A Survey of Computer Vision Research for Automotive Systems

Current autonomous driving technology heavily relies on a special sensor, Velodyne Sensor, and a detailed precision map. From Darpa Grand Challenge in 2005 to current Google Car, Velodyne Lidar sensors are used widely. They could provide 360 degree real-time 3D distance map by illuminating environment with lasers and analyzing the reflect light. The 3D distance map can be used in obstacle detection and navigation. However, according to the marketing report, Velodyne LIDAR scanner used in Google Car costs 70,000 dollars which is much higher than the average budget people expect. Another constrain Google Car facing is a high definition inch-precision map. For example, the map should include the height of traffic light. That means Google Car can only drive when the precision map is provided. It limits usage of Google Car to only small areas. Although these two technologies provide important functions for autonomous driving, they also make universalizing autonomous car become impossible.