Research On Compression Algorithm For Time Interference Hyperspectral Data

Remote sensing, as the primary means of achieving earth observation, has grown from the initial stage of the panchromatic remote sensing to hyperspectral remote sensing phase after decades of development. As an important data source, the significant difference between interference hyperspectral data and other remote sensing image is the rich multi-dimensional spectral information in earth observation. It has been widely used in target recognition and other fields. However, with the continuous improvement of spectral resolution, the data obtained by imaging spectrometer improves with rapid expansion which brings enormous pressure on the storage and transmission of data. Therefore, the study of hyperspectral data compression and encoding algorithm has a very important and real theoretical value.This thesis is based on above background, according to large numbers of domestic and foreign literature to determine the feasibility of the system as well as data compression ratio as the main research directions. It includes interference hyperspectral data preprocessing and lossless compression as two parts:(1) Considering the feature of oversampling in Hyperspectral data pre-processing section, extraction has been done to reduce the amount of data. Firstly, in order to avoid spectral leakage, a window function is commonly used for performance in contrast, select the most appropriate spectral characteristics of data over a window function; Secondly, considering the data extraction efficiency by using a polyphase filter technology, the original filter into a plurality of sub-filters to improve the processing speed; Finally, to increase board practicality of implementation, the multi-level hyperspectral data extraction is used to solve the problem caused by the data amount of the processing time.(2) For interference lossless compression part, in order to provide the basic theory of hyperspectral compression, first give spatial and spectral correlation analyzes of hyperspectral data. Spatial correlation analysis showed that although each band has a certain spatial correlation, but lower than the ordinary natural image. Therefore, the common nature image compression method couldn’t be used directly with hyperspectral data; spectral correlation analysis of hyperspectral images gives the results of the spectral correlation is much stronger than the spatial correlation. More attraction should be focused on the redundancy of hyperspectral imaging spectrometer direction. Based on the correlation analysis of hyperspectral data, the thesis proposed a new algorithm which combines the best DPCM spectral prediction and the idea of LOCO-3D algorithm, which comes from DPCM and JPEG-LS prediction method and 3D extended form of LOCO-Ⅰ algorithm.Through analyzing the process of the algorithms, the thesis introduced It has greater improvement in compression ratio and computational complexity aspects, also has stronger board feasibility.