A hyperbolic plate coating uniformity evaluation method and system based on machine vision
By using machine vision-based methods to partition and analyze images of hyperboloids, the problem of accuracy in evaluating the uniformity of coatings on complex curvature surfaces was solved, and high-precision evaluation of the uniformity of hyperboloid coatings was achieved.
Patent Information
- Application Number
- CN202511396351.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-28
- Publication Date
- 2025-12-09
- Estimated Expiration
- 2045-09-28
AI Technical Summary
Existing methods for evaluating the uniformity of coatings on hyperboloid plates are subject to large detection errors due to the complex curvature variations of the hyperboloid plates and uneven illumination, making it difficult to achieve accurate evaluation.
A machine vision-based approach is adopted to acquire hyperbolic plate images, process and calculate curvature region image sets, generate multi-view images using a neural radiation field model, extract brightness and hue information, calculate gradient difference thresholds, generate transformed images, and finally evaluate coating uniformity.
It improves the accuracy of coating uniformity detection on hyperboloids, enabling accurate assessment of coating uniformity on complex curvature surfaces and reducing the impact of uneven lighting.
Smart Images
Figure CN120876490B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of image processing, in particular to a hyperbolic plate coating uniformity evaluation method and system based on machine vision. BACKGROUND
[0002] The hyperbolic plate is a complex three-dimensional curved plate material with opposite bending radii in two perpendicular directions (one positive and one negative), resembling a saddle or a cooling tower.
[0003] The existing method often uses an image analysis method based on gray variance to evaluate the coating uniformity of the hyperbolic plate.
[0004] However, due to the complex curvature variation of the hyperbolic plate surface, it is easy to cause uneven illumination and projection distortion, so that the calculation method directly dependent on the gray variance is greatly disturbed, the detection error is significant, and it is difficult to realize accurate evaluation. SUMMARY
[0005] The present application provides a hyperbolic plate coating uniformity evaluation method and system based on machine vision to solve the existing problems.
[0006] The hyperbolic plate coating uniformity evaluation method based on machine vision provided by the present application adopts the following technical scheme:
[0007] One embodiment of the present application provides a hyperbolic plate coating uniformity evaluation method based on machine vision, which comprises the following steps:
[0008] Obtain a hyperbolic plate image, wherein the hyperbolic plate image is an RGB image;
[0009] Obtain a curvature region image set from the hyperbolic plate image, wherein the curvature region image set includes images corresponding to different curvature regions;
[0010] Determine a perspective image set from the curvature region image set, convert the perspective image set into an HIS image and extract the I channel to obtain a luminance image set;
[0011] Calculate the image quality of each image in the luminance image set, and determine a target perspective image in each luminance image set, wherein the target perspective image is the image with the maximum image quality in each perspective image set;
[0012] Extract the image corresponding to each target perspective image from the perspective image set to obtain a uniformity analysis image, convert the uniformity analysis image into an HIS image and extract the H channel to obtain a hue analysis image;
[0013] Calculate the gradient difference threshold of the hue analysis image, convert the hue analysis image according to the gradient difference threshold to obtain a converted image;
[0014] According to the transformed image, the uniformity of the uniformity analysis image is calculated, and the uniformity quality of the hyperbolic plate is determined according to the uniformity of the uniformity analysis image.
[0015] Optionally, a curvature region image set is acquired according to the hyperbolic plate image, and specifically includes:
[0016] A triangular network three-dimensional model is determined according to the hyperbolic plate image.
[0017] The curvature of each position point in the triangular network three-dimensional model is calculated based on the normal vector change rate.
[0018] The curvature of each position point is sorted to obtain a curvature sequence.
[0019] The curvature sequence is segmented using an otsu multi-threshold segmentation method to obtain a curvature segmentation section.
[0020] The first position point and the second position point in the triangular network three-dimensional model are merged to obtain a curvature region, wherein the first position point and the second position point are any position points in the triangular network three-dimensional model, and the first position point and the second position point are adjacent and in the same curvature segmentation section.
[0021] An image corresponding to each curvature region is cropped from the hyperbolic plate image to obtain a curvature region image set.
[0022] Optionally, a view angle image set is determined according to the curvature region image set, and specifically includes:
[0023] Each image of each curvature region in the curvature region image set is input into a trained neural radiance field model, and a multi-view image corresponding to each image is output by the trained neural radiance field model, wherein the trained neural radiance field model is trained according to the hyperbolic plate image.
[0024] Each image of each curvature region in the curvature region image set and the corresponding multi-view image are determined as the view angle image set.
[0025] Optionally, the image quality of each image in the brightness image set is calculated, and specifically includes:
[0026] Each image in the brightness image set is detected using an otsu multi-threshold segmentation method to obtain a segmentation threshold, and the ratio of the number of pixel points in the Nth image in the brightness image set that are greater than the segmentation threshold of the Nth image to the number of pixel points in the Nth image is determined as the highlight point ratio of the Nth image.
[0027] The entropy value of the gray level co-occurrence matrix of the Nth image in the brightness image set is calculated, and the entropy value is normalized to obtain the gray level entropy of the Nth image.
[0028] Multiply the highlight point proportion of the Nth image with the gray entropy of the Nth image to obtain the image quality of the Nth image.
[0029] Optionally, the gradient difference threshold of the tonal analysis image is calculated, specifically comprising:
[0030] Obtain the gradient of each pixel point in the tonal analysis image to obtain a gradient map;
[0031] Determine the pixel points greater than the preset gradient threshold in the gradient map as gradient pixel points;
[0032] Determine the minimum gradient value in the gradient pixel points as the gradient threshold;
[0033] Calculate the difference between the gradient mode in the gradient map and the gradient threshold, take the absolute value of the difference, and calculate the ratio of the absolute value to the maximum gradient value in the gradient map to obtain the gradient difference threshold.
[0034] Optionally, the tonal analysis image is converted according to the gradient difference threshold to obtain a converted image, specifically comprising:
[0035] Determine a first tonal pixel point from the tonal analysis image, and determine a second tonal pixel point adjacent to the first tonal pixel point in the eight-neighborhood;
[0036] Calculate the difference between the hue value of the second tonal pixel point and the hue value of the first tonal pixel point, and divide the absolute value of the difference by the hue value of the first tonal pixel point to obtain the abnormality degree of the second tonal pixel point compared to the first tonal pixel point;
[0037] When the abnormality degree is greater than the gradient difference threshold and the number of the second tonal pixel points is greater than 3, determine the first tonal pixel point as a difference pixel point;
[0038] Determine the hue value of the second tonal pixel point corresponding to the maximum abnormality degree as the hue value of the difference pixel point, and re-determine the difference pixel point as the first tonal pixel point until the abnormality degree of the second tonal pixel point compared to the first tonal pixel point is less than the gradient difference threshold, or the abnormality degree is greater than the gradient difference threshold and the number of the second tonal pixel points is less than 4;
[0039] Convert each pixel point in the tonal analysis image as the first tonal pixel point to obtain the converted image.
[0040] Optionally, the uniformity of the uniformity analysis image is calculated according to the converted image, specifically comprising:
[0041] Calculate the hue variance of all pixel points in the converted image to obtain the difference degree;
[0042] Subtract the tonal analysis image from the converted image to obtain a tonal layer;
[0043] determining the pixel points in the hue layer corresponding to the pixel points in the conversion image as conversion pixel points, and determining the region formed by the conversion pixel points as a conversion connected domain;
[0044] obtaining the sum of the conversion times of the pixel points in the conversion connected domain;
[0045] normalizing the sum of the conversion times to obtain a conversion frequency;
[0046] determining the product of the conversion frequency and a discriminant as the uniformity of the uniformity analysis image, wherein the discriminant is the negative of the difference degree and is obtained as an exponential of a natural base.
[0047] Optionally, the uniformity quality of the hyperbolic panel is determined according to the uniformity of the uniformity analysis image, and specifically includes:
[0048] determining the uniformity of the uniformity analysis image as the uniformity of the corresponding image in the curvature region image set to obtain the uniformity of each image in the curvature region image set;
[0049] determining the coating uniformity of each curvature region in the hyperbolic panel image according to the uniformity of each image in the curvature region image set;
[0050] determining the overall uniformity of the hyperbolic panel according to the coating uniformity of each curvature region in the hyperbolic panel image;
[0051] when the overall uniformity is less than a preset first uniformity threshold, it is determined that the uniformity quality of the hyperbolic panel has a major defect; when the overall uniformity is greater than or equal to the preset first uniformity threshold and the coating uniformity of the target curvature region is less than a preset second uniformity threshold, it is determined that the uniformity quality of the target curvature region of the hyperbolic panel has a major defect; and when the overall uniformity is greater than or equal to the preset first uniformity threshold and the coating uniformity is all greater than or equal to the preset second uniformity threshold, it is determined that the uniformity quality of the hyperbolic panel has no defect.
[0052] Optionally, a triangular network three-dimensional model is determined according to the hyperbolic panel image, and specifically includes:
[0053] inputting the hyperbolic panel image and the viewing angle parameter corresponding to each hyperbolic panel image into a neural radiance field model for training to obtain a trained neural radiance field model;
[0054] constructing a three-dimensional density field data through the trained neural radiance field model, inputting the three-dimensional density field data into a marching cubes algorithm model, and outputting a triangular network three-dimensional model through the marching cubes algorithm model.
[0055] The application provides a hyperbolic plate coating uniformity evaluation system based on machine vision, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, and the computer program realizes the steps of the hyperbolic plate coating uniformity evaluation method based on machine vision when executed by the processor.
[0056] The technical scheme of the application has the beneficial effects that:
[0057] In the embodiment of the application, the image obtained by the hyperbolic plate is divided into partitions, and then the image with the least light influence in each partition is obtained, the transformed image after difference amplification is obtained through transformation on the image, and then the coating uniformity evaluation value of each image is obtained according to the difference degree and the transformation difficulty of the transformed image, so that the precision of the coating uniformity detection of the hyperbolic plate is improved. BRIEF DESCRIPTION OF DRAWINGS
[0058] In order to more clearly illustrate the technical scheme in the embodiments of the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or the prior art description. Obviously, the drawings in the following description only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.
[0059] Figure 1 A flow chart of a hyperbolic plate coating uniformity evaluation method based on machine vision provided by an embodiment of the application;
[0060] Figure 2 A corresponding schematic diagram of multi-view images;
[0061] Figure 3 A structure diagram of a hyperbolic plate coating uniformity evaluation system based on machine vision provided by an embodiment of the application. DETAILED DESCRIPTION
[0062] In order to further illustrate the technical means and effects adopted by the application to achieve the predetermined application purpose, the following describes the specific implementation, structure, features and effects of the hyperbolic plate coating uniformity evaluation method based on machine vision according to the application in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.
[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application belongs.
[0064] Specifically, the application provides a specific scheme of a hyperbolic plate coating uniformity evaluation method based on machine vision.
[0065] The application provides a hyperbolic plate coating uniformity evaluation method and system based on machine vision. Figure 1 Fig. 1 shows a flowchart of a hyperbolic plate coating uniformity evaluation method based on machine vision, and the method comprises the following steps:
[0066] S101, acquiring a hyperbolic plate image, wherein the hyperbolic plate image is an RGB image.
[0067] In a specific embodiment, the hardware and parameters for better acquiring the hyperbolic plate image are as follows:
[0068] The hardware comprises: a camera: a high-resolution, global shutter industrial camera, which ensures no motion blur; a lens: a preferred image-side telecentric lens, which fundamentally eliminates perspective distortion; illumination: an integrating sphere dome light must be used, which provides uniform diffuse light and completely eliminates reflections and shadows; and a rotating table: a high-precision motorized rotating table, which needs to have an angular accuracy of seconds.
[0069] The parameters comprise: system calibration: before acquisition, a high-precision camera calibration is performed using a calibration plate to obtain lens distortion parameters; path planning: according to the camera field of view angle, the rotation angle interval (for example, 2°) is calculated to ensure that the adjacent image overlap rate is greater than 60%, so as to meet the three-dimensional reconstruction requirement; synchronous acquisition: after the rotating table reaches the predetermined angle, a hardware trigger signal is sent to the camera for exposure, so as to ensure that the angle and the image are strictly corresponding; and parameter setting: a lossless format (for example, TIFF) is used for saving; the exposure time is set to avoid overexposure; and the aperture is set to the best sharpness range.
[0070] S102, acquiring a curvature region image set according to the hyperbolic plate image, wherein the curvature region image set comprises images corresponding to different curvature regions.
[0071] In the embodiment, the curvature region image set is acquired according to the hyperbolic plate image, and specifically comprises the following steps:
[0072] determining a triangular network three-dimensional model according to the hyperbolic plate image;
[0073] calculating the curvature of each position point in the triangular network three-dimensional model based on the normal vector change rate;
[0074] sorting the curvature of each position point to obtain a curvature sequence;
[0075] segmenting the curvature sequence using an otsu multi-threshold segmentation method to obtain a curvature segmentation segment;
[0076] The first position point and the second position point in the triangular network three-dimensional model are merged to obtain a curvature region, wherein the first position point and the second position point are any position points in the triangular network three-dimensional model, and the first position point and the second position point are adjacent and in the same curvature segmentation section.
[0077] An image corresponding to each curvature region is cropped from the hyperbolic plate image to obtain a set of curvature region images.
[0078] Optionally, the triangular network three-dimensional model is determined according to the hyperbolic plate image, and specifically includes the following steps.
[0079] The hyperbolic plate image and the perspective parameter corresponding to each hyperbolic plate image are input into a neural radiance field model to be trained, so as to obtain a trained neural radiance field model.
[0080] The three-dimensional density field data is constructed by using the trained neural radiance field model, the three-dimensional density field data is input into a marching cubes algorithm model, and the triangular network three-dimensional model is output by using the marching cubes algorithm model.
[0081] Exemplarily, the triangular network three-dimensional model is used to represent the specific three-dimensional structure of the hyperbolic plate surface, which is composed of countless vertices and triangular facets.
[0082] Due to the irregular curved surface structure of the hyperbolic plate, when the entire hyperbolic plate is photographed, some areas are always reflective, and thus the uniformity of the coating cannot be directly evaluated by factors such as brightness and hue of different pixel points. Therefore, the hyperbolic plate is divided into different regions according to curvature in this embodiment, the lighting conditions of the same region are similar, and thus the uniformity can be calculated.
[0083] For each point on the three-dimensional model, the curvature of each position point in the model is obtained based on the normal vector change rate.
[0084] All the curvature values are arranged in ascending order to obtain an ascending sequence (in this embodiment, all the curvatures are sorted in ascending order, and in actual cases, the curvatures can also be sorted in descending order, which is not specifically limited here), and multiple curvature segmentation sections are obtained by using the otsu multi-threshold segmentation method on the ascending sequence. The curvature values in the same curvature segmentation section are similar, and the curvature values in different curvature segmentation sections are quite different.
[0085] On the three-dimensional model, for each point, the curvature values of each neighbor point in the eight-neighbor domain of the point and the point are compared, if the curvature values belong to the same segmentation section, the two points are merged, and through multiple mergings, multiple curvature regions can be obtained, and the curvature values of the same curvature region belong to the same segmentation section. For example, the hyperbolic plate is divided into three curvature regions of upper, middle and lower parts, the influence of light on the pixel points in each curvature region is similar, and thus the uniformity can be calculated.
[0086] For each curvature region, the corresponding region on the picture obtained by shooting is cropped to obtain a plurality of images of the curvature region, denoted as a curvature region image set.
[0087] S103, determine a view angle image set according to the curvature region image set, and convert the view angle image set into an HIS image and extract the I channel to obtain a luminance image set.
[0088] In this embodiment, the view angle image set is determined according to the curvature region image set, specifically including:
[0089] Each image of each curvature region in the curvature region image set is input into the trained neural radiance field model, and a multi-view image corresponding to each image is output by the trained neural radiance field model, wherein the trained neural radiance field model is obtained according to the hyperbolic plate image;
[0090] Each image of each curvature region in the curvature region image set and the corresponding multi-view image are determined as the view angle image set.
[0091] Exemplarily, on each curvature region image set, images under different rotation angles of the curvature region can be obtained, and adjacent images obtained by shooting have an overlapping region (based on the shooting process in S101, different rotation angles correspond to different images in the hyperbolic plate image, and one image is shot at one rotation angle, so that the hyperbolic plate image corresponding to each rotation angle can be obtained).
[0092] As shown in Figure 2 , the corresponding schematic diagram of the multi-view image is shown in Figure 2 . In Figure 2 , through the fixed view angle of the camera and the rotation of the hyperbolic plate, images under different rotation angles can be obtained; then through the neural radiance field, images under a plurality of other view angles at each rotation angle can be obtained. In the figure, only four images of the other view angles at the normal view angle under each rotation angle are shown (only as an example), and there can be a plurality of images under the other view angles, not just four.
[0093] The normal view angle image and the image of the other view angle corresponding to each rotation angle are denoted as the view angle image set of the rotation angle.
[0094] S104, calculate the image quality of each image in the luminance image set, and determine a target view angle image in each luminance image set, wherein the target view angle image is the image with the maximum image quality in each view angle image set.
[0095] In this embodiment, the image quality of each image in the luminance image set is calculated, specifically including:
[0096] The otsu multi-threshold segmentation method is used to detect each image in the brightness image set to obtain a segmentation threshold, and a ratio of a number of pixel points greater than the segmentation threshold of the Nth image in the Nth image in the brightness image set to a number of pixel points of the Nth image is determined as a highlight point ratio of the Nth image;
[0097] An entropy value of a gray level co-occurrence matrix of the Nth image in the brightness image set is calculated, and the entropy value is normalized to obtain a gray level entropy of the Nth image;
[0098] The highlight point ratio of the Nth image is multiplied by the gray level entropy of the Nth image to obtain an image quality of the Nth image.
[0099] Exemplarily, for each image in the view angle image set of each rotation angle, the RGB image is converted into an HSI image, and then the I channel is extracted to obtain a brightness image. The otsu threshold segmentation method is used on the brightness image to obtain a threshold k, and pixel points greater than the threshold k are recorded as highlight points. A ratio of the number of highlight points to the number of image pixel points (that is, a highlight point ratio) is denoted by b. For each image in the view angle image set of each rotation angle, an entropy value of a gray level co-occurrence matrix of the image is calculated, and the entropy value is normalized to obtain a gray level entropy s. The greater s is, the more abundant the texture details on the image are, and the more diffuse reflection is.
[0100] For each image in the view angle image set of each rotation angle, the product of b and s is calculated as an image quality, and an image corresponding to the highest image quality is determined as a target view angle image corresponding to each rotation angle.
[0101] S105, an image corresponding to each target view angle image is extracted from the view angle image set to obtain a uniformity analysis image. The uniformity analysis image is converted into an HIS image and the H channel is extracted to obtain a hue analysis image.
[0102] S106, a gradient difference threshold of the hue analysis image is calculated, and the hue analysis image is converted according to the gradient difference threshold to obtain a converted image.
[0103] In this embodiment, the gradient difference threshold of the hue analysis image is calculated, specifically including:
[0104] The gradient of each pixel point in the hue analysis image is obtained to obtain a gradient map;
[0105] The pixel points greater than the preset gradient threshold in the gradient map are determined as gradient pixel points;
[0106] The minimum gradient value in the gradient pixel points is determined as the gradient threshold;
[0107] The difference between the gradient mode in the gradient image and the gradient threshold value is calculated, the absolute value of the difference is taken, and the ratio of the absolute value to the maximum gradient value in the gradient image is calculated to obtain a gradient difference threshold value.
[0108] The hue analysis image is converted according to the gradient difference threshold value to obtain a converted image, specifically including:
[0109] A first hue pixel point is determined from the hue analysis image, and a pixel point adjacent to the first hue pixel point in an eight-neighbor domain is determined as a second hue pixel point.
[0110] The difference between the hue value of the second hue pixel point and the hue value of the first hue pixel point is calculated, and the absolute value of the difference is taken and divided by the hue value of the first hue pixel point to obtain the abnormality degree of the second hue pixel point compared with the first hue pixel point.
[0111] When the abnormality degree is greater than the gradient difference threshold value and the number of second hue pixel points is greater than 3, the first hue pixel point is determined as a difference pixel point.
[0112] The hue value of the second hue pixel point corresponding to the maximum abnormality degree is determined as the hue value of the difference pixel point, and the difference pixel point is re-determined as the first hue pixel point until the abnormality degree of the second hue pixel point compared with the first hue pixel point is less than the gradient difference threshold value, or the abnormality degree is greater than the gradient difference threshold value and the number of second hue pixel points is less than 4.
[0113] Each pixel point in the hue analysis image is taken as the first hue pixel point to obtain the converted image.
[0114] Exemplarily, if the coating is uniform, the hue values of different pixel points on the uniformity analysis image are close. For each rotation angle, the corresponding uniformity analysis image is converted to the HSI space, the data of the H channel, i.e. the hue channel, is extracted, and thus the hue analysis image under each rotation angle is obtained.
[0115] In view of the problem that direct calculation of the hue variance is difficult to effectively identify subtle differences, the hue difference is amplified and a converted image is generated, and the quantitative evaluation of the coating uniformity is realized by combining the difference significance and the conversion difficulty.
[0116] For the hue analysis image of each curvature region, the greater the difference degree between the hue values of the pixel points in the converted image and the smaller the conversion difficulty, the greater the non-uniformity factor in the original image itself, i.e. the greater the non-uniformity of the hyperbolic plate coating corresponding to the original image.
[0117] For each color tone analysis image, a corresponding gradient image is obtained, each pixel point in the gradient image corresponds to a pixel value representing the gradient value of the pixel point, and pixel points in the gradient image greater than a preset gradient threshold value are determined as gradient pixel points.
[0118] In a specific embodiment, the preset gradient threshold value can be obtained in the following manner: gradient values are sorted in descending order of gradient value, and the smallest gradient value in the top 5% of gradient values is determined as the preset gradient threshold value. The determination manner of the preset gradient threshold value can also be set according to actual conditions or determined according to historical data, which is not specifically limited herein.
[0119] The smallest gradient value in the gradient pixel points is denoted as a gradient threshold value, the absolute value of the difference between the gradient threshold value and the gradient mode in the gradient image is calculated, and the ratio of the absolute value of the difference to the maximum gradient value is taken as a gradient difference threshold value y.
[0120] For any one pixel point in each color tone analysis image, the ratio of the absolute value of the difference between the color tone value of each pixel point in the eight-neighborhood of the pixel point and the color tone value of the pixel point to the color tone value of the pixel point is calculated, and the ratio is denoted as the abnormality degree of the neighborhood pixel point compared to the pixel point. For any one pixel point in each color tone analysis image, if the color tone values of more than half of the pixel points in the eight-neighborhood of the pixel point differ from the color tone value of the pixel point by more than y, the pixel point is denoted as a difference pixel point, and the color tone value of the difference pixel point is converted to the color tone value of the pixel point with the maximum abnormality degree in the neighborhood of the pixel point. After multiple conversions, if there is no pixel point in the neighborhood that meets the condition, the conversion is stopped. After all the pixel points are converted, a converted image is obtained. Any one pixel point in the color tone analysis image can be a first color tone pixel point, and any one pixel point in the eight-neighborhood of the first color tone pixel point can be a second color tone pixel point.
[0121] S107, calculating the uniformity of the uniformity analysis image according to the converted image, and determining the uniformity quality of the hyperbolic plate according to the uniformity of the uniformity analysis image.
[0122] In this embodiment, the uniformity of the uniformity analysis image is calculated according to the converted image, specifically including:
[0123] The color tone variance of all pixel points in the converted image is calculated to obtain a difference degree.
[0124] The converted image is subtracted from the color tone analysis image to obtain a color tone layer.
[0125] The pixel points in the color tone layer with color tone values not equal to 0 at the corresponding positions of the pixel points in the converted image are determined as converted pixel points, and the region formed by the converted pixel points is determined as a converted connected domain.
[0126] obtaining a sum of the conversion times of the pixel points in the conversion connected domain;
[0127] normalizing the sum of the conversion times to obtain a conversion frequency;
[0128] multiplying the conversion frequency and the distinctness to determine the uniformity of the uniformity analysis image, wherein the distinctness is a negative of the difference and is obtained as an exponential of a natural base.
[0129] determining the uniformity quality of the hyperbolic plate according to the uniformity of the uniformity analysis image, specifically comprising:
[0130] determining the uniformity of the uniformity analysis image as the uniformity of the corresponding image in the curvature region image set to obtain the uniformity of each image in the curvature region image set;
[0131] determining the coating uniformity of each curvature region in the hyperbolic plate image according to the uniformity of each image in the curvature region image set;
[0132] determining the overall uniformity of the hyperbolic plate according to the coating uniformity of each curvature region in the hyperbolic plate image;
[0133] when the overall uniformity is less than a preset first uniformity threshold, it is determined that the uniformity quality of the hyperbolic plate has a large defect; when the overall uniformity is greater than or equal to the preset first uniformity threshold, and the coating uniformity of the target curvature region is less than a preset second uniformity threshold, it is determined that the uniformity quality of the target curvature region of the hyperbolic plate has a large defect; when the overall uniformity is greater than or equal to the preset first uniformity threshold, and the coating uniformity is all greater than or equal to the preset second uniformity threshold, it is determined that the uniformity quality of the hyperbolic plate has no defect.
[0134] Exemplarily, the hue variance of all pixel points in the conversion image is calculated as the difference degree c of the pixel values on the conversion image, and the smaller the value, the better the uniformity of the corresponding coating on the original image.
[0135] subtracting the hue analysis image from the conversion image to obtain a hue layer; determining the pixel points corresponding to the positions of the pixel points with non-zero hue values in the conversion image in the hue layer as the conversion pixel points, and determining the region formed by the conversion pixel points as the conversion connected domain, or the following steps can be implemented: recording the pixel points on the conversion image that obtain new hue values through conversion as the conversion pixel points, and recording the connected domain surrounded by the conversion pixel points as the conversion connected domain.
[0136] For each transformed connected domain, the number of transformations of each pixel point in the connected domain is obtained. For example, the eight-neighbor pixel points of a certain pixel point do not meet the transformation condition, and the eight-neighbor pixel points of the certain pixel point meet the condition after a certain pixel point is transformed for multiple times, so that the certain pixel point becomes a transformed pixel point. For each transformed pixel point, the more the number of transformations, the more difficult the transformation, and the better the uniformity of the original image.
[0137] The number of transformations of each transformed pixel point is counted, and then the sum of all the numbers of transformations is obtained. Then, normalization is performed to obtain the transformation frequency g. For each transformed image, the result of the transformation frequency g is taken as the uniformity of the corresponding original image, wherein the formula for calculating the discrimination is The uniformity of the hyperbolic plate region corresponding to each rotated image of each curvature region can be obtained by calculation.
[0138] For each curvature region, the uniformity of the hyperbolic plate region corresponding to each rotated image can be obtained, and the mean value of all the uniformities is taken as the coating uniformity of the curvature region. For all the curvature regions, the mean value of the coating uniformities of all the curvature regions is calculated as the overall uniformity of the hyperbolic plate.
[0139] The coating uniformity of different curvature regions, i.e., local regions, and the overall coating uniformity of the hyperbolic plate are obtained through the above process. Then, the respective uniformity thresholds can be used for judgment:
[0140] If the overall uniformity meets the preset first uniformity threshold, then it is determined whether the uniformity of each curvature region meets the preset second uniformity threshold. If the uniformity of each curvature region does not meet the preset second uniformity threshold, different curvature regions are resprayed or subjected to other operations, so that the uniformity meets the condition. If the overall uniformity meets the preset first uniformity threshold and each curvature region also meets the preset second uniformity threshold, it is proved that the hyperbolic plate meets the factory delivery condition.
[0141] Alternatively, the preset first uniformity threshold and the preset second uniformity threshold can be set and adjusted in real time according to actual conditions. In this case, the preset first uniformity threshold and the preset second uniformity threshold are not limited in value. In a preferred embodiment, the preset first uniformity threshold and the preset second uniformity threshold can be 0.7 and 0.8, respectively.
[0142] Thus, the present application is completed.
[0143] To sum up, in the embodiment of the present application, the image obtained by the hyperbolic plate through the curvature is partitioned, and then the image with the least light influence of each partition is obtained, and the transformed image with difference amplification is obtained by transformation on the image, and then the coating uniformity evaluation value of each image is obtained according to the difference degree and the transformation difficulty of the transformed image, thereby improving the accuracy of the uniformity detection of the soil layer of the hyperbolic plate.
[0144] The present application also provides a hyperbolic plate coating uniformity evaluation system based on machine vision, which is shown in Figure 3 The present application also provides a hyperbolic plate coating uniformity evaluation system based on machine vision, which is shown in
[0145] The data acquisition module 101 is used to acquire the hyperbolic plate image, wherein the hyperbolic plate image is an RGB image.
[0146] The data processing module 102 is used to acquire the curvature region image set according to the hyperbolic plate image, wherein the curvature region image set includes images corresponding to different curvature regions; determine the perspective image set according to the curvature region image set, and transform the perspective image set into HIS image and extract the I channel to obtain the brightness image set; calculate the image quality of each image in the brightness image set, and determine the target perspective image in each brightness image set, wherein the target perspective image is the image with the largest image quality in each perspective image set; extract the image corresponding to each target perspective image from the perspective image set to obtain the uniformity analysis image, transform the uniformity analysis image into HIS image and extract the H channel to obtain the tone analysis image; calculate the gradient difference threshold of the tone analysis image, and transform the tone analysis image according to the gradient difference threshold to obtain the transformed image.
[0147] The quality evaluation module 103 is used to calculate the uniformity of the uniformity analysis image according to the transformed image, and determine the uniformity quality of the hyperbolic plate according to the uniformity of the uniformity analysis image.
[0148] It should be noted that: the system provided by the above embodiment is only exemplified by the division of the above functional modules, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above described functions. In addition, the hyperbolic plate coating uniformity evaluation system based on machine vision and the hyperbolic plate coating uniformity evaluation method provided by the above embodiment belong to the same concept, and the specific implementation process is shown in the method embodiment, which will not be repeated here.
[0149] It is to be noted that the above-mentioned embodiments illustrate rather than limit the application, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the application. Such alternative embodiments are therefore considered to be within the scope of the application. In the drawings:
[0150] The various embodiments in the specification are described in progressive manner, and the same or similar parts between the embodiments can be mutually referred to, and each embodiment mainly explains the difference from other embodiments.
[0151] The above descriptions are merely preferred embodiments of the present application, and are not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.
Claims
1. A method for evaluating hyperbolic plate coating uniformity based on machine vision, characterized by, The method comprises the following steps: obtaining a hyperbolic plate image, wherein the hyperbolic plate image is an RGB image; obtaining a curvature region image set from the hyperbolic plate image, wherein the curvature region image set comprises images corresponding to different curvature regions; determining a view angle image set from the curvature region image set, converting the view angle image set into an HIS image and extracting an I channel to obtain a luminance image set; calculating the image quality of each image in the luminance image set and determining a target view angle image in each luminance image set, wherein the target view angle image is the image with the maximum image quality in each view angle image set; extracting an image corresponding to each target view angle image from the view angle image set to obtain a uniformity analysis image, converting the uniformity analysis image into an HIS image and extracting an H channel to obtain a hue analysis image; calculating a gradient difference threshold value of the hue analysis image, and converting the hue analysis image according to the gradient difference threshold value to obtain a converted image; calculating the uniformity of the uniformity analysis image according to the converted image, and determining the uniformity quality of the hyperbolic plate according to the uniformity of the uniformity analysis image.
2. The method for evaluating the uniformity of hyperbolic plate coating based on machine vision according to claim 1, characterized in that, The method of obtaining the curvature region image set from the hyperbolic plate image comprises the following steps: determining a triangular network three-dimensional model from the hyperbolic plate image; calculating the curvature of each position point in the triangular network three-dimensional model based on the normal vector change rate; sorting the curvature of each position point to obtain a curvature sequence; segmenting the curvature sequence using an otsu multi-threshold segmentation method to obtain a curvature segmentation section; merging a first position point and a second position point in the triangular network three-dimensional model to obtain a curvature region, wherein the first position point and the second position point are any position points in the triangular network three-dimensional model, the first position point and the second position point are adjacent and in the same curvature segmentation section; cropping an image corresponding to each curvature region from the hyperbolic plate image to obtain the curvature region image set.
3. The method for evaluating the uniformity of hyperbolic plate coating based on machine vision according to claim 1, characterized in that, The method of determining the view angle image set from the curvature region image set comprises the following steps: inputting each image of each curvature region in the curvature region image set into a trained neural radiance field model to output a multi-view image corresponding to each image through the trained neural radiance field model, wherein the trained neural radiance field model is trained from the hyperbolic plate image; determining each image of each curvature region in the curvature region image set and the corresponding multi-view image as the view angle image set.
4. The hyperbolic panel coating uniformity evaluation method based on machine vision according to claim 1, characterized in that, The method of calculating the image quality of each image in the luminance image set comprises the following steps: detecting each image in the luminance image set using an otsu multi-threshold segmentation method to obtain a segmentation threshold value, determining the ratio of the number of pixel points greater than the segmentation threshold value of the Nth image in the luminance image set to the number of pixel points of the Nth image as the highlight point proportion of the Nth image; calculating the entropy value of the gray level co-occurrence matrix of the Nth image in the luminance image set and normalizing the entropy value to obtain the gray level entropy of the Nth image; multiplying the highlight point proportion of the Nth image and the gray level entropy of the Nth image to obtain the image quality of the Nth image.
5. The method for evaluating the uniformity of hyperbolic plate coating based on machine vision according to claim 1, characterized in that, The method of calculating the gradient difference threshold value of the hue analysis image comprises the following steps: obtaining the gradient of each pixel point in the hue analysis image to obtain a gradient map; Determine the pixel points greater than the preset gradient threshold value in the gradient graph as gradient pixel points; Determine the minimum gradient value in the gradient pixel points as the gradient threshold value; Calculate the difference between the gradient mode in the gradient graph and the gradient threshold value, take the absolute value of the difference, and calculate the ratio of the absolute value to the maximum gradient value in the gradient graph to obtain the gradient difference threshold value.
6. The hyperbolic panel coating uniformity evaluation method based on machine vision according to claim 1, characterized in that, The conversion of the color tone analysis image according to the gradient difference threshold value to obtain a converted image, specifically includes: Determine the first color tone pixel point from the color tone analysis image, and determine the pixel points adjacent to the first color tone pixel point in the eight neighborhood as the second color tone pixel points; Calculate the difference between the color tone value of the second color tone pixel point and the color tone value of the first color tone pixel point, and divide the absolute value of the difference by the color tone value of the first color tone pixel point to obtain the abnormality degree of the second color tone pixel point compared with the first color tone pixel point; When the abnormality degree is greater than the gradient difference threshold value, and the number of second color tone pixel points is greater than 3, the first color tone pixel point is determined as a difference pixel point; The color tone value of the second color tone pixel point corresponding to the maximum abnormality degree is determined as the color tone value of the difference pixel point, and the difference pixel point is re-determined as the first color tone pixel point until the abnormality degree of the second color tone pixel point compared with the first color tone pixel point is less than the gradient difference threshold value, or the abnormality degree is greater than the gradient difference threshold value and the number of second color tone pixel points is less than 4; Each pixel point in the color tone analysis image is taken as the first color tone pixel point to obtain the converted image.
7. The hyperbolic panel coating uniformity evaluation method based on machine vision according to claim 1, characterized in that, The uniformity of the uniformity analysis image is calculated according to the converted image, specifically including: Calculate the color tone variance of all pixel points in the converted image to obtain the difference degree; Subtract the color tone analysis image from the converted image to obtain a color tone layer; Determine the pixel points in the converted image corresponding to the pixel points with color tone values not equal to 0 in the color tone layer as converted pixel points, and determine the region composed of the converted pixel points as a converted connected domain; Obtain the sum of the conversion frequencies of the pixel points in the converted connected domain; Normalize the sum of the conversion frequencies to obtain the conversion frequency; The product of the conversion frequency and the discriminant is determined as the uniformity of the uniformity analysis image, wherein the discriminant is the negative of the difference degree and is obtained as the exponential of the natural base.
8. The hyperbolic panel coating uniformity evaluation method based on machine vision according to claim 1, characterized in that, The uniformity quality of the hyperbolic plate is determined according to the uniformity of the uniformity analysis image, specifically including: Determine the uniformity of the uniformity analysis image as the uniformity of the corresponding image in the curvature region image set to obtain the uniformity of each image in the curvature region image set; Determine the coating uniformity of each curvature region in the hyperbolic plate image according to the uniformity of each image in the curvature region image set; Determine the overall uniformity of the hyperbolic plate according to the coating uniformity of each curvature region in the hyperbolic plate image; When the overall uniformity is less than the preset first uniformity threshold, it is determined that the uniformity quality of the hyperbolic plate has a large defect; when the overall uniformity is greater than or equal to the preset first uniformity threshold, and the coating uniformity of the target curvature region is less than the preset second uniformity threshold, it is determined that the uniformity quality of the target curvature region of the hyperbolic plate has a large defect, and when the overall uniformity is greater than or equal to the preset first uniformity threshold, and the coating uniformity is all greater than or equal to the preset second uniformity threshold, it is determined that the uniformity quality of the hyperbolic plate has no defect.
9. The hyperbolic panel coating uniformity evaluation method based on machine vision according to claim 2, characterized in that, The method comprises the following steps: The hyperbolic plate image and the corresponding view angle parameter of each hyperbolic plate image are input into a neural radiation field model for training, so as to obtain a trained neural radiation field model; The three-dimensional density field data is constructed by using the trained neural radiation field model, and the three-dimensional density field data is input into a moving cube algorithm model, so as to output the triangular network three-dimensional model by using the moving cube algorithm model.
10. A hyperbolic panel coating uniformity evaluation system based on machine vision, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The computer program is executed by the processor to realize the steps of the hyperbolic plate coating uniformity evaluation method based on machine vision according to any one of claims 1-9.
Citation Information
Patent Citations
Judgment method for particle uniformity of steam foaming foam plastic
CN120182613A
Automobile leather defect detection method and system based on visual detection
CN120655628A