Research On Color Visualization Of Hyperspectral Imagery Based On Isometric Mapping

With the continuous progress of remote sensing technology,hyperspectral remote sensing technology has been rapidly developed and applied in many fields.At the same time,people's expectation of display for hyperspectral are higher and higher.Under the current technical conditions,although the spatial resolution has made great strides,it has not been able to solve the problem of the heterogeneity existing in the same pixel.For any pixel,past classification methods can only classify pixels into a certain category.Sub-pixel mapping and other methods can achieve finer classification results,but for a wide range of mixed pixels can only get a clear distinction between each type of hyperspectral imagery,which is not ideal to depict the details of the distribution of each other,and this situation is ubiquitous in practical application.Although these methods have been continuously improved,the expression of hyperspectral imagery information is still inadequate.Hyperspectral imagery color visualization is a visual representation of hyperspectral images.Based on human visual characteristics,the use of color space to represent the abundant spatial and spectral information contained in an image enables the viewer to understand more clearly and accurately useful information related to handling.This visualization method of transforming abstract data information into concrete image information is of great significance for scientific decision-making and information utilization.In this paper,we focus on the hyperspectral image visualization technology,based on the isometric feature mapping algorithm to visualize the hyperspectral imagery,retain as much as possible the information contained in the original hyperspectral data,maintain the spectral distance between the pixels,combined with the color space the principle of colorimetry shows the differences in spectra between the pixels in a way that meets the human visual characteristics.The color differences of the pixels are characterized by the change of color,which is more in line with the actual situation of the objects' variety.In view of improving the visualization accuracy,a new visualization method is proposed based on the characteristics of hyperspectral images.The main research contents are as follows:Firstly,the important role of hyperspectral imagery in the field of remote sensing technology and the background and significance of the research are expounded.The current research status of hyperspectral imagery visualization technology is introduced in detail.The deficiencies of all kinds of visualization methods are pointed out.Secondly,the imaging theories of hyperspectral imagery and their data features are analyzed.The principle and process of the three classic visualization methods are presented in detail,in order to display the basic principles of visualization.Thirdly,the basic principles and implementation steps of isometric mapping(ISOMAP)algorithm are studied.In order to overcome the shortcomings of L-ISOMAP in visualization,a KL-ISOMAP algorithm is proposed in this paper to reduce the distortion of low-dimensional embedding reconstruction of hyperspectral imagery in the traditional L-ISOMAP algorithm.Improving the performance of L-ISOMAP when it comes to visualization from both visual accuracy and operational efficiency.Finally,two visualization algorithms for hyperspectral imagery based on the combination of spatial texture information and spectral information are proposed in order to improve the discrimination between pixels to achieve better visualization.Firstly,the spatial texture information is extracted respectively by Gabor filterbank and gray level co-occurrence matrix,and then is concatenated with the spectral information.Then,the KL-ISOMAP algorithm is used to reduce the dimension of hyperspectral to realize visualization.With the help of spatial information,the geodesic distance between pixels is corrected to enhance the visual effect.