Study On The Yellow River Delta Wetland Typical Vegetation Using Hyperspectral Remote Sensing

Feature extraction and classification play an important role in the research field of hyperspectral remote sensing application。Because coastal wetland is vast in area and complex in distribution, and hyperspectral remote sensing data is high in precision and great in data size, for the traditional feature extraction method, it is difficult to make full use of the spectral information in hyperspectral data, neither to extract the potential characteristics out of the expert knowledge and statistical information, it is difficult to develop a high precision classification algorithm which is apply to the coastal wetlands classification on hyperspectral remote sensing data. This situation dose not help the development of hyperspectral remote sensing technology in coastal wetland region.In order to reach the precise monitoring requirements on coastal-wetland vegetations using hyperspectral remote sensing data, the Yellow River Delta wetland is selected as the study area in this paper, based on the technology of Data Mining, hyperspectral remote sensing image and typical vegetation field spectral data are used for developing an method for extracting hyperspectral features of typical coastal-wetland vegetation and developing classification method, to achieve the goal of extracting and classifying the typical vegetation in our study area precisely. The specific contents are as follows.(a) Developing an analysis work on the field spectral features based on the separability of spectrum and the seasonal differences between the spectral featuresAccording to the characteristics of the study area and hyperspectral remote sensing data, the field spectrum of the typical vegetation in the Yellow River Delta coastal wetland is measured in order to get a representative data of the typical vegetation spectral features in study area; baced on reflectance spectrum data, the typical vegetation spectrum analyis. and feature extraction are executed. For analyzing the separability of the different vegetation spectrum, a characteristic spectruml extraction method based on Continuum removal and spectral diffrencefrom is developed, the object of which is to get a lookup table on the spectrum band separability of the typical vegetation. For analyzing the spectral feature differences between different vegetation in different season, a comparison work is performed on typical vegetation spectral features in both spring and autumn seasons based on the derivative transformation method, in order to get the differences on the position information and spectral reflectance of4kinds of spectral features, such as green light region reflectance peak, red light region absorption peak, the red edge and near infrared region reflectance peak.(b) Developing a technology of typical vegetation spectral features extracting in the research region based on data miningBased on the PROBA CHRIS multi-angle hyperspectral remote sensing image, firstly we study the preprocessing technology of the image data, and analyze the imaging effect and classification ability of different-angle images, make sure that0°angle image as the data source of feature extraction. On this basis, in order to obtain the typical vegetation spectral features and the combination rule of spectrum band in CHRIS hyperspectral remote sensing images, and then classify vegetation types in the study area, we develope a hyperspectral remote sensing feature extraction technology of coastal wetlands typical vegetation based on association mining. This technology is applied on mining association rules from the hyperspectral remote sensing data of study area, and based on the quantitative extraction rules, develop the hyperspectral remote sensing feature set of typical vegetation in the Yellow River Delta.(c) Developing a hyperspectral remote sensing classification method on the typical vegetation in the Yellow River Delta based on decision treeBased on hyperspectral remote sensing feature sets of the typical vegetation in the Yellow River Delta, combined with information on the field spectral features, the classification rule of typical vegetation in study area is extracted, and then a remote sensing classification method on typical vegetation in the Yellow River Delta is developed based on decision tree. The classification experiment on typical vegetation is done using hyperspectral remote sensing image covering the study area, and the field information and spectrum data obtained in the field reconnaissance are utilized to evaluate the validation and accuracy of classification results.The result shows that, the classification accuracy of the classification method developed in this paper is Satisfying.