Hardware Collaborative Design Of Vehicle Pedestrian Detection System Based On Vision

With the rapid and steady development of China's economy,people's living standards are increasing,and the amount of car ownership is growing rapidly.Cars bring convenience to people's lives,but also bring a lot of negative effects.The safety of road traffic is one of the most concerned problems.It is an important and basic task in the field of intelligent transportation to protect the pedestrian in the road by the vehicle auxiliary system with pedestrian detection function.Based on the study of pedestrian detection correlation algorithm and vehicle pedestrian detection system,this paper has successfully designed a vehicle pedestrian detection system with good real-time performance,high accuracy,small volume and low power consumption.The work done in this article is as follows:(1)I have read a large number of documents on pedestrian detection and pedestrian detection technology in vehicle equipment,summed up the difficulties of vehicle pedestrian detection technology,using ZYNQ chip based on ARM+FPGA architecture.Combined with the hardware conditions of AX7010 Development Board,the software and hardware design scheme of pedestrian detection system based on vision is designed.(2)A pedestrian detection framework based on centrist feature and cascaded support vector machine classifier is built.In the feature extraction phase: firstly,the original color image is converted to grayscale image,and then the image is scaled to construct image pyramid.Then the Sobel operator is used to detect the edge of the image and construct the Sobel image.Then the CT image is constructed by calculating the CT value of each pixel of Sobel image.In the detection phase: The Cascade classifier is used,and the first stage is linear support vector machine,which is used to quickly and roughly locate the candidate region,and the second stage adopts the HIK kernel support vector machine to classify the candidate region accurately and get the final pedestrian detection result.So,the whole pedestrian detection algorithm has fast and accurate effect.(3)The embedded development environment based on ZYNQ extensible platform is built,including making system startup file Uboot,equipment tree devicetree.dtb,file system andembedded Linux operating system,OPENCV image processing library and Qt Image interface library.(4)In the Centreist feature extraction,it is necessary to preprocess the image,such as image scaling and Sobel edge detection,these operations are more time-consuming.Based on the advantage of the ARM+FPGA architecture,this paper uses the ZYNQ chip to realize the hardware acceleration of the image preprocessing algorithm using FPGA,and then transmits the preprocessing image to the ARM processor for feature extraction,classifier detection and result display through the VDMA bus.The speed of the whole pedestrian detection system is improved without changing the detection accuracy.(5)using the software and hardware collaborative design method to realize the vision-based vehicle pedestrian detection system,and the performance of the system,such as accuracy,detection rate,false detection rate and testing speed.Through the analysis of test data can be achieved,the system basically achieves the expected effect.