Efficient & safety oriented environment perception of medical supplies delivery robots

With the fundamental purpose of helping medical workers reduce their workload and simplify their workflow, the medical supplies delivery robot was invented by people to help to share low technical work. In this paper, a recognition method using point clouds is proposed that provides the robot to identify obstacles during the delivery process. To reduce the time cost, this study segments the region of interest into a series of cubes arranged tightly, each cube being a spatial domain with specific edge length containing specific attributes. The cube is used as a processing unit and point clouds intensity, height, entropy, eigenvalues of the covariance matrix, the normal vector are used as the features to train the SVM classifier model for environment perception. The results show that the method can accomplish the goal of environment perception using only point clouds, with significant recognition ability and short time used.