feature recognition and parameter settings, which greatly reduces the computational complexity. Traditional methods take an average of 30ms to process an image, while this invention shortens the processing time to less than 10ms, improving detection efficiency and meeting the real-time requirements of industrial online inspection.

[0111] The foregoing description of specific exemplary embodiments of the present invention is for illustrative and descriptive purposes. It is not intended to be exhaustive, nor to limit the invention to the precise forms disclosed; obviously, many changes and variations are possible in accordance with the foregoing teachings. The exemplary embodiments were chosen and described to explain the specific principles of the invention and its practical application, thereby enabling others skilled in the art to implement and utilize various exemplary embodiments of the invention, as well as their different alternatives and modifications. The scope of the invention is intended to be defined by the appended claims and their equivalents.

Claims

1. A method for detecting orange peel defects in paint surfaces based on a visual inspection device, wherein the visual inspection device is installed in an inspection station and includes a light source and a camera; characterized in that, The testing station is isolated from natural light and is in a dark state. The light source is elongated and used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera. The camera is fixedly installed facing the object to be tested; the light source and the object to be tested move relative to each other at the detection station. At the same time, the camera acquires light strip images at a preset frame rate, and multiple frames of light strip images are stored sequentially in the detection image set. Detect orange peel defects using the following methods: ① Extract the edge lines on both sides of the light stripe for each frame of the detection image set; ② Perform the following processing on each frame of the image: Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B; Starting from the beginning of a single frame image, along the edge line on one side, find the number of edge points in the B direction pixel by pixel along the A direction. If the number is greater than 1, mark each edge point found as a defect point. Perform the same search on the other side of the edge line and mark the defect points on that edge line; ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image to complete the detection of orange peel defects.

2. A method for detecting orange peel defects in paint surfaces based on a visual inspection device, wherein the visual inspection device is installed in the inspection station and includes a light source and a camera; characterized in that, The testing station is isolated from natural light and is in a dark state. The camera is fixedly installed facing the object being tested. The light source is elongated and used to continuously generate uniform light stripes on the surface of the object under test. The reflected light from the light stripes on the surface of the object under test can meet the imaging requirements of the camera. The light source can rotate sequentially to the specified installation angle according to the preset angle step size; Each time the installation angle of the light source is rotated, the light source and the object under test are made to move relative to each other again in the detection station. At the same time, the camera acquires light strip images at a preset frame rate and stores multiple frames of light strip images into the detection image set in sequence. Multiple detection image sets were obtained under different light source installation angles; Detect orange peel defects using the following methods: 1) Perform steps ①~③ on multiple detection image sets to obtain multiple overall images: ① Extract the edge lines on both sides of the light stripe for each frame of the detection image set; ② Perform the following processing on each frame of the image: Let the direction in which the light source extends be direction A, and the direction perpendicular to direction A be direction B; Starting from the beginning of a single frame image, along the edge line on one side, find the number of edge points in the B direction pixel by pixel along the A direction. If the number is greater than 1, mark each edge point found as a defect point. Perform the same search on the other side of the edge line and mark the defect points on that edge line; ③ The images are stitched together sequentially to form an overall image of the surface of the object under test, and each defect point is marked on the overall image; 2) Summarize the locations of defect points in different overall images, mark all defect points in one overall image, and complete the detection of orange peel defects.

3. The method for detecting orange peel defects in paint surfaces based on a visual inspection device as described in claim 2, characterized in that: The angle step size can be set in the range of 10° to 90°.

4. The method for detecting orange peel defects in paint surfaces based on a visual inspection device as described in claim 1 or 2, characterized in that: Starting from the image's initial position, the method for counting edge points along direction B, pixel by pixel, along direction A on a single-sided edge line includes the following two methods: Method I: Starting from the beginning position of the image, draw auxiliary lines in direction B pixel by pixel along direction A, find the intersection points of the auxiliary lines and the single-sided edge lines. If there is more than one intersection point, mark each current intersection point as a defect point. Method II: Determine whether the installation angle of the light source is horizontal / vertical; If not, first calculate the angle required for the light source to rotate to horizontal / vertical, and based on the angle value, rotate all edge points in the image to make the light strip horizontal / vertical; then proceed to step I. If so, proceed directly to step I; Step I is as follows: starting from the starting position of the image, obtain the coordinate components of each edge point in the horizontal light strip in the vertical direction, or obtain the coordinate components of the vertical light strip in the horizontal direction. If there are identical coordinate components, the edge points corresponding to the identical coordinate components are marked as defect points.

Images