The Study Of Sampling And Storage Methods Based On Compression Sensing

With the continuous development of information technology,how to deal with information data has become a hot topic.The theory of compression perception breaks the shannon sampling theorem and provides a new direction for information processing.The increasing amount of signal is also putting a lot of pressure on the application of compressed sensing.In addition,the signal characteristics are not evenly distributed in the signal,how to solve the problem of uneven distribution of signal characteristics while considering the signal characteristics,and also bring new research directions for the study of compression sensing.This paper mainly studies the application of compression sensing theory in image processing and data storage of wireless sensor network,The main contents include the following three aspects:(1)This paper summarizes the traditional theory of compression perception and summarizes the advantages and differences of compression perception compared with traditional compression methods.And the existing some of the method of block compression sensing in the field of image is introduced,and the evaluation standard of image reconstruction quality is given.At the same time,the representative algorithm of compression sensing in sensor network data storage is described.(2)A new quadratic block compression sampling method based on significance is proposed.The method firstly preprocesses the image to obtain the image significance information.Based on the obtained significant information,the block is further aggregated on the basis of the conventional image uniformly blocking compression sensing,and the similar blocks are clustered together to balance the distribution of the feature of image block and optimizes the allocation of the sampling rate.The experimental results show that the proposed method under the same sampling value,compared with other methods,improve the image reconstruction quality,make the PSNR,MSSIM and other evaluation indicators of reconstruction image are obvious higher than the traditional algorithms,while the visual effect of the important areas are looks better.(3)A p-probability random walk data storage strategy based on compressed sensing for wireless sensor network is proposed.This strategy combines the compression sensor theory with the random walk model,and changes the node storage structure,which makes up for the shortcomings of nodes in the previous method that can not deal with specific small areas alone.The method first initializes each node to form a individual date packet and a storage packet.Then,in the coding and storage phase,each source node transmits data to its neighbors with p probability,and then stops the data transmitting by t steps.After each step,the node receiving the data stores the data,and then transmites the received data to the neighbor with p probability;Finally,in the reconstruction phase,the sampled node,codes data in the query range and upload to mobile sink node.The experimental results show that the algorithm proposed in this paper reduces the error of the reconstructed data of the specified area,and greatly reduces the amount of data generated for reconstruction.