DIB: Piled Man-made Object Detection and Pose Estimation in Point Cloud Blocks

Object detection and pose estimation are fundamental modules in robotic applications, many of these objects are man-made like mechanical parts. Although have been researched widely in recent designs, detecting objects from a cluttered pile is still challenging. In this paper, a robust method for the detection and pose estimation of randomly piled man-made objects in blocks of point cloud is presented which increases the detection efficacy significantly. The approach begins with building blocks from the point cloud, each of which contains one object. Then object detection and pose estimation are performed in the blocks using 3D primitive shapes. We evaluate the performance of our approach in comparison with state-of-the-art methods. Experiments show that the proposed system detects objects efficiently and accurately in the presence of noise and occlusion.