Research And Implementation Of Underwater Intelligent Optical Wireless Communication Channel Monitoring Technology Based On Binocular Vision

Underwater optical wireless communication(UOWC)technology,due to its advantages of high speed,low latency,low power consumption,and high security,has great potential in applications such as underwater environmental monitoring,energy exploration,military communications,and is one of the key support technologies for realizing the future air space sea integrated communication network.Currently,research on underwater optical wireless communication mainly focuses on underwater complex channel modeling,optical transceiver acquisition alignment tracking(APT),and related modulation and coding techniques.Due to the weak diffraction effect of optical signals,underwater wireless optical communication mainly relies on direct LoS links.When the links are shielded by underwater obstacles,the communication performance will deteriorate dramatically.Currently,the occlusion problem of underwater wireless optical channels can only be solved through redundant channels from the network layer,that is,switching to the corresponding protection or recovery link immediately after occlusion occurs.However,real-time communication usually requires the network to have a strong self-healing ability and be able to automatically recover communication services in a short time.Therefore,research is conducted on how to predict the occlusion of underwater wireless optical links,It has important theoretical significance and application value.This paper creatively proposes an implementation scheme for underwater target detection and tracking based on binocular vision to address this issue.It improves the YOLOv5 target detection network framework to address the problems of underwater target detection difficulty,weak generalization ability,and data set scarcity.The main research content and innovation points of this article are as follows:(1)Aiming at the shortcomings of general target detectors in underwater applications,an improved underwater target detection method based on YOLOv5 was proposed.This method utilizes image style migration technology to expand underwater dataset samples and improve the generalization ability of the algorithm;Improve underwater target detection accuracy based on attention mechanism and detector head expansion.In this paper,the effects of various improved methods were verified through ablation experiments,and the effectiveness and feasibility of the algorithm framework was verified through testing the entire improved algorithm.The improved underwater target detection method proposed in the paper has significantly improved robustness and accuracy compared to the benchmark YOLOv5 model.(2)Aiming at the problem of link occlusion early warning in underwater optical wireless communication,an underwater channel environment monitoring system based on binocular vision,target detection and tracking was designed and constructed.The improved YOLOv5 and Deep SORT algorithms are used to detect and track underwater targets,and the PROSAC based ORB algorithm is used for stereo matching and threedimensional reconstruction.According to the system requirements and the distribution characteristics of ORB feature points,a method of adaptive border filtering feature points is proposed to filter redundant feature points within an object,greatly reducing the amount of matching computation and improving the efficiency of the algorithm.Finally,a series of demonstration experiments were completed on the platform,including target detection,tracking,occlusion judgment,and other tasks,verifying the feasibility and effectiveness of the system.