Research on object tracking and recognition based on multi-sensor fusion

A new framwork is presented for the maneuver target locating, tracking and recognition by integrating the radar and machine vision information. The supority of the two kinds of sensors is used, and the performances of realtime and object tracking and recognition accuracy are highly improved by integrating advanced Hausdorff-based three dimentional (3D) model matching technology, non-linear unsented Kalman filter and MMSE-based data fusion algorithm. Experiments on 3D simulation scenario show the proposed method can effectively track and recognize maneuvering targets strongly.