Cleaning Robot Vision System Based on RGBD Camera and Deep Learning YOLO-based Object Detection Algorithm

Abstract

Autonomous cleaning for a wide environment is a big challenge due to energy and time consumption. This paper introduces a novel vision strategy for the auto-guided cleaning vehicle to wash dirty areas in the indoor environment. A modified YOLOv3-based deep learning algorithm is utilized to detect trash objects and the location of soiled regions. Furthermore, a dataset with various uncleaned floors is collected to deal with solid, liquid, or reflected trash under the supervised fashion. The experimental results confirmed that the effectiveness of the proposed autonomous cleaning system could handle the detected uncleaned areas with low energy and time consumption compared with other cleaning systems.