Research On Pedestrian Detection Technology In Vehicle Assisted Driving System

With the rapid development of the country's economy,the number of vehicles is constantly increasing.Due of many accidents caused from driver's subjective mistakes,the safety of pedestrian is threatened seriously.The vehicle assisted driving system can help the driver to control the vehicle,which can determine intelligently whether the target that appears in front of or around the vehicle is pedestrian.Pedestrian detection is important technology in auxiliary driving systems because man's life is more precious than anything.Pedestrian detection method is a valuable and challenging topic in the field of computer vision because pedestrian with both rigid property and flexible property,whose appearance is easily affected by such factors as clothes,occlusion,scale,posture,and viewed angle.The main work of this thesis is as follows:(1)Pedestrian detection technology based on static images is researched.The gradient direction histogram feature(HOG)is analyzed in the image feature extraction step,and support vector machine(SVM)algorithm is used as a classifier to test on the INRIA data set.The Deformable Parts Model is an extension of the HOG feature,which can effectively solve the uncertainty of the pedestrian pose.The thesis analyzes the implementation process of the model and tests the model performance on the VOC 2007 data set.(2)Methods of Pedestrian detecting in video streaming are researched in this thesis.The frame difference,optical flow and Gaussian mixture model are analyzed here and their advantages and disadvantages are compared.The Gaussian mixture model and EM algorithm are also studied in detail.A pedestrian detection system is developed based on Visual C++ 2017 programming environment,which can detect existing pedestrians in the stream and track the detected pedestrians.(3)Finally,the application of deep learning in target detection is studied.The Single Shot Multibox Detector(SSD)uses different default frame sizes on different output layers.The experimental results show that the algorithm has high detection efficiency and fast detection speed,which can finish pedestrian detection operation under certain conditions.