A Unique Method for Detecting Grounds in the Indoor Environment

In this paper, a unique method for detecting and distinguishing the grounds, walls, and ceilings in the indoor environment using single depth image is presented. Our method combines clustering and recognition to estimate the planes show in the depth image. A depth image correction method is proposed. Clustering process using k-means algorithm. The number of clusters is determined by elbow method. Recognition process using HOD (histogram of distance) to classify the planes. The classification of grounds walls and ceilings is very useful in mobile robot navigation, people detection and many other situations. Different from other existing methods, our method is fully automated, and only need one single depth image, no need to know the position and posture of the camera. On the basis of a certain number of experiments, we demonstrate the usefulness and robustness of our method. Our method is ready to be used in many other situations like path planning, target following, map reconstructing, etc.