Super Edge 4-Points Congruent Sets-Based Point Cloud Global Registration

With the acceleration in three-dimensional (3D) high-frame-rate sensing technologies, dense point clouds collected from multiple standpoints pose a great challenge for accuracy and efficiency of registration. The combination coarse registration fine has been extensively promoted. Unlike requirement small movements between scan pairs registration, can match scans with arbitrary initial poses. state-of-the-art methods, Super 4-Points Congruent Sets algorithm based on Sets, improves speed to linear order via smart indexing. However, lack reduction scale original limits application. Besides, coplanarity bases prevents further search space. This paper proposes novel method called Edge address above problems. proposed follows three-step procedure, including boundary segmentation, overlapping regions extraction, selection. Firstly, an improved vector angle is used segment aiming thin out clouds. Furthermore, extraction executed find contour. Finally, selects conforming distance constraints candidate set without consideration about coplanarity. Experiments various datasets different characteristics have demonstrated that average time complexity by 89.76%, 5 mm than algorithm. More encouragingly, experimental results show be applied restrictive cases, such as few massive noise. Therefore, this faster more robust under guarantee promised quality.