Research On Drainage Networks Extracting From Multi-source Remote Sensing Data Based On Heuristic Information

GIScience (Geographic Information Science) is an interdisciplinary science which applies techniques and methods of computer science, surveying and mapping science and mathematics to geography. The core content of GIScience is redefining, reforming and resolving the existing domain issues of spatial data with GIS software and its environment. GIS is a tool and can be seen as a decision support system. GIS can process, analysis, model, store and express the geospatial data. Extracting drainage networks from DEM (Digital Elevation Model) is a powerful tool of hydrologic analysis. It is an important research field in DTA (digital terrain analysis), and one of important function modules and research contents of GIS. Extracting drainage networks play an important role in many applications of GIS, such as sediment analysis, land erosion analysis, environmental analyses and hydraulic engineering planning.With the development of sensor and platform technology of RS (Remote Sensing), the data of RS have a huge raise in terms of quality, quantity and variety. While, it is inherent heterogeneity, complexity and accuracy characteristics of geospatial data and fuzzy and uncertain expression of geographic phenomenon that present challenges for the automated derivation of fully connected drainage patterns. The computer intelligent information processing technology has obvious advantages in these areas, especially when dealing with the practical application.In this thesis, there is an intensive study on the methods of extracting digital drainage networks from DEM and other remote sensing data. This thesis is threaded by the practical application of intelligent information processing technology and tries to find the linking point between the intelligent information processing technology and the digital drainage network extraction and DEM data processing. The main achievements involve the following aspects:1. There is an intensive study and comprehensive summary on the existing methods of extracting digital drainage networks from DEM and other remote sensing data. Firstly, the thesis briefly describes the DEM and its characteristics, including the definition of DEM, related research and application of DEM. Secondly, the thesis describes the general workflow of extracting drainage networks in detail, including preprocessing the DEM (preprocessing the pits and the flats), generating flow direction matrix, flow accumulation matrix and drainage networks. Finally, the thesis deeply describes two steps, preprocessing the DEM and generating flow direction matrix. The two steps are the working emphases of the thesis. 2. The thesis proposes a method to evaluate the extracting digital drainage networks based on buffer analysis. There is a certain degree of surface configuration error in the results of the existing methods for extracting digital drainage networks from DEM and other remote sensing data. Usually, the methods would generate two kinds of results. The first kind of result is flow accumulation and specific catchment area. The second kind of result is terrain structural line and drainage networks in graph display. There is no standard method to assess the quality of the second kind of result. This thesis creates different buffers around the referenced river network and calculates the differences in shape and position between the extracted digital drainage network and the referenced river network. The method can evaluate the result globally, objectively and quantitatively. The proposed method has the character of agility, visualization and simple calculation. The later experiments prove that the method is effective. The method obtained the same results with the other evaluation method.3. The thesis formalizes the procedure of processing the pits and the flats as a path searching problem. The pits and flat areas are areas where the flow direction of cells cannot be determined with reference to eight neighbouring cells. The pits and flat areas are the biggest obstacles of extracting drainage networks from DEM.While, the pits and flat areas exist in all DEMs and handling pits and flat areas effectively is a key step and preprocessing work of drainage network extraction. The proposed method marked the pits and flat areas as"PIT"(the PIT is start node and the outlet of the PIT is target node) and selected successor and cost fuction. The procedure of dealing with pits and flats were formalized as a path searching problem. The pits and flats could be processed in one procedure and the procedure is very simple.4. The thesis proposes a method to deal with pits and flat areas with heuristic information. Under the framework of heuristic searching, the proposed method not only defined searching cost function which included vertical distance (elevation) and horizontal distance (path length), but also proposed the evaluation function which selected a large number of cells helps to assess the regional aspect surrounding the selected cell. The evaluation function ensures the proposed method only tries to select cells that most likely to lead to the direction towards the outlet quickly. Meanwhile, the proposed method modified the DEM and determined the flow dirction simultaneously. The proposed method corrected fewer cells and fewer slopes than the other methods and maintained the original landscape features. To examine the suitability and performance of the proposed method, the experiments were carried out on actual DEM data. The results were evaluated by the proposed evaluation method and others. The experimental evidence suggests that the proposed method can extract closer drainage networks to the reference river network than the other methods. The proposed method is feasible, effective and robust.5. Under the framework of heuristic searching, drainage networks are extracted from Multi-source remote sensing data based on heuristic information. If the DEM data are the only data source, all the methods for processing the pits and flat areas are powerless. The extracted drainage networks cannot match with the ground truth. Introducing new sources of information is a new idea for extracting the drainage networks. This thesis extracts the main drainage line from satellite imagery and the main drainage line is regarded as heuristic information of the selected cell. The information can control the position of the extracted drainage networks on certain degree and help the method obtain the better results. To examine the suitability and performance of the proposed method, the experiments were carried out on actual DEM data and LandSat-7, ETM+ band 5 data. The experimental evidence suggests that the proposed method is feasible, effective and robust.Finally, the software implemented by the thesis uses the standard data format and can be compatible with the main business open-source and GIS software. The software is powerful tool for hydrologic analysis and other practical application.