Extraction Of Structural Geomorphometrical Information From Digital Elevation Models

Terrain is a fundamental feature in various research areas, which controls the local redistribution of energy and water. It has an important impact on the hydrological, soil, biological and geomorphological processes. In the recent years, the fast developing technologies in electronics and aeronautics give people the opportunity to re-understand our earth, especially terrain. The key is digital terrain data.Cartography and GIS provide a representation and management platform for the location based phenomena and further facilitate the spatial analysis and engineering applications. Among these, terrain plays an important role which sets a reference framework for other geographical features in the digital environment. Digital terrain analysis is coined as a research area which studies data model of digital terrain data, information derivation from elevation data, application in geosciences. The success in data acquiring technology poses a challenge for all the research topics because the availability of voluminous data with high accuracy and resolution makes human being lost. The recent research on promoting high-level representation from the primary digital elevation data models, which include contour, TIN and regular grid, are an active area. A basic result is surface network of terrain. The surface network, which has many its origins as structural lines, surface specific lines, Warntz Graph, Pflatz Network and contour tree, has lots of roles in automatic processes on terrain information. The geographical information generalization is an example which employes surface network as the terrain's abstraction and generalize it instead of contour for the effectiveness. Currently the main type of the digital terrain data is regular grid. Nevertheless, surface network is not the only choice in all situations for grid. Contour has its name in representing terrain in an adaptive way and quality communication of geomorphological information. However the efficiency of contour threading algorithm in grid has to improve because of the data volume and user command.This paper studies the algorithm of valley lines extraction from grid data. Existing methods which can be classified into local concavity, water accumulation and polynomial surface analysis have limited success in different situations. Our algorithm combines the first two types methods. It has three stages in general. Firstly water accumulation is used to extract the primary valley lines. The original implementation of this algorithm has some potential to refine. We proposes a new depression filling algorithm which does not use the combining of depression and the window resizing. The filling process starts from the pit point and find the depression range in a circular way and the final boundary is not wider two points than the actual one. The threshold strategy is core idea of simulating valley as flowing process dominating area. But it is so subjective that the result is not pleasing because it may contain too detail valley lines in the lower area and miss a lot typical valley lines in the higher area. The second stage of our algorithm, which includes bottom-up and top-down analysis, serves as the extension based on the primary valley lines acquired in the first stage. The local concavity method is used here for its simplicity and robust in the area which contain typical valley lines. But the latter one is noise sensitive and produces clutter result. So we need clean it to an acceptable one in the third stage which first organizes thevalley lines in Strahler order based the flow direction of each point on the extracted valley lines in previous steps. Then we classify the clutter of valley lines into different categories and design corresponding rules to clean them. The minor valley lines in the final result can be pruned based on length and watershed area. The conversion from raster to vector is easy based on Strahler order and could have more application.As noted above contour line is a more adaptive way to represent terrain than grid and even has advantages in some situation. Here we design new algorithm for extracting contour lines from grid based on interval-tree index and bucket sorting index. The experiment showed that it is much more efficient that ever ones. Some tests recommend that latter index have better performances in construction and query.The further study is concentrated on the spatial relation of contour lines because we believe that the efficient application of contour lines must have both geometric data and the relation information. The framework we proposed is based on the directional adjacency of contour lines which define twelve kinds basic relation patterns. The relation is built based on the triangulation of original contour lines. The effectiveness of our framework is demonstrated by the applications. The first is on automatic elevation labeling which has many practical meaning in the conversion of contour lines from traditional media to digital ones. The second is identification and connection of broken contour lines which can be encountered in same process. The last one is about the construction of contour lines bounded area in which the open contour lines is hard to deal with in existing methods.