Research On Moving Target Tracking Technology Of Quad-rotor UAV Based On Vision

As the one of the key technology of UAV monitoring system,the recognition and tracking of moving targets can be used as an important basis for UAV to execute flight missions by capturing the location and trajectory of targets,through the moving target is recognized and tracked in the real time by the design algorithm.The basis of the target tracking in Aerial Robotics Competition,a set of quad-rotor UAV target tracking control system is designed to optimize attitude estimation algorithm and image processing algorithm,and reduce the system operation time,so as to improve the real-time performance of target tracking.Firstly,in order to build the whole quad-rotor target tracking hardware system,the overall design structure of the hardware system is planned,and the requirement of the hardware system is put forward.Then,on the basis of the demand,the hardware system is divided into flight control system and visual tracking system,then the design and selection of hardware modules have been completed.Secondly,the target tracking algorithm and software design are studied.The target tracking algorithm based on kernel correlation filtering is emphasized,the algorithm uses the circle matrix to solve ridge regression problem,and the diagonalization with Fourier is introduced into the ridge regression solving process,which greatly improves the computational efficiency.At the same time,the kernel function is introduced on the basis of linear regression to improve the performance of the classifier.In addition,aiming at the problem of target tracking algorithm scale fixed,we use one-dimensional scale filter to estimate the change of target scale,which effectively improves the accuracy of target recognition.The software system uses a PID control law structure that inputs the relative position error as the track loop.Thirdly,to achieve the stability of quad-rotor UAV tracking control,the gradient descent algorithm with high accuracy,high computational efficiency and fast calculation is proposed.The algorithm is developed in quaternion style to describe and fuse the data of gyroscope,accelerometer and magnetometer to achieve the fast estimation of attitudes,the effectiveness of the algorithm is verified by flight experiments.Finally,the performance evaluation and fight function verification of the target tracking algorithm are carried out.Integrating the hardware and software design,carried out the whole system power function test and flight environment test,and analyzed two sets of static and dynamic tracking data.The experimental results show that the quad-rotor UAV target tracking control system designedin this paper has stable flight ability,at the same time,it can achieve high-precision and real-time tracking of moving targets.