The Design And Simulation Of A Smart UAV Vision System Based On Deep Learning

Within the past decades,unmanned arial vehicle(UAV)has gained worldwide popularity,and is applied to both military and civilian area.The gradual expansion of UAV application requires that UAVs be smarter than simply fly.The application and promotion of machine learning techniques and computer vision makes it easier to improve flight control,object detection and recognition,and obstacle avoidance.However,the techniques of machine learning and computer vision used by UAVs are still focused more on traditional algorithms.This paper is aimed to build a smart UAV vision system convenient for developing and deploying computer vision applications.We focus on the implemention of computer vision on embedded systemsm and propose that synergetics be applied to study and optimize deep neural network and implement a neural network based on synergetics.First we delve deeply into synergetics and synergetic neural network,introduce the basic approach,mathematical principle and core concepts,and explain synergetic neural network in detail.Next,we introduce the application and implementation of machine learning and computer vision in present UAVs,discuss the research and application of deep learning on embedded systems and mobile terminals.We then introduce the hardware and software available for vision tasks on UAVs,mainly focused on real-time operating systems and build an application development platform for computer vision on a real-time system.About ground control station(GCS)of UAV,we first introuce the present design of GCSes and then propose that deep learning and synergetics be integrated into ground stations.Finally based on this UAV vision system framework,an algorithm for autonomous landing of UAV is proposed and simulated,as an application of this smart UAV vision system.Software simulation is then carried out to prove the applicability of this algorithm.