Sub-Pizel Quantization by Applying Iterative Error Analysis Algorithm to Hyperion Data: Case Study of Northwest of Yun-Nan, China
The multiform earthsurfaces in Northwest of Yun-nan, which has complex condition of climate and geology, can not be accurately classified and identified by traditional multispectral image due to its resolution limitation in spectral and space, especially under the condition that precise geological survey, training set and rule base building are insufficient or absent. Hyperspectral image with plenty of continuous spectral information can overcome the limitation. In this paper the Hyperion image, a sort of hyperspectral image, makes good use of abundant correlatively spectral information and makes it possible to classify and efficiently distinguish the sub-pixel of earth surface in Northwest of Yun-nan. In order to extract sub-pixels automatically without any experiential knowledge, the article applies Iterative Error Analysis algorithm to hyperspectral image to obtain various sub-pixels spectral information in northwest of Yun-nan that can not easily be distinguished generally, such as road, jokul, sandstone etc. In the experiment, a series of preprocessing to the Hyperion image is firstly implemented and the sub-pixels are then extracted by iterative error analysis algorithm. Finally fractional abundance maps of mixed pixels are calculated and sub-pixels are quantified. Through our implementation, the purpose of quantization and identification of sub-pixels and unmixing of pixels in northwest of Yun-nan is achieved without any experiential knowledge. The paper illuminates that using iterative error analysis algorithm, the sub-pixel of complicated northwest of Yun-nans earth surface can be effective classified and identified unsupervisely with Hyperion data, and the experiment provides an ideal and practical method to detect sub-pixel automatically in the absence of experimental data.
Gao Wei Liu Tianle Liu Xiuguo F J Liu
国际会议
武汉
英文
127-131
2008-12-19（万方平台首次上网日期，不代表论文的发表时间）