Hongbo Kang

Title: Hongbo Kang: Advancing 3D Reconstruction of Dynamic Crowds

Introduction

Hongbo Kang is an inventor based in Tianjin, China. His work includes a patent on reconstructing the positions, poses, and shapes of people in large-scene videos. The patent record provided lists 1 patent.

Latest Patents

Kang’s listed patent is “Method for Dynamic 3D Crowd Reconstruction from a Large-Scene Video.” It introduces DyCrowd, a framework designed to reconstruct hundreds of people from video footage, even when dense crowds and frequent occlusions make individual movements difficult to track.

The method uses a top-down strategy involving pre-reconstruction, matching individual movement sequences, and multi-stage iterative optimization. It also introduces group optimization with an asynchronous motion consistency loss. By grouping people with similar trajectories, the approach uses clear, unoccluded movements to help recover the motion of individuals who are obscured, addressing challenges caused by prolonged occlusion.

The work also presents VirtualCrowd, a virtual benchmark dataset intended to address the shortage of ground-truth human reconstruction labels for large-scene videos.

Career Highlights

The described invention combines crowd reconstruction, motion-sequence matching, and group-based optimization in a unified framework. Its focus is on recovering detailed 3D information about many people from complex video scenes.

Collaborations

Kang’s listed coworkers include Kun Li and Jian Ma.

Conclusion

Hongbo Kang’s patent presents a framework and benchmark for reconstructing dynamic crowds from large-scene videos, with methods designed to improve recovery when people are frequently occluded.