Information Extraction Method Of The Analysis Based On Spectral Characteristics Typical Yeyahu Wetland Plants

Vegetation is an important component of the wetland system, which is sensitive indicator to wetland ecosystem health or not, so the vegetation classification is the wetland research foundation. Remote sensing methods have been widely applied to detection of wetland vegetation. The purpose of this study is how to extract the vegetation data faster, more accurate.The traditional multi-spectra data could identify big objects, but it can't the requirement to identify wetland plant community. Vegetation spectral data with abundant spectrum (thousands of bands), and high spectral resolution (<10nm), and specific diagnostic features, are possible to be used to improve the classification accuracy of the wetland plant community. Object-oriented method breakthrough the method of only based on remote sensing spectrum information, This method can be combined with the organic advantage of spectral information, it making full use of remote sensing images feature, and include geometry, shape, texture and so on, to identify species and distributed to different species, different segment can be identified. Yeyahu wetland is the study area in this research, on the base of typical plants'field spectral reflectance data and Geoeye-1 image data, by Comparative Analysis method and one-way ANOVA method, obtained the extraction parameters, and then guide for extract Geoeye-1 image information by Object-oriented method.It including:①analyze the spectral characteristics of five typical vegetations (Phragmites australis. Typha orientalis Presl. Scirpus planiculmis Fr. Schmidt. Medicago sativa Linn. PopulussimoniiCarr.) 3 key growth period of study area, explore the optimal strategy for identification and classification of wetland vegetation, including identifying trie best time to identify the characteristic parameters;②Unifies object-oriented method and analysis result of the typical plant community spectrum index of reflection data, to identify the five typical wetland plant on the base of Geoeye-1 image data, Using remote sensing data's feature, near-infrared wave band ratio, wave band ratio, shape index, brightness and so on, form a set of extraction typical plant community model in research area.Overall, the results shows that the spectral reflectance of five typical plants with "two-convex-three-concave" character, and the highest speed of the reflectance increasing is located nearby 550nm and 720nm, besides band 720 has the highest peak value. And the typical smaller concaves are located nearby band 500nm,720nm and 900nm, besides band 550nm-720nm rang has the strongest absorption strength (the maximum absorption intensity).From the different period different cover type, through comprehensive analysis of the spectral reflectance curve, absorption index, a vegetation index, using the follow results guide to orient-objected extraction:①Select july remote sensing data as extraction data;②The most easily identifiable is Medicago sativa Linn community, the green band and shape are the most obvious characteristics. The followed by Typha orientalis Presl.community, the green band is the best identification feature. The next is Phragmites australis and PopulussimoniiCarr. Community, the obvious feature is NIR band. And last, the most easily identifiable is green band to Scirpus planiculmis Fr. Schmidt. The extraction results showed that the accuracy of Medicago sativa Linn and Scirpus planiculmis Fr.Schmidt are best, the user accuracy of 91%, 100%, the overall accuracy is 90%, Kappa Coefficient is 0.88.In general, the analysis of the field reflectance data of wetland plants helped us to guide the object-oriented method for high resolution remote sensing image information extraction, the precision and speed are better than traditional object-oriented approach (totally dependent on experience to classification), this study can provide scientific and technical support to extract wetland plants.