FPGA-Based Object Detection for Autonomous Driving System

Autonomous driving systems require a real-time detection of the objects including pedestrians and obstacles on the roads. Object detection by image processing is a popular approach for autonomous driving systems. However, there are complex reflections caused by multiple light sources and objects on the roads, which disturb the robust and real-time object detection. This paper describes an FPGA-based object detection method for autonomous driving systems using multiple CMOS cameras. Our system is implemented on Xilinx Zynq-7020 SoC-FPGA, in which real-time processing of tracking white-lines for lane-keeping and obstacles/pedestrians detection on the roads are executed by the hardware on the Programmable Logics, and the whole system is controlled by a software on the Processing System (CPU) written in Python language.