Research On Distributed Video Coding Based On CS In Coal Mine Underground

With the continuous improvement of the information level in coal mine underground,the interaction information between people,machines and environment collected by sensor nodes is increasing.How to efficiently process massive data information and reduce the energy consumption of sensors in coal mines with limited resources has become a hot issue.The distributed video coding framework adopts the codec mode with simple coding and complex decoding method,which transfers the complexity of the encoding end to the decoding end and is suitable for use in wireless sensor networks with limited resources.The distributed compressed video sensing technology with compressed sensing added to the distributed video codec framework further improves the compression performance of the system.In order to obtain a more sparse representation and more general applicability of the video signal,the sparse basis strategy is improved;In order to generate more accurate side information and improve the reconstruction quality of the non-key frame,the weight of the side information generated by the motion estimation method is adaptively selected;In order to select a more representative key frame in the video sequence to further improve the reconstruction quality,the frame group structure of the video sequence is studied.The specific research contents are as follows:(1)In order to solve the sparse representation problem to obtain a more sparse representation of the video frame signal,a hybrid sparse strategy combining linear DCT basis and block prediction basis is proposed to solve the problem of sparse representation.First,the video frame is subjected to equal large non-overlapping block processing,and is divided into different types of blocks by using efficient classification decision criteria,the appropriate sparse basis strategy is adaptively selected according to the result of the block classification.This sparse representation scheme can achieve more sparse representation and more general applicability of the video signal while reducing complexity.(2)In order to generate more accurate side information to improve the quality of non-key frames,an adaptive weighted side information generation scheme based on block classification is proposed according to the correlation between different blocks of video frames.First,at the decoding end,different types of block classification are performed by using correlation between different blocks in the two adjacent decoding key frames,and different weights are selected according to different types of blockclassification,and two adjacent decoding key frames are used to generate side information.The side information generation scheme proposed in this paper makes full use of the correlation between different blocks of video frames,which makes the generated side information more accurate,and the reconstruction quality of non-key frames is further improved.(3)Key frames play a crucial role in the recovery of the entire video sequence,and the frame group structure determines the selection of key frames.In order to improve the representativeness of the first frame keyframe in the GOP group,an adaptive GOP grouping strategy is proposed.According to the severity of motion between adjacent frames to perform block classification,calculate the difference between the number of different blocks and compare it with the preset threshold,and determine the grouping relationship between the current frame and the reference frame according to the different application scenarios.At the same time of adaptive grouping,in order to balance the simple design intention of the encoding end and the reconstruction accuracy,the maximum value of the GOP packet length is limited.The performance of the proposed algorithm is verified by comparing the adaptive GOP grouping algorithm and fixed GOP grouping algorithm based on different application scenarios.