A Point Cloud Registration Method Based on Improved Beetle Antennae Search Algorithm and ICP

Since a single stereo vision sensor cannot completely represent the whole picture of the object, it is necessary to use point cloud registration to realize the whole construction of the object, so as to help the robot complete subsequent operations according to the point cloud information of the complete object. Aiming at the defects of the ICP algorithm in the traditional point cloud registration algorithm, which takes a long time and requires a high initial pose of the point cloud, a point cloud registration method based on improved beetle antennae search algorithm and ICP is proposed. In this method, the initial point cloud pair to be registered is preprocessed by subsampling, etc., and then the application of Moth-flame optimization algorithm to the population is used for reference in rough registration to improve the beetle antennae search algorithm, so that the point cloud pair has a better initial pose. Finally, the KD-TREE is introduced in the process of accurate registration to improve the ICP algorithm and achieve the final point cloud registration. The experimental simulation results show that compared with the traditional ICP algorithm, the registration accuracy of the proposed algorithm is improved by 64.1% on average, and the registration efficiency is improved by 82.0% on average, which effectively improves the ICP algorithm's low speed and accuracy in point cloud registration.