A Robust Recognition Method for Occlusion of Mini Tomatoes Based on Hue Information and the Curvature

Abstract

In this paper, we propose a robust recognition method for occlusion of mini tomatoes based on hue information and the curvature. This method is used for a managing system using robots for hydroponic that we have proposed. In this system, robots need to recognize mini tomatoes to manage farmlands. In a lot of cases, mini tomatoes are covered partially by leaves or other tomatoes. Thence, the system needs a mini tomato recognition method in the situations including occlusion. First, this method detects red areas using hue information from a source image. Second, the method detects contours from the areas by using contour tracking. Finally, the method judges whether contours are mini tomatoes or not by using the curvature. We compared our method with circle detection method using Hough transform. Experimental results showed that the recognition rate of our method was 78.8%. On the other hand, the recognition rate of the comparative method was 47.9%. Therefore, we consider that the proposed method is appropriate for mini tomato recognition in the situations including occlusion.