Research On Intelligent Obstacle Avoidance Of Small Quadortor Uav Based On Machine Vision

Quadrotor UAV are the most widely used and most popular drones.Since traditional drones avoiding obstacles and planing paths rely on manual operations,it is difficult to accurately control the drone flight.Therefore,this paper studies the construction of intelligent indoor 3D maps and obstacle recognition based on depth cameras.This obstacle avoidance system can monitor indoor environment in dark conditions,especially the relief site terrain monitoring,disaster path detection,etc,accurately detecting the characteristics of the indoor environment can also avoid unnecessary injuries or deaths,and accelerate disaster relief efficiency.In this paper,quadrotor is used as the experimental platform for indoor intelligent obstacle avoidance research of drones.The following work has been carried out:Aiming at the problem of image processing delay of obstacle avoidance for drones,the visual image acceleration scheme was studied,and a visual image acceleration scheme based on FPGA was designed.First,the speed of drone is relatively fast,in order to ensure the realtime nature of image processing,analyze and compare vision acceleration schemes,select FPGA processors as the acceleration of drone vision detection.Use the advantages of FPGA's unique parallel processing data to accelerate the visual image,design the corresponding visual image algorithm and write the program using Verilog language.The visual acceleration experiment uses an edge detection algorithm,parallel data processing for image processing,at the same time,the resources inside the FPGA are used for visual acceleration processing,thereby ensuring the real-time nature of obstacle detection.Aiming at the problem of 3D map construction of small quadrotor UAV in complex indoor environment,a new type of indoor 3D map construction scheme is proposed.For a complex indoor environment,especially when the indoor light is insufficient or not available,using ordinary cameras cannot normally obtain indoor obstacle information,this article uses a structured light camera to detect the indoor environment,it can obtain image information of obstacles and the depth information of it in the room regardless of whether there is a light source.The algorithm uses an anti-sensor model,the anti-sensor model has a simple structure,with high detection efficiency,it can accurately detect indoor obstacles and build indoor 3D maps,laying a foundation for obstacle avoidance and path planning of drones.Aiming at the obstacle detection and flight path planning during the obstacle avoidance of UAV,the YOLO algorithm is used to detect the category information of the obstacles,and the TSP algorithm is used to conduct the path planning research of the UAV based on the indoor 3D map construction.A mathematical model of obstacle avoidance control is established in a hypothetical scenario,and the best flight path is obtained through simulation analysis and comparison.