Research On Vehicle Detection Method Based On FPGA Acceleration

In recent years,with the rapid development of Chinese national economy,Chinese national income has been greatly improved,the people's pursuit of quality of life has been continuously improved,and family cars have changed from the original luxury goods to a common commodity.With statistics,the number of cars in China reached 340 million in 2019.With the rapid growth of the number of cars,traffic problems in various cities in China have become increasingly serious,so it is very important to install road monitoring systems for the main roads in cities.The main function of the road monitoring system is to classify and count the vehicles in the video,so it is especially important to develop a low-cost and powerful vehicle detection system.This paper is to study the structure of FPGA+ARM of Xilinx Zynq-7020 platform,and design a road condition real-time monitoring system based on ZYNQ chip.Compared with the traditional embedded image processing system using ARM chip and DSP microprocessor,the Zynq chip of ARM+FPGA structure can realize the video processing of high-speed cooperative work of software and hardware,and has the advantages of fast image acquisition and high bandwidth of image processing system.This design utilizes the advantages of parallel processing of FPGA to realize fast and efficient preprocessing of images;software flow control through ARM processor and operation of target recognition algorithm for HAAR feature extraction using machine learning model;ZYNQ chip comes with The memory to achieve efficient work of software and hardware,in pursuit of processing speed,while reducing the power consumption of the hardware.The main work of this paper is:(1)Programming the FPGA part of the ZYNQ chip,using the Vivado Hls tool to design and write the accelerated IP core of the FPGA chip,which has the function of converting the video image into a format and enhancing the contrast algorithm for hardware acceleration.(2)Design a data transmission channel for the FPGA chip and the ARM chip in the ZYNQ chip,so that the FPGA chip can transmit the pre-accelerated video image to the vehicle detection system running on the ARM chip.The specific work is to design a VDMA IP core using the Vivado Hls tool.(3)Configure the Linux system in the ARM chip part of the ZYNQ chip and write the vehicle detection system program.Mainly for configuring the Linux kernel and documentation,and porting the OpenCV library.The vehicle detection system trains the Cascade cascade classifier through the Haar feature extraction algorithm to classify and track the vehicles in the video.