Shaolei Liu

Fudan University, Shanghai Medical College of Fudan University, Zhengzhou University of Light Industry

Papers

Total Citations

295

H-Index

About

Shaolei Liu is a computer vision and robotics researcher whose work centers on 3D point cloud processing, deep learning-based geometric matching, and multimodal data fusion. Liu is best known for pioneering a robust point cloud registration framework grounded in deep graph matching — a contribution that has garnered over 280 citations across related publications and variants, signaling substantial influence within the field. This work directly addresses one of the most persistent challenges in 3D scene understanding: the sensitivity of learning-based registration methods to outliers, which can corrupt correspondence estimation and degrade alignment accuracy. By leveraging graph matching principles within a deep learning pipeline, Liu's framework advances the reliability of point cloud registration even under conditions of high outlier ratios and without requiring good initialization — a practically critical scenario in robotics and autonomous perception. More recently, Liu has extended research interests toward multimodal image fusion for enhanced 3D reconstruction, combining novel registration strategies such as IKKD-tree with fusion architectures like CSTDFusion. Collectively, Liu's body of work reflects a consistent drive to make 3D spatial understanding more robust, scalable, and applicable to real-world environments where data imperfections are unavoidable.