쿼드 트리 분해를 이용한 딥러닝 기반의 소형 카메라 모듈 덮개 결함 분류

Abstract

Compact camera modules are used in many small electronic devices. Defect detection is essential because the cover protecting the module is exposed to the environment outside. It also defects classification to provide feedback to the production process and prevent further defects from occurring. Since the size of the defect is small compared to the size of the module, we adopted the method of extracting the inspection patch. To reduce the inspection time, we employed the compact camera module defect classification system using quadtree decomposition. The module area was detected using template matching, and the unnecessary background was deleted through bit operation. By binarizing the detected module areas, defect candidates were found through quadtree decomposition. The defects were then classified through deep learning. Compared to the previously proposed patch-based method, the inspection time could be reduced.