Implementation With Software And Hardware Of Pedestrian Detection System Based On ZYNQ

With the rapid growth of vehicles, the road traffic safety has attracted widespread attention. The Pedestrian Detection system in car, as an import part of driver assistance system, can protect pedestrians effectively. The speed of a car is fast, and the space and energy of a car is limited. So it has important practical significance if we design a Pedestrian Detection system which has high speed, high accuracy, small size, and low energy consumption.This design achieve the function of identify pedestrians by CENTRIST (CENsus TRansform HISTogram) and SVM (Support Vector Machine)on the base of the current research situation at home and abroad. The pedestrian in the video frame are identified by the method of window scanning.Image acquisition, pedestrian detection, and high-definition display are accomplished in this system by the way of combination of hardware and software in the platform of ZYNQ of the Xilinx. IP of hardware acceleration are realised by FPGA in this system. CENTRIST feature extraction and the design of SVM classifier are accomplished by ARM in this system. This implementation not only makes full use of the logic resources and parallel computing ability of FPGA, but also takes the advantages of the ability of accomplishing complex algorithm and building up operating system of the ARM processors. The main works of this paper as follows:(1)Establish an embedded development environment on ZYNQ of Xilinx, which includs making Linux kernel(zImage)and root file system, transplanting of OpenCV and Qt library.(2)The Programmable Logic in ZYNQ achieves these key modules, including the IP of hardware acceleration, the video transmission channel between ARM and FPGA, and the IP of hardware acceleration can accelerate algorithms by Vivado_HLS, including image scaling, grayscale, and detecting edge.(3)The Processing System in ZYNQ achieves two parts. First of all, it completes some complex algorithms, including CENTRIST and SVM. Secondly, it completes the design of driver program of the VDMA IP.This design completes a Pedestrian Detection system which is applied in cars on the platform of ZYNQ by the way of the combination of hardware and software. The system has the characteristics of high accuracy, good real-time performance, small size, and friendly interface.