Design And Implementation Of A Quadrotor With Vision-based Landing

Quadrotor is widely used in agriculture,commerce,military industry and other fields,since its small volume,low cost and strong adaptability.Its theoretical research involves rigid body dynamics,aerodynamics,control theory,inertial navigation,signal detection and filtering and other disciplines,quickly and accurately obtain the quadrotor attitude and position is one of the elements directly decides the quadrotor flying quality.In general,quadrotor navigation and positioning uses GPS,indoors,close to the ground,or some bad GPS signal environment,the navigation precision is reduced.Visual landing is a visual navigation method,which can realize quadrotor navigation,and automatic landing on special occasions.Therefore,it is of great theoretical and practical value to study the attitude and position algorithm of quadrotor with vision-based landing technology.Firstly,in view of the traditional attitude and standard EKF attitude algorithm not high accuracy and poor dynamic performance problems,an improved EKF algorithm is proposed.Motion acceleration restrained with dynamic step gradient descent algorithm is introduced into the standard EKF,improves the problem of slow convergence caused by the fixed step gradient descent algorithm,and he problem of poor accuracy caused by motion acceleration in the strong maneuver.And the improved EKF algorithm separates the magnetometer to calculate the yaw angle only.A positional algorithm of inertial fusion GPS and barometer is designed.Secondly,a fast and accurate image recognition algorithm is designed for the problem of Harris corner detection large amount calculation.Through transformation of all corners in a region into a single value corner,instead of the local non-maximal suppression algorithm in the Harris,reduces the amount of calculation.Designed attitude and position conversion algorithm of quadrotor vision mode,and visual landing control algorithm.Finally,for the problem of barometer is affected by airflow and accelerometer is sensitive to vibrationthis,a box flight control unit is designed.The IMU through 3M glue structure into the damping system inside of the box light control unit.Thus avoiding the interference of the airflow to the barometer and the influence of the frame vibration on the accelerometer.On this basis,the independent design and construction of the experimental platform,The effectiveness and practicability of algorithm and design of the system in this paper are verified.The platform includes: a quadrotor,a box flight control with damping system,a vision processing module and a vision debugging module.The results show that improved EKF has faster calucation speed,higher tracking accuracy,stronger magnetic disturbance resistance.Target image recognition algorithm has higher identification accuracy;And eventually this paper achieves quadrotor visual landing function.