Research On The Front Vehicle Detection Method And Hardware Implementation For Intelligent Vehicle Application

With the advent of the trend of driverless cars,the driving safety of intelligent vehicles has attracted more and more attention.At present,the research on the front vehicle detection algorithm plays an important role in driverless system.Therefore,this thesis studies the front vehcile detection technology.In order to improve the accuracy and detection efficiency of the current vehicle detection algorithm,this thesis has carried out the front vehicle detection method based on the new Haar feature and the improved cascade iteration algorithm,optimized the Haar feature template and cascade iterative algorithm,and realized the detection and recognition process of the front vehicle combined with the embedded development platform.Firstly,this thesis briefly expounds the research background and significance of front vehicle detection technology,investigates the relevant research status at home and abroad,and summarizes the work of the full text.Secondly,the front vehicle detection system based on DE1-SoC is built,including using USB camera as image acquisition equipment,embedded development board DE1-SoC as image processing equipment to realize the improved algorithm function,and VGA display as detection result display equipment.Then,the commonly used vehicle detection algorithm is investigated.Based on the current research on Haar feature + AdaBoost algorithm,a new Haar feature for the vehicle in front is proposed to train the classifier.The weight updating rule of the strong classifier in the AdaBoost iterative algorithm is proposed to reduce the over fitting phenomenon in the training process.A vehicle width matching algorithm is proposed to screen the area to be detected to improve the efficiency of the detection process.According to the proposed algorithm,the video image in the driving environment of the front vehicle is experimentally simulated,and compared with mainstream vehicle detection algorithm.Combined with the target detection evaluation standard,the detection accuracy and rate of the proposed algorithm are evaluated.It is proved that the proposed algorithm can improve the detection rate to 97.4% compared with the traditional detection algorithm.Finally,the transplantation of the front vehicle detection method is completed by using the embedded platform DE1-SoC based on ARM + FPGA architecture.Through the built vehicle detection system,the vehicle images and videos in the database and the actual environment are detected respectively,and multiple groups of experimental data are obtained,analyzed and summarized from the aspects of evaluation index,power consumption and processing time.It is proved that the front vehicle detection algorithm proposed in this thesis can be applied and developed combined with embedded system and has good detection effect and portability.