High Accuracy and Simple Real-Time Circle Detection on Low-Cost FPGA for Traffic-Sign Recognition on Advanced Driver Assistance System

This paper describes a hardware oriented algorithm and its conceptual implementation for real-time traffic signs detection system on automotive oriented FPGA. The speed limit sign area on a grayscale video frame is detected through a two-stage simple computation process. Rectangle Pattern Matching roughly detects global luminosity sharing feature between rectangle and circle for Region of Interest (ROI). Then, Circle Detection roughly votes local pixel direction of circle inside the detected ROI in binary image for circle confirmation. The proposed system achieves 83 full HD fps and over 99% accuracy even in difficult situation such as rainy night. It occupies around 50% the hardware available on proposed Xilinx Zynq automotive FPGA, which has 85 K logic cells, 53.2 K LUTs, 106.4 K registers and 506 KB BRAM, and so be able to apply to Advanced Driver Assistance System on common vehicles.