참외 비파괴 품질 판정을 위한 3차원 영상 선별 시스템 개발

초록

To meet the increasing consumer demand for high-quality produce, fruit sorting have become essential processes in the distribution chain. External quality factors such as color, shape, and surface defects are key criteria for fruits, typically assessed by labor or 2D imaging systems. In the case of oriental melon, additional inspection is required to detect internal defects such as fermentation. Currently, agricultural products processing center rely on labor to identify fermented fruits by observing whether the fruit sinks or floats in water, and deformed fruits are also sorted manually due to the limitations of 2D imaging. This study proposes an automated 3D vision-based sorting system and evaluates its performance. The system consists of an input section, imaging system, discharge section, return conveyor, and a main controller. Five RGB-D cameras are used, which is then processed using calibration and image processing algorithms to reconstruct 3D models. Based on the 3D data, the system calculates density and volume ratio to identify fermented and deformed ones. Experimental results showed a classification accuracy of 98.88% for fermented and pre-fermented melons and 95.83% for deformed melons. These findings indicate that the proposed 3D vision sorting system is a viable solution for siorting process automation.

키워드