Research On The Tracking Algorithm Of The Unmanned Ground Vehicie Based On Machine Vision

An unmanned ground vehicle(UGV)is an autonomous land vehicle system,which consists of environmental perception,path planning,automatic control without personnel operation.Assisting or replacing mankind,the vehicle system could complete the certain mission in complex terrain conditions.Accurate real-time object tracking is the prerequisite and guarantee of multi-vehicle coordination system in the application scenario of multi-UGV coordination and dispatching.When tracking the target,the UGV based on traditional fixed camera has the shortcoming of narrow field of view,which increases the difficulty of tracking process and the complexity of the system.Aiming at the shortcomings of the traditional fixed camera performing on the process of object tracking,this paper studies the UGV tracking algorithms based on PTZ(Pan/Tilt/Zoom)camera machine vision,which apply a free-turning PTZ camera on the UGV testbed designed and developed all by our laboratory.First of all,analyze the existing UGV architecture and replace the fixed camera with PTZ camera.Combined with the characteristics of tracking mission and vehicle equipment type,an UGV architecture based on PTZ camera is designed.Secondly,this paper studies the visual servo control algorithm based on PTZ camera and designs the PTZ motion control algorithm based on image coordinate offset feedback.According to the relative position,the algorithm can drive the camera turning to track target so that it can ensure the target is always in the center of the image.This method effectively solves the problem of losing tracking target because of the narrow field of view of traditional fixed camera.Thirdly,a modified POSIT iteration algorithm is studied for estimating the position and attitude,and according to this theory,a tracking algorithm based on monocular vision and inertial navigation fusion is designed.Applying with Extended Kalman Filter(EKF),the algorithm fuses the visual data and the inertial data to realize accuracy pose estimation with respect to the target.Furthermore,it helps to realize real-time object tracking.Fourthly,the control system of the UGV steering gear is studied.A local path planning method for real-time autonomous adjustment of S-curve is designed.The method can be used to determine the tracking behavior of UGV on the basis of pose estimation.Then this paper studies the control algorithm of the UGV steering gear based on the incremental PID method,and the optimal decision is made to send the steering instruction to the steering gear which will drive the UGV to go forward.Finally,the UGV object tracking system based on PTZ camera is implemented and several experiments have been carried out to verify the feasibility through the whole process.The experiment results show that the UGV system based on PTZ camera can complete the mission of object tracking accurately and efficiently.