A method for testing the grinding quality of tracks used in LED taping machines

Abstract

This invention relates to the field of image data processing technology, specifically to a method for detecting the grinding quality of LED tape and reel machine tracks. The method includes: acquiring images of the LED tape and reel machine track to be inspected, classifying the images using an improved support vector machine model to obtain corresponding labels, and judging the grinding quality based on the labels. The improved support vector machine model introduces class weights, which are calculated based on the content richness of each image to be inspected. Content richness is measured by the difference between the image's grayscale gradient and color channel information. Specifically, the difference between the image's grayscale gradient variance and the color channel grayscale mean is used to calculate the content richness of each image, thereby affecting its class weight. This invention solves the problem of low accuracy in detecting the grinding quality of LED tape and reel machine tracks.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology. More specifically, this invention relates to a method for inspecting the grinding quality of tracks used in LED tape and reel machines. Background Technology

[0002] With the rapid development of the LED industry, LED taping machines are playing an increasingly important role in the production process. As a key component, the LED taping machine track is responsible for transporting and securing the LED strips; therefore, the processing quality of the track directly affects the efficiency of the entire production line and the product qualification rate. To ensure high-precision processing of the LED taping machine track, surface quality inspection has become a crucial step. However, during the track grinding process, due to equipment precision limitations, differences in worker operation, and variations in materials, uneven wear, scratches, or other defects may appear on the track surface. These defects are often difficult to identify efficiently and accurately using traditional manual inspection methods.

[0003] Therefore, automated and precise inspection of the grinding quality of LED tape-and-reel machine tracks has become a pressing technical problem in current production processes. By employing advanced image processing and machine vision technologies, various defects on the track surface can be quickly and accurately identified, thereby improving production efficiency and product quality, and reducing the occurrence of defective products. Support Vector Machines (SVMs) are a supervised learning method widely used in image classification, pattern recognition, and other fields. In image binary classification tasks, SVMs distinguish between two classes of samples by finding the optimal hyperplane. Their classification performance depends on factors such as the feature representation of the data, the choice of kernel function, and class imbalance. The introduction of class weights is an important means of solving the class imbalance problem, allowing for optimization of the classifier's performance during training by adjusting the misclassification costs of different classes.

[0004] Existing technologies can effectively alleviate class imbalance by adjusting class weights, but this method has some drawbacks. Choosing appropriate class weights often depends on human experience or parameter tuning, and different weight settings may lead to overfitting or underfitting, which may result in low accuracy of classification results in practical applications, and consequently, low accuracy in the grinding quality inspection of LED tape and reel machine tracks. Summary of the Invention

[0005] To address the problems of low accuracy and insufficient reliability of detection results mentioned in the background art, the present invention provides the following solution.

[0006] This invention provides a method for detecting the grinding quality of LED taping machine tracks, comprising: acquiring images of each LED taping machine track to be inspected; classifying the images to be inspected using an improved support vector machine model to obtain labels; and judging the grinding quality of the LED taping machine tracks based on the labels; wherein, the improved support vector machine model includes class weights, the first... The class weights of the image to be detected and the first The content richness of the images to be detected is positively correlated with the sum of the content richness of all images to be detected; the content richness is related to the content richness of the first image. All pixels in the image to be detected The difference between the grayscale value and the grayscale mean in a channel, and the variance of the grayscale gradient are positively correlated with the first channel. All pixels in the image to be detected The difference between the grayscale value and the mean grayscale value in a channel is inversely correlated.

[0007] The above technical solution can intelligently adjust the attention to different images during the classification process, enhance the sensitivity to highly complex images and subtle defects, and reduce interference to low-information areas. Through the improved support vector machine model and combined with the dynamic weight allocation of image content richness, the detection accuracy of the grinding quality of LED tape and reel machine tracks is improved.

[0008] Furthermore, the first Class weights of the image to be detected for, In the formula, For the first The content richness of the image to be detected. The total number of all images to be detected. This is an empirical constant.

[0009] Furthermore, the first Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected.

[0010] The aforementioned technical solution significantly enhances the ability to quantify image complexity by introducing a content richness calculation method based on the variance of gray-level gradients and the difference in gray-level mean values ​​between color channels. By combining gray-level gradients and gray-level differences between color channels, it can more precisely measure detail changes and potential defects in images, especially when processing images with complex reflective properties such as metal surfaces, effectively capturing subtle surface variations. This not only enhances the response to highly complex images but also effectively improves the sensitivity to minute defects in images, helping support vector machine models to more accurately adjust weights during classification, thereby improving the accuracy of judging polishing quality.

[0011] Furthermore, the first Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected. For the natural constant An exponential function with base 0. For the natural constant A logarithmic function with base 0.

[0012] The aforementioned technical solution further optimizes the method for measuring image content richness by introducing logarithmic and exponential operations, combined with the gray-level gradient variance and the gray-level differences between color channels. Through precise quantification of image details, it can more effectively capture subtle changes in images, especially in complex images such as defects or surface irregularities on metal surfaces. This computational method enables more sensitive identification of highly complex areas and potential quality problems, thereby improving the accuracy and robustness of defect detection. Particularly when dealing with subtle differences in images, nonlinear operations enhance sensitivity to details, avoiding interference from low-information areas. This allows the model to better cope with diverse detection scenarios in practical applications, improving the quality control capabilities of production lines and ensuring more stable and accurate detection results.

[0013] Furthermore, the labels are divided into damaged and non-damaged types.

[0014] Furthermore, an alert is issued when the label is deemed defective.

[0015] Furthermore, CCD or CMOS cameras are used to acquire images of the LED tape and reel tracks to be inspected.

[0016] Furthermore, it also includes performing grayscale processing on the image to be detected.

[0017] The aforementioned technical solution, by introducing a grayscale processing step, effectively simplifies image complexity, removes interference from color information, and allows the model to focus more on the image's structural and textural features. Converting the image to grayscale reduces computational load, increases processing speed, and enhances sensitivity to grayscale changes and edge details without losing crucial information. This is particularly important for defect detection, as many defects in grayscale images typically manifest as abrupt changes or shifts in grayscale values. Grayscale images can more clearly highlight these subtle differences, thereby improving detection accuracy and robustness. This is especially true when processing metal tracks with high reflectivity or complex surface textures, enabling more precise identification of potential problems and optimizing overall quality control and detection efficiency.

[0018] Furthermore, the warning is issued via a buzzer.

[0019] The aforementioned technical solution utilizes a buzzer for early warning, providing immediate and intuitive feedback to ensure operators can detect grinding quality abnormalities or defects in the shortest possible time. This audible alarm effectively improves production line response speed, preventing product quality issues or production delays caused by undetected defects. Furthermore, the buzzer's high audibility in noisy production environments ensures operators can clearly receive warning information even in complex or busy work environments, thereby reducing the risk of human error and improving overall production efficiency and safety.

[0020] Furthermore, the CCD camera or CMOS camera is mounted above each LED tape and reel track.

[0021] The beneficial effects of this invention are as follows:

[0022] This invention effectively improves the detection accuracy of track grinding quality in LED taping machines by combining an improved support vector machine model and content richness analysis. By classifying images and assigning different category weights, it can more accurately distinguish tracks of different quality, especially issuing timely warnings when damage is detected, preventing production defects caused by quality issues. Furthermore, image acquisition is performed using a CCD or CMOS camera, and grayscale processing enhances the stability and consistency of image processing. This method enables real-time monitoring of track grinding quality during production, ensuring production efficiency and product quality. Simultaneously, a buzzer alert facilitates rapid response and improves safety in the working environment, reducing the risk of human error. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for inspecting the grinding quality of an LED taping machine track according to an embodiment of the present invention. Detailed Implementation

[0024] An embodiment of a grinding quality inspection method for LED taping machine tracks.

[0025] like Figure 1 The flowchart shown is a method for grinding quality inspection of LED taping machine tracks according to an embodiment of the present invention, including the following steps:

[0026] S1: Acquire the images to be inspected on each LED tape and reel machine track.

[0027] In one embodiment, to achieve efficient detection and monitoring of the LED taping machine track, a high-resolution CCD or CMOS camera can be used for image acquisition. These cameras are precisely mounted above the LED taping machine track to ensure that the entire image to be inspected is captured from the optimal angle. The advantage of using a CCD or CMOS camera lies in its superior image clarity and imaging quality. CCD and CMOS sensors have high photoelectric conversion efficiency, enabling them to capture extremely fine image details, making them particularly suitable for use in complex industrial environments. This high-quality image can effectively avoid false detections and missed detections caused by image blurring or excessive noise.

[0028] The acquired images to be detected are then converted to grayscale. Grayscale conversion, a fundamental operation in image preprocessing, eliminates redundant information in color images, thereby reducing computational complexity and significantly improving processing efficiency, especially in large-scale image data processing. Converting color images to grayscale not only reduces data storage requirements but also simplifies and improves the accuracy of subsequent image feature extraction, edge detection, and defect analysis.

[0029] S2: The image to be detected is classified using an improved support vector machine model to obtain a label.

[0030] In one embodiment, the improved support vector machine model includes class weights, the first... Class weights of the image to be detected for, In the formula, For the first The content richness of the image to be detected. The total number of images to be detected is represented by a single class weight. By introducing class weights to improve the support vector machine model, the weights of each image can be dynamically adjusted based on its content richness, allowing the model to focus more on images with higher information content during training. This method effectively improves the accuracy and robustness of the model when processing complex image data, especially in image sets with significant differences or noise. By assigning higher weights to images with higher content richness, the impact of low-quality data on model performance is reduced, thereby enhancing the overall classification effect and improving the accuracy and reliability of detection.

[0031] No. Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected.

[0032] It's important to note that orbitals are typically made of metal, and metal surfaces have unique light reflection characteristics. In the visible spectrum, red light (R channel) has a longer wavelength than blue light (B channel), making it more easily reflected by metal surfaces. When defects exist on the orbital surface, such as cracks, scratches, or pits, these defects alter the reflectivity of the metal surface, causing a difference in red light reflection at the defect location compared to the normal area. This makes the defects more noticeable in the R channel. Furthermore, the morphological characteristics of the defects themselves, such as cracks and scratches, often cause minute deformations or damage to the metal surface. These deformations affect the direction of light scattering and reflection. Because red light has a longer wavelength, it is more susceptible to these minute changes, resulting in a more pronounced contrast in the R channel. In contrast, blue light has a shorter wavelength and is less sensitive to minute deformations or damage, so defects are less noticeable in the B channel. Therefore, when there is a significant difference between the R and B channels in an image, it can be inferred that there is a high probability of defects in the image.

[0033] Furthermore, the presence of defects increases the number of edge pixels in an image, and edge pixels typically have larger grayscale gradients. Therefore, the larger the gradient variance of pixels in an image, the more edge pixels there are, the richer the edge information of the image, and the more complex the overall content. This characteristic further indicates that by calculating the variance of the grayscale gradient in an image, the richness of the image content can be effectively reflected, and it can help identify and locate defective regions in the image.

[0034] By introducing the difference between the image gray-level gradient variance and the mean gray-level of the color channels, the content richness of each image to be detected is calculated, thereby quantifying the complexity of the image. This approach allows for a more accurate measurement of the spatial and color information distribution differences within the image, thus assigning greater weight to images with higher content richness. This treatment not only improves the model's attention to highly complex images and avoids excessive influence on low-information images, but also effectively enhances the accuracy of image classification and anomaly detection, especially in complex scenes, contributing to the optimization of the model's robustness and adaptability.

[0035] In another embodiment, the first Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected. For the natural constant An exponential function with base 0. For the natural constant A logarithmic function with base 0.

[0036] By combining the gray-level gradient variance and the exponential function of the gray-level value differences in color channels, the richness of image content is further measured more accurately. Introducing logarithmic and exponential operations effectively balances gradient changes and color channel reflectance differences, giving higher weight to regions with significant variations in the image during calculation. This approach not only improves sensitivity to subtle defects but also adjusts the response to different types of defects through nonlinear functions, making the model more stable and accurate when dealing with complex image data. This enhances the accuracy and robustness of defect detection and image classification, especially when handling complex or noisy images, effectively avoiding false positives and false negatives.

[0037] The labels are divided into damaged and non-damaging types;

[0038] S3: Determine the grinding quality of the LED taping machine track based on the label.

[0039] In one embodiment, when a damaged grinding quality label on the LED taping machine track is detected, an immediate warning will be issued to ensure that necessary adjustments are made promptly during production. This warning is delivered via a buzzer, quickly and effectively attracting the operator's attention and ensuring the problem is addressed in the shortest possible time. As an intuitive and easily identifiable alarm method, the buzzer not only rapidly transmits information in noisy industrial environments but also indicates the urgency of the problem through changes in audio frequency, thus helping operators make appropriate responses.

[0040] Timely early warnings can significantly reduce the risk of production failures and substandard product quality caused by grinding quality issues, preventing greater losses due to undetected defects. Furthermore, this early warning mechanism enhances the intelligence level of the production line, enabling equipment to respond automatically to problems, thereby reducing reliance on manual inspection and improving reliability.

[0041] This invention utilizes an improved support vector machine model combined with image content richness assessment to accurately detect the grinding quality of LED tape and reel machine tracks. By classifying and discriminating the acquired images, defects or undesirable phenomena during the grinding process can be precisely identified, improving the accuracy and efficiency of detection. Furthermore, by introducing grayscale value difference and gradient variance analysis into grayscale processing and image classification, the sensitivity to different types of defects is further enhanced, enabling the detection process to effectively distinguish between normal and undesirable areas. In addition, combined with an early warning mechanism, timely alerts can be issued when quality problems are detected, improving the production line's response speed and operators' real-time monitoring capabilities, thereby effectively avoiding potential production quality risks and ensuring the stability of the grinding process and product consistency.

[0042] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0043] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for inspecting the grinding quality of tracks used in LED taping machines, characterized in that, include: Images of the LED taping machine tracks to be inspected are collected; the images are classified using an improved support vector machine model to obtain labels, and the polishing quality of the LED taping machine tracks is judged based on the labels. The improved support vector machine model includes class weights, the first... The class weights of the image to be detected and the first The content richness of the images to be detected is positively correlated with the sum of the content richness of all images to be detected; No. Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected. For the natural constant An exponential function with base 0. For the natural constant A logarithmic function with base 0; The content richness and the first All pixels in the image to be detected The difference between the grayscale value and the grayscale mean in a channel, and the variance of the grayscale gradient are positively correlated with the first channel. All pixels in the image to be detected The difference between the grayscale value and the mean grayscale value in a channel is inversely correlated.

2. The method for testing the grinding quality of the track used in LED taping machine according to claim 1, characterized in that, No. Class weights of the image to be detected for, In the formula, For the first The content richness of the image to be detected. The total number of all images to be detected. This is an empirical constant.

3. The method for testing the grinding quality of the track used in LED taping machine according to claim 1, characterized in that, No. Content richness of the image to be detected for, In the formula, For the first The variance of the gray-level gradient of the image to be detected. For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The first in the image to be detected Pixels in Grayscale value under the channel, For the first All pixels in the image to be detected Mean gray level under the channel For the first The total number of pixels in the image to be detected.

4. The method for testing the grinding quality of an LED taping machine track according to claim 1, characterized in that, The labels are divided into damaged and non-damaged types.

5. The method for testing the grinding quality of an LED taping machine track according to claim 4, characterized in that, If the label is defective, an alert will be issued.

6. The method for testing the grinding quality of an LED taping machine track according to claim 1, characterized in that, Images of the LED tape and reel tracks to be inspected are acquired using a CCD or CMOS camera.

7. The method for testing the grinding quality of an LED taping machine track according to claim 1, characterized in that, It also includes performing grayscale processing on the image to be detected.

8. A method for detecting the grinding quality of an LED taping machine track according to claim 5, characterized in that, The warning is issued via a buzzer.

9. A method for detecting the grinding quality of an LED taping machine track according to claim 6, characterized in that, The CCD or CMOS camera is mounted above each LED tape and reel track.

Images