Real-time lane departure and front collision warning system on an FPGA

Driving safety is a top priority when designing intelligent automobiles. In this paper, we present an FPGA-based real-time system for lane departure warning (LDW) and front collision warning (FCW), which requires intensive computation and memory access for high resolution video input from a front-view camera. Our system first converts gray scale video input into desired binary video images by applying Sobel filter, Otsu´s threshold binarization and dilation. Hough transform is then applied to detect lanes for LDW. The front vehicles are detected for FCW by searching shadow area on the images. Our experimental results demonstrate the FPGA design is fully functional and it can process about 160 frames per second for 720P video input, which exceeds the real-time processing requirement.