Obstacle Avoidance Control System Design For Quadcopters Based On Machine Vision

Quadcopters,as a kind of UAV with new structure and excellent performance,has a wide range of applications.Autonomous obstacle avoidance is an urgent need for UAV to realize intelligent flight.Obstacle avoidance based on machine vision is a hotspot of research.This thesis presents the methodology based on machine vision for solving the problem of obstacle avoidance of UAV by the following work.Firstly,this paper studies the principles and steps of binocular-vision & ultrasonic ranging,and studies the current multi-sensor information fusion technology.Then,the data collected by the two obstacle sensing methods are merged.Secondly,a machine vision perception system was designed to acquire the obstacle distance information during UAV flight.Then,an airborne image processing software for binocular-vision calibration,matching and depth extraction was built based on MFC,and then fusion with the ultrasonic distance information,the final output obstacles distance information.Then,a mathematical model of Quadcopters is established and the parameters of dynamic system are identified.A hierarchical multi-loop PID control structure is designed to realize the position control and attitude control of UAV.Aiming at the problem of navigation control in obstacle avoidance route,a navigation control algorithm based on recursive virtual reference target point is proposed,which can track the desired trajectory.In order to solve the problem of mission state scheduling in route,An extended mission scheduler is proposed to realize the switching and recovery of task status in obstacle avoidance route.Finally,the stability validation and analysis of the obstacle-sensing system are carried out on the flight platform.Then,the obstacles fusion algorithm and the obstacle-bypass control algorithm are validated in the specific environment and achieve the dynamic detection of obstacles and simulation of obstacles around the flight control.