Research And Design Of Node Autonomous Localization In Underwater Optical Wireless Communication Network Towards Dynamic Enviroment

In recent years,the research on underwater optical wireless communication network goes deeper.In this field,how to realize the autonomous localization of communication network nodes in a dynamic and open underwater environment is one of the research hotspots.Among multiple node autonomous localization technologies,Simultaneous Localization and Mapping(SLAM)technology has attracted extensive attention because it does not depend on the prior calibration of environments.However,moving objects such as fishes will lead to wrong extraction and matching of environmental features,resulting in serious error drift in pose estimation,which eventually leads to localization loss and mapping failure.On the other hand,in order to effectively monitor and control the working status of underwater communication nodes,the design of underwater communication network monitoring and realistic control platform also has important application value.In view of the above two problems,this paper mainly completed the following two innovative work:Firstly,a semantic and geometric features collaborated RGB-D SLAM method for underwater dynamic environments is proposed.Most current visual SLAM methods rely on static assumptions about the environment and thus are prone to failure in dynamic environments.In this RGB-D SLAM method(the SLAM method using RGB-Depth sensors):Firstly,a semantic extraction and segmentation module for underwater dynamic images is designed based on deep learning network,which can remove moving or potentially moving object regions in environmental images;Secondly,based on geometric feature comparison and reprojection error measurement in adjacent environmental frames,the detection module is designed,which can remove dynamic regions with known or unknown semantics in environmental images;Finally,adaptive improvements are proposed based on the ORB-SLAM2 framework(the second generation SLAM framework based on Oriented FAST and Rotated BRIEF features),including key frame optimization strategy in semantic segmentation module and trace thread allocation adjustment strategy in geometry detection module.This method was evaluated on public datasets.The results shows that the proposed method can maintain robust pose estimation in dynamic environments and achieve better error drift control.Secondly,a realistic positioning and monitoring platform for underwater optical wireless communication network is designed and implemented.At present,the underwater observation system in the industry lacks the ability of visual monitoring and feedback control of environmental conditions,and the ability of communication network node control and communication quality analysis for underwater optical wireless communication network.The platform designed in this paper is based on Web 3D rendering technology.According to the basic map data of the environment and the quality data of the deployment network,it supports the analysis and control of the communication environment and communication nodes of underwater optical wireless network with the change of node posture.Based on the typical Web engine,the platform can achieve good rendering capability and interactive experience.