A Neural Network Accelerator System for Traffic Lights Recognition Based on ZYNQ

Abstract

In unmanned driving, the recognition of traffic lights and other critical signals depends on real-time performance. Due to the unique structure and low power consumption of FPGA, it has a wide application prospect in the field of artificial intelligence. Based on this, this project designs a hardware and software co-design neural network accelerator system for application in autonomous driving using ZYNQ7020 FPGA development board. The neural network accelerator allows simple configuration of the parameters, scale, and algorithms of the architecture, while providing satisfactory performance. This paper also proposes a set of pre-processing image processes that can improve recognition accuracy. After being tested and verified, this system has high accuracy and good performance, which can meet the needs of autonomous driving.