The PAVE Vision team has begun gearing up for the 2009 IGVC. We will be continuing our stereo-vision approach to obstacle and lane detection, and this year we will be using a Videre Design STOC (Stereo On Chip) color stereo camera. This camera’s innovative design includes an onboard Xilinx Spartan 3 FPGA that can stream live disparity maps to the PC. This frees up resources on our computer for depth map analysis and all the other processes that control the robot.

The image above shows a sample frame from the camera. The source image is at the bottom-left, and the disparity map calculated onboard the camera is shown at the bottom-right. The center window shows the 3D reconstruction of the scene, with the source image mapped onto the point cloud reconstructed from the disparity map.

As the semester progresses, the Vision team will provide updates on the modifications we are making to improve our obstacle and lane detection systems. We look forward to demonstrating our single-sensor approach again at the competition, and hope to make our system even more successful than last year.