Research On Vegetation Classification And Extraction Technology Of Hyperspectral Image

Classifying and extracting vegetation using hyperspectral image is an important aspect of vegetation hyperspectral remote sensing application studies.It is difficult to meet the needs of practical application when classifying vegetation with traditional classification algorithms.Therefore,the application of hyperspectral vegetation indices and multiple kernel learning methods are studied systematically for vegetation classification and extraction in this dissertation.Vegetation classification method based on weighted hyperspectral vegetation index and vegetation fine extraction method based on multiple kernel support vector machine are proposed respectively with theoretical analysis and experimental validation.The major works implemented and the goals achieved are listed as follow:1.Firstly,the significance and research contents of vegetation RS and the advances in hyperspectral RS technologies are illustrated.Then analysis of the methods of hyperspectral image vegetation classification and extraction is given,and several difficulties which need to be resolved in its application are summarized.2.The vegetation spectral characteristics are studied based on the spectral curve and spectral transform.Bands are selected based on the spectral characteristic of vegetation with hyperspectral data bands response of vegetation.Experimental results have shown that this method can effectively reduce the dimension of hyperspectral data and maximize the retention of vegetation spectral characteristic.3.The construction principal of broad and hyperspectral vegetation indices is analyzed,and a weighted hyperspectral vegetation index is designed to reduce the impact of soil and enhance the adaptation of vegetation index on hyperspectral data.The experimental result has showed that the weighted hyperspectral vegetation index can extract vegetation information from hyperspectral image effectively.A vegetation classification and extraction method based on weighted hyperspectral vegetation index is proposed to reduce the impact of non-vegetation in the processing of vegetation classification.To start with,weighted hyperspectral vegetation index is used to extract vegetation from hyperspectral image in this method,and then SVM classification algorithm is carried out to classify vegetation.Experimental results have shown that this method can reduce the impact of non-vegetation on vegetation classification,and improve the overall classification and extraction accuracy.4.The principles of nuclear methods and multiple kernel learning are analyzed,and it has been pointed out that it has some the disadvantages of single kernel method when dealing with all samples in complicated conditions,while the multiple kernel method can get better treatment results.Vegetation classification algorithms based on multiple kernel SVM is proposed as the characteristics of vegetation samples distribution are concerned Experimental results have shown that this method can improve the accuracy of fine vegetation classification and extraction compared with SVM and other traditional classification methods.