Simultaneous clustering classification and tracking on point clouds using Bayesian filter

Simultaneous Clustering, Classification and Tracking (SCCT) maintains many challenges, especially for point cloud data. SCCT is an essential process to facilitate the autonomous mobile systems. We present a novel unified framework from the object extraction to tracking with real-time performance. The framework can be described as five sub-tasks: ground extraction, clustering, recognition, tracking and representation. We compare the adjacent two frames to solve dense tracking and motion estimation. The state of each clustered object (moving or static) is estimated by using Spatial-Temporal methods. The distinguish objects with different features are extracted. Conditional Random Field and Bayesian filter are adopted to solve the data association problem. All the algorithmic modules have been tested on both outdoor actual environments and indoor simulation situations. The results indicate the efficiency and effectiveness of the proposed method.