Self-Shadow를 이용한 푸른 사과 낙과 검출 시스템 개발

This paper proposes an algorithm for detecting apples to measure the number of apples on the ground through analyzing the apple of self-shadow. The proposed algorithm explores whether the apple candidate object has a round self-shadow. In order to educe colors appropriate for green apple areas, CIE L*a*b* color space was used to educe colors. In order to educe green apple characteristics strong against lighting change, modified census transform (MCT) was used. Then using AdaBoost Learning algorithm, characteristics data on green apples were learned and generated. And self-shadow is used as the weight to apply the light reflection characteristics of the ball-like shape. With generated data, detection of green apple areas was made. The proposed algorithm had a higher detection rate than existing image processing algorithms and minimized false detection.