Vegetation Classification Of The Basin Of Shule River Based On NDVI Time Series

Because of different influence of climatic conditions and human activities, vegetation formed different surface ecological covered landscape in the different regions.Also, vegetation have some feedbacks on the climate and human activities by the selectivity nutrient uptake, differences between the photosynthesis and respiration features. Vegetation type research is an important part in the study of land surface process model. Impedance inversion of surface, surface roughness, leaf area index and other parameters are important parameters of model requires,which was producted by vegetation type indirectly. Vegetation play an important role in the global and regional material circulation and energy cycle. And the vegetation type is related to the human society resource utilization, environmental comfort of life, food production security and other issues, so the vegetation type research has always been a hot spot of research.In this paper, the main research method is based on vegetation index of time series data.The main idea is different vegetation types have different phenological periods and characteristics.the phenological characteristics differences can becomethe difference of NDVI curve, including time difference and value differences, if they are extracted from these differences, it can be used as an effective basis to the vegetation classification. In this paper, the vegetation classification research is in Shule River basin.I using the image data of MODIS products,which are time-series NDVI images,and a total of23scene. The time span is one year of2013.This product is generated with maximum synthesis, which have time resolution of16days after treatment. After denoising by refactoring (such as minimum value alternative, hants filtering method), I extract12feature band such as the amplitude, phase, NDVI net cumulants, spring season, the first principal component. After correlation analysis, five band with information redundancy are found. Using the field survey of quadrat point as the training data, I study the components of vegetation types with See5.0automatic extraction method of decision tree. I using the data of and landcover products and vegetation map products as the verified classified which is made of the Chinese academy of sciences.The main research contents:the comparison between the existing vegetation type classification system; Time series data refactoring; To extract features from the time-series NDVI curve wave band method are studied; The applied research to extract the classification based on decision tree classification and components.Research conclusion of this paper:1, the NDVI data of time-series is an effective data to vegetation classification. Whether the time-series NDVI as classification data directly or used to extract the band as a categorical data, our predecessors have done relevant research. In this paper,I extracted12characteristic bands tostudy the classification of vegetation and good results have been achieved. Based on time-series data of vegetation remote sensing research has become a research hotspot. 2, Noise of time-series data can not purify with the method of single one.Because of likely to be noise pollution and the source of the noise is different,we should be combined through a variety of denoising method.In this paper, I use three ways to remove the noise,which are maximum synthesis, the minimum replacement and HANTS filtering processing.3, Shule River basin has single vegetation type.The main vegetation types:t he desert vegetation,desert grassland, alpine meadow grassland, artificial vegetati on。 The main formation: Alhagi formation, Lycium ruthenicum Murr formation, Achnatherum splendens formation, Tamarix formation, populus simonii formatio n, Phragmites formation, vegetation coverage of Research area is as follows:de sert grassland accounts for17.1%of the area in the study area, alpine meadow and steppe24.8%, swamp and aquatic vegetation, temperate deciduous broad-1eaved forest area are below1%, mountain and valley thickets,3.5%of cultivat ed land area of about2%, the rest without vegetation area area of more than50%.