Research Of Traffic Sign Detection Based On ZYNQ-SoC Chip

In the automatic drive technique,traffic signs detection and recognition is a significant topical subject in this area.The traditional algorithms of traffic signs detection are realized on computers.However,with the development of embedded platform and the ability of multiple cores hardware systems which combined FPGA and ARM within one chip rises these days,the applications of traffic sign detection on embedded platform showed more portability,more practicability and more universality.In this paper,it comes a new realization of traffic sign detection on SoC based on the situations in Image detection above.The main research contents are as follows:(1)The researches of feature of traffic sign detection based on HOG featuresHistogram of Oriented Gradient Features are formed by computing and analyze of local histogram of oriented gradient in an image.This paper takes 1000 images with many kinds of traffic signs from the German Traffic Sign Detection Benchmark as a positive set,and 1000 images which do not contain any traffic sign in the same size as a negative one.The HOG features are described as the whole training set which are used to train the BP neural network.(2)The researches of the BP neural network classifierThe algorithm of Error Back Propagation neural network is researched.The ability of network is raised to the request by adjusting its abilities of self-studying and generalizations,also the weight coefficient.Principal Components Analysis method is used to reduce the dimensionalities of the vector.(3)The application of theoretical detection algorithm based on ZYNQThe advantages and disadvantages of some kinds of traffic sign detection algorithm on different embedded platforms are analyzed.It comes up there are deficiency of traditional SoC,the realization of ZYNQ which is produced by Xilinx has more feasibility.Its architecture combined ARM and FPGA provides more possibility synergism between software and hardware.