Xudong Lv

Harbin Institute of Technology, Hangzhou Dianzi University

Papers

Total Citations

232

H-Index

About

Xudong Lv is a leading researcher in autonomous systems, specializing in multi-sensor fusion, 3D perception, and robot navigation. His most significant contributions lie in LiDAR-camera calibration, where he has pioneered deep learning-based methods that eliminate the need for laborious manual setups and complex calibration targets. His highly cited work, "LCCNet: LiDAR and Camera Self-Calibration using Cost Volume Network" (175 citations), introduced a cost volume network that achieves robust, targetless calibration, a critical precondition for effective sensor fusion in self-driving and robotics. He further advanced this field with "CFNet: LiDAR-Camera Registration Using Calibration Flow Network" (52 citations), which improved registration accuracy through a novel flow-based architecture. Lv’s recent work, "DMCL: Robot Autonomous Navigation via Depth Image Masked Contrastive Learning" (5 citations), tackles the challenge of learning robust state representations from high-dimensional pixel inputs for deep reinforcement learning, addressing a key bottleneck in autonomous navigation. By automating and enhancing sensor calibration and representation learning, Lv’s research directly enables safer, more reliable autonomous vehicles and robots, with his methods widely adopted for their efficiency and accuracy in real-world applications.