Research And Implementation Of Channel Quality Assessment In UVLC System Based On Machine Vision

Underwater Visible Light Communication(UVLC)is considered to be a strong candidate for underwater wireless communication in the future because it can transmit data while taking into account lighting.In order to address the interference problem of time-varying channel to visible light communication system performance,the adaptive transmission mechanism can correspondingly change the transmission mode according to the obtained channel quality indexes,so as to maximize the transmission efficiency.However,it is difficult to deal with the complex and volatile underwater environment using only conventional adaptive transmission mechanisms with known signal distortion.There is a certain internal mechanism correlation between underwater optical imaging theory and underwater optical communication system,that is,the attenuation of light in the transmission medium not only affects the communication quality,but also affects the image quality.Therefore,this paper innovatively proposes to introduce dark channel prior technology in imaging theory to estimate channel quality indexes of underwater wireless optical channel,which can be directly estimated quickly and accurately according to the sampled underwater scenario images.However,the existing underwater dark channel prior technology still has some problems in dealing with the complex and diverse underwater environment,so it is an enormous challenge to construct a universal underwater channel estimation framework based on machine vision.This paper conducts in-depth research on the quality estimation of underwater visible light channel based on machine vision technology.The innovation of this paper is primarily reflected in the following two aspects:·A channel quality evaluation method based on fuzzy mean clustering(FCM)for underwater visible light communication system is proposed.The effect of channel quality estimation based on underwater dark channel prior technology depends mainly on the accuracy of estimation of background light and transmission map.In the complex underwater scene,there are some bright foreground objects,such as fish and other organisms.In the underwater dark channel theory,its high pixel value will interfere with the estimation of background light,thus affecting the final channel estimation effect.In this paper,based on the fuzzy mean clustering image segmentation method,the foreground interference object is separated from the area where the background light is located,and then the first 0.1%of the brightest pixel in the background light area is selected as the global background light,so as to avoid the influence of background light information from the foreground object,and effectively improve the accuracy of estimation.Experimental results show that this method can obtain higher estimation accuracy than traditional methods in typical underwater scene image processing.·A channel quality estimation method based on cycle-consistent generative adversarial network for underwater visible light system in dim background is proposed.For the existing underwater dark channel prior technology in the dark and unilluminated underwater scene,the introduction of artificial light source and the original extremely low pixel distribution will have a great impact on the channel estimation,so that the transmission mapping estimation obtained by it has a very high value that does not correspond to reality.Generative adversarial network is composed of generator and discriminator,and is trained by game theory so that generator can generate"false samples" matching with real sample data distribution.In view of the failure of existing theories in dim and unilluminated scenes,this paper designs an image cross-domain translation algorithm based on cycleconsistent generative adversarial network,and uses the above improved underwater dark channel algorithm to estimate the channel quality of translated images.The experimental results show that the proposed method can reflect that with the increase of the distance from the camera,the channel quality presents a trend of attenuation,to our best knowledge,which is not achieved by the existing dark channel algorithms.