Nagqu Grassland Spectral Analysis And Recognition Based On Remote Sensing Technology

With the development of economy and technology, the national macro decision-making, resource survey, environment and disaster monitoring and so on.The areas affecting the development of national economy needs data support, and require the data on the macroscopic, continuous availability in time. Remote sensing technology have more ability, so closely watched. Remote sensing, geographic information system and global positioning system (GPS), has been more and more used in our daily life.This study measured the spectra of typical grassland community in Nagqu area. According to the spectral reflectance curve of Nagqu grassland types of Kobresia humilis, Kobresia pygmaea, Kobresia littledalei, Stipa purpurea, based on the first order derivative method and continuum removal and spectral feature extraction methods on grassland spectral feature extraction, the same plant in different coverage and the same coverage under different vegetation of grassland spectra features. At the same time the use of hyperspectral data environment of small satellite; grassland type RapidEye satellite image in experimentation area were classified, depending on the texture characteristics of high spatial resolution satellite, which greatly improved the accuracy of classification. This research work for the physical and chemical properties, in-depth study of alpine vegetation coverage, vegetation identification, classification and guide the satellite remote sensing monitoring of grassland degradation, remote sensing inversion model establishment, forage evaluation has certain reference significance. The results show that:(1) This research in typical grassland communities in naqu region the measured spectrum. Typical grass type Kobresia humilis to naqu area, northern song grass grass, small song, purple flower pin thatched spectral reflection curve, and by applying the method of first derivative spectral feature extraction methods such as envelope to division of spectrum feature extraction, the grass are the same in different vegetation coverage and under the same grass spectral characteristics of different vegetation coverage. Also used RapidEye satellite image classifying experimental area of grassland types, depending on the high spatial resolution satellite texture features, greatly enhance different vegetation canopy spectroscopy has special curve, spectral reflectance in visible light band is purple flower s. grandis grass, small song grass, northern song, the near infrared band reflectance spectrum is, in turn, small song esparto grass, grass northern song and purple flower; Northern song grass red edge position can be recognized, but not able to distinguish between small song esparto grass and purple flower; Visible light wave band spectral reflection curve characteristics can be used to identify the alfalfa stipas and red edge position identifiable northern song of grass, and little song esparto grass and purple flower, can’t rely on red edge position for identification.(2) The vigorous growth of Kobresia in550nm and719nm derivative are up to a maximum value, the fastest growth rate of original reflectance, reflectance value increases with increasing coverage, a derivative of red edge slope increases with coverage; with the leaves yellow, Kobresia growth in "green peak" and the red edge spectral reflectance is exuberant period decreased, a "blue shift of red edge, green peak decline phenomenon" (3) In the infrared band at low coverage, canopy spectra with spectra characteristics of soil, moisture absorption Valley features are not clear, high coverage is on the contrary, the visible spectral reflectance increases with increasing coverage. With the development of Kobresia into recession, all fell in the "spectral reflectance of green peak" and red edge position, the visible band Kobresia canopy and the different stages of the chlorophyll content is directly related.(4) Combination of spectral noise reduction and traditional multi-spectral classification method is feasible, and the classification accuracy is higher than the classification accuracy depends only on the spectral characteristics.(5) Through the spectral noise reduction and traditional multi-spectral classification method to combine with its feasibility and universality; but in the hyperspectral spectral classification algorithm, only to rely on the spectral features of hyperspectral data, is far from a high accuracy; should consider adding other information based on the spectral characteristics of the differences, to provide classification accuracy.(6) For multispectral images, decision tree based on spectral characteristics classification accuracy is lower. Adding texture features in the classification model, can greatly improve the accuracy of classification.