An Intelligent Garbage Sorting Robot System Based on Machine Vision and Knowledge Base

Garbage disposal has always been a widely concerned environmental problem. At present, garbage sorting mainly depends on manual work, with low efficiency, poor accuracy and easy to harm the staff. One way to solve the problem is to develop robots with garbage sorting function instead of manual operation. This paper presents a design scheme of intelligent garbage sorting robot system based on machine vision and knowledge base. The coarse localization of garbage is firstly detected by the YOLO-v5 algorithm with the RGB camera and the bounding box of the garbage can be obtained in the image coordinate system. Then, the depth information of the garbage is retrieved by the depth camera, and the garbage coordinates are obtained. After the coordinate conversion, the position of the garbage is set as the robot navigation target and the trajectory planning is carried out. When the robot reaches the target position, the YOLO algorithm is then utilized to refine the classification and the position of the garbage and the robot arm equipped on the robot is utilized to pick up the garbage. After the garbage recognition, the garbage type and the corresponding garbage bin location information can be retrieved according to the knowledge base. At the same time, it stores the garbage information of the current position in the knowledge base to facilitate the subsequent optimization of the cruise track. Finally, the robot navigates to the corresponding bin location, puts down the garbage, and completes the garbage sorting task. Experimental results show that this method can accurately identify garbage and effectively improve the efficiency of garbage sorting.