Defect Detection Method Based On Machine Vision Capacitive Screen Non-visible Area Leads

As an important part of human-computer interaction, touch-screen has been widely used due to the rapid growth of mobile devices. Quality testing of Capacitive Screen is becoming increasingly important since Capacitive Screen quality directly affects the competitiveness of products. Traditional Capacitive Screen Defect Detection is implemented manually, but relying on artificial vision testing is becoming difficult to meet the current requirements of efficient industrial production. So, detection of defects in the invisible areas based on Computer Vision is studied in this dissertation.First of all, the Capacitive Screen defect samples and the Defect Detection requirements are analyzed to develop the defect detection process. The defect detection process is made up of Image Preprocessing, Defect Detection, Defect Segmentation, Feature Extraction and Defect Identification. Secondly, both Gradient Inverse Weighting Filter and Median Filter are used in image filtering and then the noise is effectively suppressed. On the basis of the filtering results, the histogram is then enhanced to improve the image. Then, the Bilinear Interpolation is used in the established Projection Transformation Model to solve the problem of Projection Transformation. After that, SIFT algorithm is used to match the image and find out the matching feature points. With the analysis on the differential results. Defect Detection can be implemented. After detection, image segmentation algorithms, such as edge detection algorithms based on Phase Congruency and Gradient, Otsu segmentation and Hough Transform, are studied. As a result, the defect can be extracted from the background by choosing the appropriate segmentation algorithm. At last, feature extraction methods, on the basis of contour tracing and the invariance of image translation, rotation, scaling, are studied. The defects can be classified by improving the KNN Classifier, which includes:using feature weighting methods, clustering of samples and assigning different weights to neighbor samples according to the distance and density.The results indicate that Image Preprocessing algorithms can significantly improve the image quality. The Defect Detection algorithms and Image Segmentation algorithms can be successfully achieved. Defect Identification algorithms based on KNN Classifier is stable and correct which is able to meet the requirements of Capacitive Screen Defect Detection.