Research On Key Technologies Of UAV-USV Collaboration Based On The Assistance Of Computer Vision

With the development of the technology of micro-drone,Unmanned Aerial Vehicle(UAV)has been given more and more functions,which are needed in reality.Compared with the vision system in Unmanned Surface Vehicle(USV),the vision system in UAV can acquire the broder view,and the limitations of the area where a UAV can reach are fewer than USV.Therefore,the onboard vision system in UAV can improve the ability of USV.Also,in order to ensure the autonomy of the whole system,the UAV should autonomously land on the USV after completing the missions.In view of above problems,this paper proposes three vision-based algorithms,which can help the UAV complete the missions better,including image enhancement,floating objects detect and autonomously visionbased landing.Furthermore,the system is constructed in real world and performs well.(1)The flight environment of the UAV is complex and diverse.The images acquired in the environment with low light intensity,such as cloudy days,are unclear and the contrast is low.It is more difficult to analyze and process the image subsequently.Therefore,the first step is to enhance the image,including image denoising,and image enhancement.This paper designs an end-to-end deep neural network to complete the image enhancement task.(2)This paper makes use of the characteristics of low-altitude aerial image captured by drone to develop a floating objects detection algorithm based on saliency map.(3)For the overall autonomy of the system,this paper designs a vision-based algorithm and control system of the UAV for the autonomous landing on the USV.The research results show that the control system designed in this paper performs well in the experimental process and has important application value.