Intelligent Water Boundary Extraction From Remote Sensing Image

It is one of the most important research fields of Photogrammetry and Remote Sensing toextract ground objects from remote Sensing imagery automatically. However, after more than40years’ effort, the automation degree of ground objects extraction is still not satisfactory.Identification and extraction of water were still carried out by operators’ interpretation andcollection, though it’s a type of important ground objects. It’s inefficient and manual. Aiming atsolving the above-mentioned problems, this thesis starts off with automatic extractionexploration and practical production and takes each kind of water in optical remote sensingimagery as research targets. The intelligent extraction of water body and the problem of itsboundary positioning are particularly studied. The key points mainly include: Active ContourModel, Establishment of comprehensive water index, water body extraction based on spectralcharacteristic classifier, water body classification based on space characteristic classifier andautomatic boundary obtainment.Main work and innovations are ordered as follows:1. The present state of the research on water factor extraction in optical remote sensingimagery was analyzed.2. Spectral characteristic and space characteristic were deeply analyzed and the influence ofwater depth and suspension content were discussed to constitute a solid basis for water boundaryextraction.3. Human-computer synergy theory was analyzed to direct human-computer interactivestrategies. Two interactive methods were proposed with a view to precision and robustness.4. Active contour model theory was successfully applied to extract water boundary and itsmerit and demerit were analyzed, an automatic beginning active contour model combing withregion growth technique was carried out.5. Four kinds of spectral-relation models and five kinds of water index models weresummarized and analyzed and a kind of Comprehensive Water Index (CWI) was put forward.6. Taking decision tree as a basic frame, spectrum classifier using spectrum characteristicwas designed to to carry out the automatic extraction of water, space classifier using spacecharacteristic was designed to carry out classification of water extraction result.7. A series of techniques, such as region label, isolated small region elimination based onshape characteristic and area characteristic, holes fill and boundary obtainment were adopted torealize automatic extraction of water boundary.