Location:Home > Engineering science > Information and Communication Engineering > Research on Multiscale Analysis-Based Target Detection in Hyperspectral Imagery

Research on Multiscale Analysis-Based Target Detection in Hyperspectral Imagery

Downloads: []
Tutor: ZhangJunPing
School: Harbin Institute of Technology
Course: Information and Communication Engineering
Keywords: Hyperspectral images,Target detection,Multi-scale analysis,Curvelet transform
CLC: TP391.41
Type: Master's thesis
Year:  2007
Facebook Google+ Email Gmail Evernote LinkedIn Twitter Addthis

not access Image Error Other errors

With the development of imaging spectroscopy, hyperspectral image has entered into the stage of effective treatment and use of hyperspectral data acquired. Hyperspectral images with high spectral resolution, narrow band width, the large amount of information to the high spectral diagnostic ability to distinguish and detect surface features of the target, therefore hyperspectral image target detection research has been widely appreciated. However, the large amount of data, data dimensional high target small to detect a lot of difficulties, some of the traditional target detection methods have been unable to achieve good results, it is necessary to study the new hyperspectral image target detection algorithm. In this context, the subject the following aspects. First, the characteristics of hyperspectral image target detection method. After the analysis of the characteristics of spectral resolution, the correlation between the spatial correlation and band, using feature extraction data dimensionality reduction, the specific use of principal component analysis and independent component analysis. Then introduced the RX detection algorithm, and the ROI extracted based on the higher moments and principal component target feature to select two improvement measures to further improve the detection performance. Secondly, multi-scale geometric analysis application in target detection, hyperspectral image target detection method based on Curvelet transform. Curvelet-represented by multi-scale geometric analysis method has good directivity, fast convergence and expression of the sparsity than wavelet transform is more suitable for processing image signals. The thesis elaborated Curvelet Transform and its implementation, focusing on research with the method of the Curvelet Transform enhanced hyperspectral image target feature, and by comparing the experimental validation of the effectiveness and superiority of the algorithm. Finally, multi-resolution analysis method to study the spatial, spectral and radiometric resolution for target detection. Through the introduction of hyperspectral imaging spectrometer can learn Institute with the image of the imaging principle, process and parameters. Then separately proposed multi-resolution analysis of the spatial, spectral and radiation, in order to obtain high resolution spectral image. The test results showed that the resolution be detected to produce different degrees of impact, based on the need to determine the appropriate resolution to achieve a good compromise between the detection effect and imaging costs.
Related Dissertations
Last updated
Sponsored Links
Home |About Us| Contact Us| Feedback| Privacy | copyright | Back to top