Research On Key Methods And Technologies Of Quadrotor Vision Drone Based On STM32 Development Board

As the main representative of unmanned gyroplanes.The quadrotor,which is also known as the quad-rotor unmanned aerial vehicle(uav),has made a qualitative development in all aspects.While the uavs are rapidly put into practical use and industrialization,it is now heading for microminiature civil area gradually.On the benefit of its symmetric structure,lightweight resulted from using MEMT(Micro-electro-mechanical technology),characteristic of vertical taking off and landing(VTOL)and eliminating itself’s reaction torque through rotor wing’s rotation.The uav not only gets extra attention from the civil area and academe,butgradually highlights its scientific value.This paper mainly uses the uav as the experimental prototype,makes it combine with vision recognition technique,and research its application value in the field of uav and visual technology cross,which includes:First,this paper introduces the uav’s flight control principle,the needed coordinate system,several representation methods of the flight attitude and builds its dynamic model basing on the characteristic of combining with vision recognition technique.Second,this paper deeply analyzes and researches on the uav’s hardware structure,designs the main control system including the drive circuit,the controller,sensors,the power supply circuit and the wireless communication system and the remote signal,assembles the prototype and tests its stability basing on actual flight.Then,this paper puts forward a method of using the feedback current of motor driving to improve the stability of the experimental prototype,and applies it in the uav’s speed controller.This method works by putting an inner control loop into every driver and obtaining the same motor speed with any given control input.Then test this method by the actual operation.If uav’s euler Angle is close to the origin of coordinate and the sensors works well,the uav is able to complete positioning,cruise and even autonomous recognition and tracking of specific objectives.Finally,the paper puts forward a target tracking method based on SIFT features and particle filter.During a certain period,the full occlusion of the target results in the target model feature point being all deleted and no new feature points added,then use the method of no update the target model,keep the characteristics of the latest used output target,and use this characteristics to update the target model.Then use this model to match the candidate matching models,which ensure the rapid and accurate matching.At last,track the target around the location of the target matching.