Integrative Analysis And Evaluation Of The Interpretation Features In Remote Sensing Image

The research on land use/cover change plays an important role in the global change project, and timely and accurate acquirement of land use/cover information is significant to governmental decision-making and scientific research. With the diversification of remote sensing platform and the improvement of spatial resolution, remote sensing has already become an indispensable technique in land use/cover research. Interpretation signature, as the linkage between image information and ecological characteristics of land use/cover, has become one of the main limitations in automatic and accurate classification. Analysis of the classification ability of the single feature and integrated function of multi-features are the precondition to realize accuracy and automatic information extraction of remote sensing.With the Landsat TM image obtained in June, 2000, as the main data source, 14 land-use types with larger area percentage were selected as the research objects according to the common three-level classification hierarchy of land use/cover investigation. On the basis of the investigation, the main researches are:(1) Based on the current supervised classification procedure, we analyzed and assessed the classification capability of original spectrum, vegetation index and texture. The results show that the spectral feature is the dominant feature sources, the integration of 5 bands or 6 bands could get the best classification; textural feature is the second feature sources, and the multi-band combination of texture feature has the best classification accuracy; the role of vegetation index was relative poor.(2) Based on the segmentation technique, landscape metrics was extracted and then used to comparative analysis along with other object-oriented features. The results indicated that the function of spectral feature and texture was consistent to that of pixel-based. The classification capacity of shape and landscape metrics was worse by themselves, so they could only act as auxiliary information, and the effect of landscape metrics was better than that of the shape obviously.(3) Based on the comparative analysis of the capacity about object-oriented and pixel-based classification, an object-oriented pixel classification was provided and the results shown this method could integrate the virtues of two kinds of classification methods.