Multi / Hyperspectral Remote Sensing Images, Spectral Decomposition Of The Research And Application

Due to the limit of spatial resolution in remote sensing images, very often, one pixel may cover hundreds of square meters with several ground objects, and becames a mixed pixel. In this situation the scattered energy is a mixing of the endmember spectra. By decomposing mixed pixel to typical ground objects (endmembers) in fractional proportions (abundances), we can acquire information in sub-pixel level to improve the accuracy of ground object recognition, and to realize the quantitative analysis of remote sensing images. So the decomposition of mixed pixels is very meanful for the ground object classification with high accuracy and ground target detection in the multi-spectral and hyper-spectral remote sensing images, and it has become a hot research topic in the remote sensing area in these years. Focusing on the aforesaid issue, the article has made a lot of research, and the innovations are as follows:1. A new algorithm for estimating the distribution of impervious surface with remote sensing images is proposed in this paper. As the endmember spectral variability of urban areas results in a poor decomposition accuracy, a brightness normalization method is applied to reduce brightness variation. In consideration of the characteristic that various land covers possess of different absorb and reflect ratio, we apply the land surface temperature (LST abtrained from the 6th band of Landsat TM/ETM+data) to eliminate the pervious surface components in the impervious surface component. The land surface temperature assistant normalized pixel purity index (T-NPPI) method performances an ideal accurancy in extracting the impervious surface distribution for urban areas.2. This paper proposes a new method based on Fisher discriminant null space (FDNS) for the decomposition of mixed pixels in hyperspectral imagery. Traditional Spectral mixture analysis assumes that each endmember must have a constant spectral signature, however, the endmember signatures are always variable. In order to restrain the effect of endmember spectral variability, FDNS searches a linear transformation of the spectrum, which makes that endmember spectra have no variability inside each endmember group but large differences among different endmember groups. As the negative impact caused by endmember spectral variability on unmixng accuracy can be decreased to a larger extent by using the transformed spectra, FDNS has a promising accuracy for the decomposition of mixed pixels in hyperspectral imagery. 3. This paper applied the linear spectral mixture analysis (LSMA) technique to moderate-resolution imaging spectroradiometer (MODIS) remote sensing data which has been sparely seen in mixed pixel decomposition research presently. In order to achieve the appropriate endmeber spectra, PPI was used in high purity pixels extraction, then the mean spectra of the choosed pixels were substituted for the endmembers to unmixing the MODIS data by fully constrained least squares (FCLS). Results show that MODIS data can be effectively applied to remote sensing dynamic monitering and land cover classification.