Study On Digital Geomorphy Of Fuzzy Datasets Based On Geomorphometry

The earth surface is the most active interface of human activities. As a basic element of the earth surface system, the landscape directly influences the human activities. Therefore, research on landscape morphology has been being under the geologist's attention. During the past three decades, the Digital Elevation models (DEMs) are widely used for computer numerical analysis of terrain. Use the regular grid DEM dataset to extract the slope, aspect, convexity, concavity and so on, and automatically get the physiognomy attributes from the neighboring relationship. It is widely applied in civil engineering, geology, geomorphology, hydrology, environmental science, landuse, soil erosion and vegetation cover. Especially in physiognomy and hydrological research, the application of the special charts based on DEM research is more common.Many efforts have been made to examine how to define landform units from DEMs. Depending on the geomorphic processes which are considered, landforms may be modeled quite differently, we described landforms through a hierarchical subdivision of the land surface into relief units with homogenous gradient, aspect and curvature as well as form elements/facets with homogenous plan and profile curvature. Landform was described as hierarchical entities into four levels: landform elements, landform types, physiographic systems and physiographic regions, of which landform elements and types are the most important.For a long time, landform was compiled from aerial photographs and topographic maps, however, those are time-consuming and arduous, and the region of study range is restricted. So the work has remained in the regional detailed research of stage. With the rapid development of aerospace and the booming of the laser radar technique, methods of data collection make great progress, the land surface information collected become more various, such as the shuttle radar topography mission of NASA which gets global digital elevation models as 3 arc-second, and collaboration of Ministry of Economy, Trade and industry of Japan and NASA, GDEM was produced by stereo image. Those provid us reliable data for geomorphometry from regional scale to global and planet.However, with technological development, access to data capabilities, the data demonstrated a high-dimensional, high information content of features in the data to extract information on to us with new challenges. In this paper, the characteristics, summarized from previous experience, trying to find a unified framework, explore the landscape in the form of classification at the same time, an effective solution to the problem of fuzzy data extraction.Digital computer and geographic information systems precludes any size and spatial resolution surface geometry based on classification of the regional barriers. Fine terrain elevation information is now automatically parse through, usually elevation data to form a regular grid digital elevation models to express.In this study, use the ASTER-DEM data of 1 arc second (mid-latitudes of about 26m) resolution which was released on June 29. 2006 by METI and NASA, and put forward a method of integrating geomorphology parameterization and self-organizing neural network for classification research of landscape form. This method can be effective for the vast amounts of data dimensionality reduction, and automatic to get landform classification threshold, reducing the subjective impact of human involvement. To some extent, broking the limits on the number of classification of landscape form parameterization methods and also reducing the misclassification phenomenon of landform types. Changchun region geomorphic study achieved satisfactory results.Through the quantitative geomorphological classification research of Changchun,the main research results are showed as follows:(1) improved fitting method of surface,a partial quartic equation was proposed. Deduced formula of slope, cross-sectional curvature, maximum curvature and minimum curvature, and produced corresponding map of Changchun. (2) Integating geomorphology parameterization and self-organizing map to construct classification model;(3) To achieve quantitative geomorphological classification, types and thresholds are given by the self-organizing map neural network by training, classification scheme is determined by a computer program automatically.A rapid expression of geomorphological classification has been achieved by study on the landscape, and the workload of geomorphological map artificial interpretation has been reduced, which greatly improves the work efficiency; classification threshold is computed automatically, which not only enhances the spatial information of different scales of flexibility, but also improves the classification accuracy for environmental research, precision agriculture, geological hazard assessment, geological structure and basin delineation study to provide accurate background parameters; This research has practical significance on civil engineering, urban planning, management of nature reserves and mineral surveys, etc.