Forest Vegetation Classification Of Huanlong Mountain Using Multi-source Remote Sensing Data

Application of remote sensing data for the forest resource survey, forest resource management and other services is the trend of the forestry development. Remote sensing image has the characteristics of convenient, wide range, strong timeliness. Spatial and spectral information of remote sensing images are important information source of forest resource survey. Combined remote sensing method with geographic information system(GIS) and global position system(GPS) have been widely used in forest production work.Resources satellite three is the first independent civil high resolution stereo mapping satellites of China. Its panchromatic data resolution is 2.1 meters; multispectral data resolution is 5.8 meters. It has the characteristics of high resolution and low price. To apply ZY-3 image in forestry research can save costs and get rid of the dependence on foreign high resolution images. In this study, we used ZY-3 panchromatic and multispectral data, Landsat 8 OLI data and SPOT5 multi-spectral data to study the different characteristics of different data in Huanglong Mountain forest vegetation classification.(1)Through multi-source remote sensing image fusion, we can improve the spatial resolution of the image, rich spectral information of images, make image clearer, help to improve the quality of the image. Gram-Schmidt fusion is ZY-3 panchromatic and multispectral images, ZY-3 panchromatic and SPOT5 multi-spectral image best fusion method, PCA transformation is the best image fusion method of ZY-3 panchromatic and Landsat 8 OLI images.(2) ZY-3 images using support vector machine classification have the highest classification accuracy. Using supervised classification method(maximum likelihood and support vector machine), the classification accuracy of each images can be up to 90%. Supervised classification accuracy was significantly higher than unsupervised classification.(3) Classes in second level of forest vegetation in the study area choose ZY-3 panchromatic and multi-spectral fusion image, using support vector machine classification, its overall accuracy was 84.83%, Kappa coefficient was 0.7836. The classification accuracy is higher than the maximum likelihood classification when we use support vector machine classification. Selection ZY-3 fusion image and Landsat 8 OLI fusion image classification accuracy is higher than SPOT5 fusion image.(4) ZY-3 fusion image combining NDVI and texture information, its overall classification accuracy was 87.85%, Kappa coefficient was 0.8269. The overall accuracy and Kappa coefficient results were higher than only use ZY-3 fusion images.For the analysis of the results the following conclusions can be drawn: the forest vegetation information extraction accuracy of ZY-3 image is higher enough in study area. If we can make full use of inexpensive domestic high-resolution satellite effective, it will reduce the cost of forestry research and production and get rid of the dependent foreign high resolution images. Using Support vector machine method to class high-resolution images, the classification accuracy is higher than other methods. Extraction NDVI and texture information help to improve forest vegetation classification accuracy of the study area.