无序多视角点云的自主配准方法

A feature match based approach is proposed to improve the performance of multi-view registration. This approach can achieve global registration of unordered range scans. It calculates and matches feature descriptors to achieve pair-wise registrations. In addition, an effective rule to judge the reliability of pair-wise registration results is designed. Moreover, a model augmentation method is proposed to use reliable results of pair-wise registration to augment the object model. Multi-view registrations are accomplished by alternately executing the pair-wise registrations and judgments, and model augmentation. Experimental results on available public data sets and comparisons with some state-of-the-art approaches show that the proposed approach increases the registration efficiency by about 5 times on average and largely reduces registration errors.