Research Of Spectral Unmixing Technique And Its Application In The Oil Quantity Analysis

With the mass spectral information, hyperspectral remote sensing technology plays a huge role in geological exploration, military applications, detection of vegetation, ocean remote sensing and other areas. However, due to the limitations of spatial resolution and the complexity and diversity of the various types of surface features, images, in addition to pure pixel, there may exist a mixture that contains a variety of surface features pixel, and thus can not achieve the accuracy of remote sensing applications.Spectral unmixing technique is to use the high spectral band information from the pixel level to enter the sub-pixel level, to decompose the basic unit-endmember of the mixed pixel, and to calculate the proportion of these endmembers. By the pretreatment of spectral unmixing, object classification and recognition is more accurate, so there is a broad space for mineral prediction, urban layout analysis, chemical composition and detection. This article attempts in the study of marine oil spill detection using spectral unmixing techniques, to analyze oil and water content of the hyperspectral oil spill image, to improve the accuracy of oil spill detection.Paper based on the mechanism of hyperspectral mixed pixel and linear spectral mixing model, using endmember distribution spatial correlation and endmember spectra of non-negative and continuity constraints, to improve the spectral unmixing technical accuracy. Researches include the following aspects:Firstly, analysising of hyperspectral unmixing process and a variety of endmember extraction methods, proposing the existing N-FINDR endmember selection algorithm to process the hyperspectral oil spill image, and using the correlation between endmembers to improve the N-FINDR and to achieve accurate extraction of endmembers and reduce computation.Secondly, create mixed-pixel image to verify the improved N-FINDR. Use the spectral curves of real water and gasoline to produce the mixed image of the hyperspectral, and then determining the proportion of each endmember by spectral decomposition with the improved N-FINDR. In the end, get the content of oil in the image.At last, select the actual hyperspectral oil spill image, for image preprocessing as well as the main ingredient to reduce the dimension. And then use the improved N-FINDR to achieve mixed pixel decomposition, and used for the oil analysis of the actual image.The results of simulated data and actual test data verified the algorithm proposed by this paper is effective. The research of this paper also shows that the using of spectral unmixing to get the endmembers'content is feasible for the using of oil analysis.