Research And Application Of Secure And High-Efficient Video Transmission Technology Based On Deep Compressed Sensing

With the rapid development of new sensors,high-performance embedded processors,5G communications,edge computing and deep learning,some new intelligence paradigms have emerged in the field of wireless sensor networks(WSNs).Among them,wireless multimedia sensor networks(WMSNs)is a typical representative of new intelligent WSNs.Compared with traditional WSNs,WMSNs can collect more dimensions and larger volumes of data,thus providing a spatial perspective that traditional WSNs lack,which enables WMSNs to penetrate into emerging fields such as intelligent transportation and intelligent security.As a new generation of intelligent WSNs,the data transmission of WMSNs is faced with severe technical challenges:the efficiency of highthroughput video data transmission and the security and privacy protection of visually sensitive multimedia data.In view of the above technical problems,based on the theory of deep compressed sensing and chaos theory,this paper carries out relevant research on the video encoding and decoding method based on deep compressed sensing and the secure and efficient transmission of video data in WMSNs.The main research achievements and innovations of this paper are as follows.(1)New high-efficient video transmission technology based on deep compressed sensingIn the research of WMSNs data transmission,the two most noteworthy technical problems are efficient video coding and low delay decoding methods respectively.On the one hand,resource-constrained wireless sensor nodes cannot encode the original video signals with high complexity for a long time.On the other hand,the traditional compressed sensing algorithm has limited ability to recover large images and cannot meet the time efficiency requirements of video surveillance system.Therefore,this paper proposes a new video transmission technology,which mainly includes a video sampling method based on block compressed sensing and a video reconstruction method based on deep compressed sensing.At the encoding terminal,the proposed sampling method can effectively reduce information redundancy within and between video frames,and reduce the fluctuation of image quality between adjacent recovery frames.On the decoding side,this paper proposes a convolution neural network(BD-FRN)model based on the traditional fast iterative shrinkage thresholding algorithm,and uses this model to reconstruct video frames in parallel,which can significantly improve the recovery speed of video data.The simulation results show that compared with some existing video transmission schemes,the proposed technology has more advantages in terms of reconstruction effect and operation efficiency under the same sampling rate.(2)Secure and high-efficient video transmission system based on chaotic compressed sensingTo enhance the security of data transmission in WMSNs,a lightweight encryption method based on pixel block is proposed.This method combines chaotic system with traditional lightweight image encryption algorithm to encrypt the sampled video data twice,aiming to ensure the integrity of data during transmission and make up for the shortcoming of traditional compressed sensing security.Further,the application of the new video transmission technology of deep compression sensing in WMSNs is studied,and the secure and efficient transmission system architecture of chaotic compressed sensing video is designed.Finally,the efficiency and security of the proposed system are verified by analyzing the key space size,image visual features,gray histogram,image entropy and correlation between adjacent pixels.