Improvement of accuracy in extracting surface features information is significant and sophisticated. This paper improves the method of extracting vegetation information by selecting the best bands group in West Liao River Basin. The medium-resolution of Landsat TM image with the solution of 30 meter obtained in 2010 were selected as the data source. During the extraction process, Principal Component Analysis is used to separate key information from background noise, which reduces the data redundancy. With the consideration of vegetation chlorophyll information, containing more information, the second principal component was selected to analyzing the bands correlation coefficient. Normalized difference vegetation index (NDVI) was chosen as one component. By calculating the correlation coefficient of band1 to band5, band7, the second principal components and NDVI, we found band1, PC2 and NDVI have the least correlation. Maximum Likelihood method of supervised classification is used to classify the surface features on basis of band1, PC2, NDVI and band5, band4, band3 combination image, respectively. The result shows that the overall accuracy of classification based on the new bands combination increased by 6.45% than based on original band. The main reasons are that the new band combination can eliminate texture interference and has the little correlation coefficient.
展开▼