A Novel Method for Registration of MLS and Stereo Reconstructed Point Clouds

Cross-source (CS) point cloud registration is a prerequisite for effectively leveraging the complementary infor- mation of multiple 3-D sensors. However, existing point cloud registration methods have primarily focused on the registration of mono-source point clouds and typically fail to register CS data with varying noise patterns and capture characteristics. In this article, we present a new algorithm for CS point cloud registration between mobile laser scanning (MLS) point clouds and stereo-reconstructed point clouds (SPCs). Our method has two key designs. First, we design a novel descriptor with in-plane rotation equivariance by leveraging the accessible gravity prior, yielding strong descriptiveness, better robustness, and improved efficiency. Second, based on the noise pattern of SPCs, a novel disparity-weighted correspondence scoring strategy is proposed to strengthen the registration accuracy. In comparison to existing registration baselines, our method achieves a 32.6% higher regis- tration recall (RR) on CS datasets of KITTI and KITTI-360 and a 23.1% higher RR on mono-source datasets of KITTI. Notably, our method also outperforms RANdom SAmple Consensus (RANSAC)-based methods in terms of computational efficiency with a 10 ? 70 speedup. The source code and datasets have been available at https://github.com/WHU-USI3DV/MSReg.