Design Of Target Recognition And Tracking System For Unmanned Platform Based On Machine Vision

Researchers around the world are attracted to the rapid development of machine vision in recent years.Modem urban warfare will be changed by the development of object detection and tracking.The study of machine vision in Unmanned Ground Vehicle(UGV)will enhanced the combat capability of the army and modernize military.Research object UGV is exemplified by this thesis.The object detection and tracking machine vision system and pan-tilt control are analyzed and designed.(1)In view of the object of study in the thesis,vehicle and person is chosen to be detected.The characteristics of vehicle and person are analyZed First,global feature is used to detect and locate the object by OpenCV.Next,Adaboost algorithm is used to train model and detect object.YOLO based on CNN is used to detect object more exactly.To track any object,Camshif lgorithn and the kalman filter modified Camshift are studied,the anti-interference ability is improved.KCF algorithm is used to improve the tracking ability of the system.(2)Graphical user interface is designed by MFC on Windows.Communication protocol of serial port between PC and lower computer is designed to ensure the stable and accurate control.A few of programs are transplanted to DJI manifold lnbedded Ubuntu to enhance maneuverability.(3)Pan-tilt is analyzed to verify the motors,STM32F405RG is used to design the control system.The speed of MPU6050 is used to design speed loop PID control,mechanical angle from motors is used to design position loop PID control,the stability of pan-tilt is improved by dead zone control.(4)Testing experiment is done to verify control ability of pan-tilt.The object detection and tracking system testing experiment is analyzed by using statistics,the tracking ability of the syten is verified.