Research On Redundant Data Removing Algorithms Of Remote Sensing Images

With the rapid development of the space remote sensing technology, remote sensing images, which have broadly applied in multiple areas, such like weather forecast, resource survey, environmental monitoring, disaster prevention and military reconnaissance, are playing an important role in national production and national security. The remote sensing image processing technology is not only one of the key technologies for promoting the development of space remote sensing, but also an effective method to improve the effectiveness and interpretation of the remote sensing data. It has become the research frontier and hotspot in the fields of remote sensing science and technology.This thesis discusses several redundant data removing algorithms of remote sensing images and focused on the compression technology for reducing remote sensing data, the cloud detection technology for removing the redundant data, and the image fusion technology for fusing the spectral and spatial information. The main innovation points of this thesis can be concluded as following:1. By analyzing the statistical characteristics and spatial characteristics of the multi-spectral images, this thesis proposes an image registration method which combines the phase correlation and the affine transformation and effectively improves the correlation of the images. With regard to the multi-spectral image compression, this thesis also proposes a two-dimensional wavelet coding method which combines KLT spectral correlation and low complexity. Shown by the simulation, the proposed method increases the PSNR value with an average of2.0dB comparing to the JPEG2000spectral image compression method. Under the scenario of implementing on orbit, the proposed method can also increase the PSNR value with an average of1.7dB higher than using the JPEG2000when the quantitative processing is operated on satellite and the float point processing is operated by ground-based computers.2. In the visible light image detection technology, by analyzing the feature of the thick cloud images, the main characteristics of the cloud can be distinguished from the object on ground, such like luminance and texture. Based on these characteristics, three features with large differentiation degree can be distinguished by experiments, which are the bright pixel ratio, the average gradient, and the two moments angle of the gray level co-occurrence matrix. Using the pattern recognition tool which supports the SVM, the related data can be received and the categorizer can be formed by performing large sample-amount research. Operating the cloud detection simulation on one track of remote sensing image with the categorizer, it can be indicated that the proposed method is able to perform great cloud detection, remove the redundant data, and satisfy the desire of the engineering application.3. Fusion of panchromatic image and multi-spectral image is one of the key technologies in remote sensing image fusion. Based on the analysis of the main fusion methods and fusion rules, this thesis proposes a Curvelet transformation-based fusion algorism. This algorism combines the advantages of the HIS transformation algorism which is able to retain great spectral characteristics and the Curvelet transformation which is able to retain great spatial information. At the meantime, we compare the methods based on5typical evaluation criteria. With regard to the spatial information retaining, it indicates that the entropy values of each spectrum of the proposed method are higher than the other three methods and the average gradient value is only a little lower than the wavelet fusion algorithm. With regard to the spectral similarity, the U1QI value and the ERGAS value of the proposed method are higher than the ones of the other three methods, but the similarity coefficient is a litter lower than the wavelet fusion algorithm. Overall, the proposed method is superior to the HIS fusion algorithm, the wavelet fusion algorithm, and the PCA fusion algorithm. Since this panchromatic image and multi-spectral image fusion technology is used to fusion multi-images into one image, from a certain perspective, this method is also an effective redundant data removing technology of remote sensing images.