Research On The Defect Detection Method For Mobile Phone LCD Panel Based On Subtraction Operation And Saliency Fusion

The traditional manual detection of the surface quality of mobile phone LCD panel is difficult to meet the contradictory demands of production efficiency and absolute quality due to the impact of fast production lines,complex products and fuzzy defects.Demands for fast and accurate mobile phone LCD panel defects detection in industrial production lines have made more precise requirements for traditional defect automatic detection algorithms that can easily detect high-contrast dirt as defects,leading to the problem of high defect over-detection rate.In this paper,an automatic defect detection method for mobile phone LCD panel based on subtraction operation and saliency fusion of different domains is proposed.The main works are as follows:(1)In order to overcome the problem of large input image size and susceptibility to interference from the background of non-detection areas,a method of localization of interest based on image gradient features and morphology is used to reduce the computational cost for subsequent defect detection algorithms.This method uses Sobel edge gradient detection and fills the contour with image morphology to quickly divide the image into the area to be detected and the area not to be detected,and uses image product operation to eliminate the background interference of the detection area.(2)For the problem of detecting small defects in the body of the region of interest,an automatic defect detection method based on improved subtraction operation is proposed.This method performs a weighted subtraction operation on the image to be detected and the template image,and subtracts the corresponding bit pixels of the two images to obtain the difference information.In the core problem of template construction,this paper proposes that the image to be detected is directly subjected to high-core median filtering to obtain the template image,thereby omitting the registration of the spatial position of the image to be detected and the template image,reducing the algorithm processing time and avoid The influence of noise caused by external factors such as light changes is discussed.(3)Considering the high defect over-detection rate in the defects detection results of subtraction operation,a fine defect detection method based on the saliency fusion of spatial and frequency domains is proposed.Based on the spatial processing of the image,the algorithm of visual attention mechanism is integrated,and the image is transformed into the frequency domain processing to lighten the dirt of the plane.Then,by multiplying the defects saliency map in the frequency domain and the defect saliency map in the space domain,the regions with high saliency probability are enhanced,and the regions with low saliency probability are weakened.(4)Finally,the algorithm proposed in this paper is verified on the 360 images collected on the industrial site.The experimental results show that the defect detection method proposed in this paper effectively reduces the problem of defect overexamination.The defect over-examination rate is only 12.3%,and the defect detection rate is as high as 95.7%,which can meet the actual detection requirements.