視覚情報に基づくトマト果実収穫ロボットの行動戦略（第二報）

―赤外線画像と鏡面反射を利用した果実検出技術を中心に―

抄録

This paper presents a novel ripe tomato fruit detection method within a tomato cluster, as a part of behavior strategy of tomato harvesting robot. Proposed method uses infrared images and specular reflection based on the optical characteristic of tomatoes. The fruits of the tomato clusters are detected separately on the captured infrared images, searching the strong response at the center of fruits caused by specular reflection. To recognize the ripe tomatoes from the unripe fruits and other background objects, RGB images were used. The method was evaluated using infrared and RGB images captured by Xbox One Kinect sensor in a greenhouse environment. The result shows the detection accuracy of 88.1 percent of the proposed method.