| The ocean occupies about 71% of the earth’s surface area and has rich reserves of energy,mineral and biological resources,most of which have not been developed and utilized.With the continuous development of human society,people’s demand for energy and resources is increasing,and countries all over the world focus on the development and utilization of the ocean.Unmanned Surface Vehicle(USV)is an unmanned platform with environment awareness and autonomous decision-making capabilities.It can perform a variety of military and non-military missions with the equipment and system it carrys.The surface unmanned vehicle can identify and confirm the surrounding environment through various sensors,plan the route and avoid obstacles according to the predetermined tasks,and implement many functions such as environmental detection,hydrological exploration,accurate attack,etc.It has high intelligence.The intelligent surface unmanned vehicle will play more and more roles in the future,and the research on the surface unmanned vehicle has important practical significance.In this paper,the visible and infrared image fusion method of the perception system of the surface unmanned vehicle is studied.The purpose is to obtain the fusion image with high clarity and information content,which can support the subsequent recognition tasks of the unmanned vehicle.The main research contents of this paper are as follows:First of all,this paper introduces the research background and practical significance of this topic,reviews the development of domestic and foreign surface unmanned vehicle,and expounds the research status of visible light and infrared image fusion technology.Secondly,the visible and infrared cameras are calibrated,and the image enhancement algorithm is applied and improved according to the actual scene of surface unmanned vehicle application.For problems such as fog environment and insufficient exposure,image enhancement algorithm is selected and improved to improve the clarity of captured images and provide convenience for subsequent algorithm research.Then,the existing visible and infrared image fusion algorithm framework is introduced and validated.according to the existing algorithms,several new fusion strategies are proposed to fuse the visible light and infrared feature information extracted by the neural network.The generated image can fully retain the texture edge features of the infrared image,and the scene description is more perfect.Finally,the proposed fusion strategies are tested and verified.Fusion images are generated according to the data obtained by the surface unmanned vehicle.The effectiveness of the proposed fusion strategies is verified by measuring and comparing the effects of several fusion strategies with a variety of indicators commonly used in the image field. |
