Urban obstacle detection for unmanned aerial vehicle using stereo vision


To develop a lightweight and robust stereo vision system that can support autonomous navigation and collision avoidance of the UAV.


As UAV technologies get more easily accessible, it is inevitable that such machines are being used more frequently and affordably by commercial users in urban built up areas. Current VTOL systems are usually flown within the visual line-of-sight and require human intervention to avoid obstacles. It is envisaged that as regulations and technologies progress, it may be possible to allow autonomous operation of UAVs flying in urban or congested environment through the capability of obstacle detection and avoidance for safety purpose. Currently, effort has been made on applying visual SLAM, efficient stereo vision and optimal path planning using lightweight processor so as to allow practical usage on our in-house developed UAV to avoid the obstacles for commercial applications. The operating system is implemented such that the synchronized image sequences captured by the vision camera are first transformed into disparity map, which is subsequently used to generate laser scan messages. These messages are essential for the evaluation of potential obstacle ahead as well as the determination of optimal flight path.

                                                          Processing board and stereo vision mounted on the UAV


                                         Generation of disparity map using synchronized images captured by stereo vision

               2D map illustration of obstacles detected at a distance of 2m (left) and 16m (right), respectively. 
                       (Red: obstacle, green + dark red: location of the camera integrated into a UAV)

Principal Investigators

Associate Prof. Wang Han (NTU)

Telephone: 6790 4506
Office: S2-B2b-49
Mr. Paul Tan (STE)

Telephone: 6660 1052
Office: S1-B4a-03