A Novel Binary Descriptor for 3D Registration of Point Clouds from Low-cost Sensors

For 3D rigid data matching, local feature description is a fundamental research topic. It remains a challenge, however, to balance the local shape descriptor between matching accuracy and efficiency. Towards this end, we propose a novel feature representation of 3D local surfaces called the Binary Weighted Projection-Point Distance (BWPD). A rigorous comparison with state-of-the-art approaches validates the superiority and feasibility of our approach across contexts and 3D sensors.