Study On Front Vehicle Detection Method Based On DSP

In recent years,with the rapid development of technology and the rapid increase of vehicle ownership,road traffic safety accidents have brought great losses to people’s lives and property.Nowadays,it has become the focus of widespread concern.The emergence of automotive active safety technology provides a new direction for improving road traffic safety and has received unprecedented attention.As an important part of the technology,the front vehicle detection technology is the basis for realization of functions such as adaptive cruise,anti-collision warning,automatic emergency braking and so on.It is also an inevitable requirement for the realization of intelligent networked vehicles in the future.This paper focuses on the complex working conditions of highways and other high-grade highways,and conducts researches on the problems of front vehicle detection system and the embedded transplantation of their algorithms based on the machine vision.First,make a deep research on the design and Simulation of the vehicle detection algorithm.The advantages and basic principles of Adaboost algorithm and the construction process of cascade classifier are analyzed.On the basis of traditional cascade Adaboost algorithm,aiming at the robustness requirement of systems under different illumination conditions,the samples are preprocessed by histogram equalization and the classifier model suitable for this system is trained.Matlab and VS simulation are used to study the detection effect of the algorithm under various complex conditions.For the complex environment of Expressway and urban road,in view of the demand for high detection speed of the system,a vehicle detection method based on multi range ROI area selection is proposed,which improves the real-time performance of the system.A vehicle detection system platform is set up.The hardware development platform with DM642 as the core chip is selected,and its module is analyzed.The integrated development environment CCS is used to develop the program,and a new embedded operation system DSP/BIOS is introduced,which shortens the development time of the system.Then the transplantation and optimization of the vehicle detection algorithm are carried out.The main transplantation module of this algorithm is introduced,and the hardware implementation process of each module is described in detail.The internal structure of cascadeclassifier is analyzed,and the initialization process is achieved by defining the pyramidal structure.According to the detection effect after the transplantation,a specific optimization scheme including scaling image,changing the scanning step from the algorithm level,and floating-point fixed-point,hardware memory management and so on from DSP levels is given for the embedded platform.The vehicle detection system is tested through offline database and practical testing.The experimental results of the system under various complex conditions show that both accuracy and real-time performance have achieved good results,and the system has good robustness and real-time performance,and it also has high engineering application value.