| Blisters and scratches are the main influencing factor to the quality of the glass-shell. In this article, some detecting methods for the defect of glass-shells by computer vision are studied.Whether the image of the glass-shell is clear or not is very important for the computer detection. Based on the reflection theory, the computer can get the defects in the images with high quality.Four detection methods of defect are presented in this article, which is respectively based on the characters of the image gray level, the gray histogram, the edge of the defect, and the curve fitting. This article compares these methods and presents the advantages and disadvantages of each method. It is known that if there is defection on the glass-shell, it must be identified exactly. In order to extract the defect, some image segmentation methods are studied, such as differential histogram method, Otsu method, and histogram method. Comparing with other segmentation methods, differential histogram method is more practical to get defect information in the glass-shells. After the defects are extracted, the glass-shells are classified based on the shape of defects by pattern recognition methods.Based on the requirement of the detection, a detecting system with illuminating instruments is designed. The software for detecting the defect is developed by Visual C++6.0, which is very friendly to the users.In the end, this article studies the theory of phase measurement method. Based on analyzing the structure of the system, a fast calibration method is presented to get the relationship matrix between the height and the phase. This article also studies how to measure the shape of the glass shell by the profilometry based on the projecting grating method. |
