Development Of Light Shield Defect Detection System Based On Machine Vision

Nowadays,smartphones and tablets are gaining popularity rapidly.Also,the screen production demand is increasing quickly.The light guide plate is an important part of the screen.It's quality directly affects the display effect of the screen.In the manufacturing process of silk screen printing,chemical etching,laser processing and bumper processing of light guide plate,due to factors such as raw material composition,equipment use,processing technology and worker operation,defects will inevitably appear on the surface.Defects include highlights,line scratches,crushes,shadows and other processing defects.At present,the detection of the light guide plate still stays at the stage of manual detection.Not only the labor cost is increased year by year,but also the detection requires strong light irradiation.Working for a long time is extremely harmful to workers' physical and mental health.On the other hand,since the manufacturing precision of the light guide plate is relatively high,the defects of the light guide plate are generally very small.Some lines have a scratch width of less than one pixel,which exceeds the limit of manual naked eye detection.Therefore,this thesis conducts indepth research on the requirements of quality inspection of light guide plate manufacturers and develop a machine vision-based light guide plate defect detection system,including hardware and software.Finally,a large number of experimental tests are carried out to verify the reliability of the algorithm.The main researches are as follows:(1)Analyze the optical characteristics of the light guide plate and design a general scheme of the light guide plate defect detection table based on machine vision,which involves mechanical structure design and vision system.In the mechanical structure design,it is necessary to ensure that the detection process does not cause secondary damage to the light guide plate.The light guide plate is grasped by the suction cup on the robot hand.The suction cup is placed on the loading platform.A linear guide rail is adopted and the loading platform is adopted.When the light guide plate to be tested is moved from the guide rail to the other end at a constant speed,putting the laser sensor under the sensor,the sensor gives a trigger signal to the line camera for image acquisition.In the vision system,according to the optical characteristics,imaging features and defect characteristics of the light guide plate,a machine vision detecting device composed of a 16 K line camera,a lens,a multi-angle light source and the like is determined.(2)Design of automatic partitioning algorithm for light guide plate imageDue to the 16 K line array camera,the line image of one light guide plate is close to 400 MB.The light guide points are unevenly distributed.The light entrance side is sparse and the other side is dense.The detection standards of different areas are different.It will cause serious misdetection problems,if a unified algorithm is used.In this thesis,a density-based image autopartitioning algorithm is proposed.Based on the body part of the line image of the light guide plate,the light guide point is extracted by OTSU threshold segmentation.The light guide point density of different regions is calculated.So,the image is divided into multiple sub-regions.The light guide plate is tested with different algorithms.(3)Design of conventional defect detection algorithm for light guide plateFor the conventional bright spots,crushes,scratches,this thesis designs a two-side Gaussian direction derivative filter to convolve the light guide plate image.Reconstruct the convolution result by using two mean-valued filters of different sizes to remove the interference of the normal light-guiding point and obtain the defect-enhanced image.Furthermore,using the binarization threshold operation to extract the suspected defect connected region to confirm whether it is a defect based on analyzing the regional characteristics.Then It can be classified.Finally,a large number of experiments were performed on the light guide plate by using the collected image at the industrial site.The experimental results show that the detection accuracy for bright spots,crushes and line scratches has a high detection accuracy.The comprehensive accuracy rate reaches 93.4%.(4)Design of detection algorithm for slight line scratch defect of light guide plateBased on the analysis of the cause and imaging characteristics of the slight scratch on the light guide plate,this thesis proposes a method for detecting the slight scratch of the light guide plate.It is based on multi-directional Gabor filtering and sub-pixel analysis.Firstly,in order to highlight the defect area,a multi-directional Gabor filter is designed.Furthermore,the sub-pixel image analysis method is used to accurately segment the suspected defect area from the background image.Finally,the area shape feature is analyzed to accurately extract the slight line scratch defect.The experimental results show that the algorithm has high operating efficiency and high accuracy,which can effectively detect slight line scratches,and the comprehensive accuracy rate reaches 95%.