A Research On LADAR-Vision Fusion And Its Application In Cross-Country Autonomous Navigation Vehicle

The temporal-spatial consistency and the multi-level fusion algorithms in LADAR-vision fusion have been investigated in this dissertation. This project is the core of the environment perception system of the compact cross-country autonomous navigation vehicle developed by our university. The following results and innovations are included in this dissertation:1. The temporal synchronization method for autonomous navigation multi-sensor data fusion is proposed, which can achieve the synchronization of the sampling time of multiple sensors with high precision.2. The LADAR-camera direct calibration method based on controllable-plane constraint is proposed, which can realize the direct calibtation of the LADAR scan points and camera image pixels. The feasibility and validity of the calibration are verified by experiments.3. The image fusion algorithm with multi-feature temporal sequence, which comprises pixel level fusion and decision level fusion for multi-layer LADAR is proposed, so that the detection of small obstacles in rough terrain is realized.4. A pixel level fusion algorithm for LADAR and monocular vision is proposed, and the grass detection is achieved with SVM classification; The LADAR and monocular vision fusion algorithm using road information has been proposed to enhance the efficiency of the obstacle detection, which satisfies the requirement of realtime for autonomous navigation in rough terrain.5. The compact fusion and loose fusion of LADAR and stereo image pair are proposed. The fusion of the LADAR and stereo vision on decision level is realized to increase the robustness of obstacle detection, which insures the safety of the vehicle's driving autonomously in rough terrain.Based on the above results, the environment perception system of the Autonomous Navigation Vehicle can recognize small obstacles (30cm X 30cm X 30cm) within 12 meters in front of the vehicle, and the processing cycle is less than 100 milliseconds. These research results have made great contribution to our Autonomous Navigation Vehicle's being able to round obstacles in medium level rough terrain at the speed of 5-20km/h firstly in China.