Research On Color Visuralization Of Hyperspectral Image

Hyperspectral Image(HSI)is a general trend of the development of remote sensing technology application,and relevant study has attracted attention of scholars worldwide with its wide application in various fields.Therefore,requirements on HSI application effect have become increasingly strict.Visualization technology,as a bridge between data and observers,plays a key role in the application and development of HSI.HSI color visualization technology is an intuitional presentation method for HSI,it is based on the human visual characteristics,uses color spaces to represent the abundant spatial and spectral information contained in images,and helps observers to understand and process relevant useful information more clearly and accurately.Such visualization method that changes data information to image information is of great significance to scientific decision-making and information utilization.First of all,this study will start from the research background and status quo of HSI visualization,it will introduce HSI’s characteristics and common display methods,sort out and classify the existing visualization methods and standards,and unify the former scattered theoretical systems of HSI researches while standardizing some commonly used design purposes.Then,this study will discuss some commonly used HSI analysis methods,propose a low-complexity distance calculating method and apply it to spectral unmixing,this method will be based on LSMM geometric method to solve the high-complexity drawback of the existing algorithms.After that,based on the existing HSI analysis methods,HSI visualization technologies will be divided into two categories including a visualization method based on pixels analysis results and a visualization method based on raw HSI with respective descriptions based on data features.These two categories of visualization methods will be summarized as follows:For one thing,the visualization method based on pixels analysis results.This is a visualization method that will first use the existing HSI pixel analysis methods to process and analyze the original HSI and then display its analysis results.In accordance with different analysis methods,this category will be divided into two conditions of a visualization method based on hard classification results and a visualization method based on soft classification results.Based on the data characteristics of HSI hard classification results,this study will.propose an automatic color distribution method based on spatial correlation,with its output image focusing more on displaying inter-class separability.And,based on soft classification,this study will propose an HSI visualization method with distance preserving property.This method will be based on spectrum linear mixing model and color space linear mixing model,then follow the human visual characteristics,and will guarantee good visual recognition between different categories and between mixed pixels of the same category.In addition,for the analysis results of some special pixels,it will be possible to follow the above two methods to make corresponding adjustments in accordance with the characteristics of data analysis method such as HSI visualization method based on sparse representation.For another,visualization method based on raw HSI.This is a visualization method that will use different data analysis or processing methods to process original HSI and simultaneously display them Three visualization methods are proposed in this situation.The first one,a color dynamic HSI visualization method which will be based on CMF method.It use dynamic colors to express ground objects so that output images will have strong visual effects.This method will not only have easy calculation but also meet real-time requirements,and will also generate images close to true color images.The second one,for the single-mode and multimode images with multiple images.This study will propose an interactive color visualization method that suits HSI.This methods will use interactive method to mix multiple images to output more useful information,and provide color display of HSI statically or dynamically in the same interface to achieve HSI information exploration with purposes.The third one,this study will also design a Class-Orieted Visualization Method for Hyperspectral Imagery which will be based on manifold alignment and combine human visual sensory characteristics,and use category information and supervised information to complete HSI visualization so that the generated images will make full use of supervised information and pixels of the same category will have good distance retention characterstics.Finally,with the purpose of increasingly accurate and strict display and application requirements at present with output images displaying objects’ spatial and spectral information simultaneously,an HSI multi-level color display method will be proposed.This study will first use a visual attention model to do priority classification of objects in raw image,then follow the obtained different priority to determine the area’s processing method,and further use the previously described one or more visualization methods as required to achieve display of Regions of Interest(ROI)in macro,local,and micro levels through multi-display policy.The display result of this method will meet the requirement that spatial and spectral information display in the same image,and will highlight certain category while keeping the other display characteristics unchanged.