Defect Detection In LCD Screen Based On Machine Vision

At present,in the LCD panel industry,most manufacturers still adopt the traditional manual vision inspection(HVI)method in the detection of LCD screen defects.This method is subject to the subjective factors of the inspectors and the external environment,that is,the lack of LCD display defects.Objectively quantified standards make it difficult to ensure product quality and detection efficiency is extremely low.Therefore,it is imperative to study stable and efficient defect detection methods.With the global attention to intelligent manufacturing and the development of machine vision in recent years,the introduction of machine vision inspection into industrial production processes in order to achieve more accurate and efficient detection has become the mainstream of industrial development in various countries.The LCD defect detection system involved in this article is based on the machine vision technology,rational use of various image processing methods to detect and quantify the display defects of the LCD screen on the assembly line.The test system has been tested and analyzed on the 55-inch LCD TV production line of the plant.The test results obtained can also meet the needs of the manufacturers for LCD screen defect detection.The work of this paper in the detection of LCD screen defects mainly has three aspects:(1)The detection system is mainly used to perform a one-pixel screen defect detection on a 55-inch 4K LCD TV,and it is difficult to find such a high-resolution and affordable industrial product under such a large-size and high-precision demand.The camera collects images on the LCD screen.Therefore,in the actual detection process,six 500-pixel industrial cameras are used for image acquisition,and these images are finally stitched to create a complete LCD image for subsequent images.deal with.In the section of image stitching,this article uses a 4K resolution image with six checkerboards for teaching calibration.Each camera then uses the relationship between the feature points of the captured image and the standard image to obtain six single-pieces.The sex matrix transforms the image taken by the camera into a specified splicing size through perspective transformation using a homography matrix,and then splicing and synthesizing it into a complete liquid crystal television image by using an image mask method.Finally,in order to extract the part of the screen in the image,a region line scan and a least-squares straight line method are used to find the four edges and vertices of the screen,thereby obtaining an image containing a complete LCD screen.(2)Influenced by the production process and physical results of the LCD monitor,when using an HD industrial camera to perform image acquisition on the LCD screen,the image will have regularly arranged horizontal and vertical texture stripes.This striped background information will interfere with people on the one hand.The visual observation on the other hand will have a serious impact on subsequent defect segmentation.Therefore,after careful study and analysis of this texture stripe,this paper proposes a method based on frequency-domain low-pass filtering to suppress the background stripe.After analyzing and comparing several commonly used filters,a frequency-domain Gaussian low-pass filter is used to filter the image of the LCD screen in the frequency domain.The results show that the background information of the filtered image is effectively suppressed.(3)Segmentation and quantification of LCD defects are the core tasks of the system,but it is necessary to do one step of image enhancement before segmentation.This paper proposes an image enhancement method based on DOG algorithm to improve the contrast of defects.In defect segmentation,this paper compares the advantages and disadvantages of commonly used segmentation algorithms for LCD defect segmentation.Then it proposes a dynamic threshold segmentation algorithm that is suitable for such defects,and effectively extracts the defect pixel points to aggregate the shape profile information.Finally,we quantify the defects.According to actual project requirements and industryrelated indicators,this paper specifies several core parameters that characterize defects to objectively quantify defects in LCD screens.