Uzaktan algılanan resimlerde üst düzey yapıları bulmak için genel doku modelleri

ABSTRACTGENERALIZED TEXTURE MODELS FORDETECTING HIGH-LEVEL STRUCTURES INREMOTELY SENSED IMAGESEmel DoğrusüzgoM.S. in Computer EngineeringSupervisor: Asst. Prof. Dr. Selim AksoyJune, 2007With the rapid increase in the amount and resolution of remotely sensed imagedata, automatic extraction and classification of information obtained from suchimages have been an important problem in the field of pattern recognition sinceremotely sensed imagery is a critical resource for diverse fields such as urban landuse monitoring and management, GIS and mapping, environmental change andagricultural and ecological studies. This thesis proposes statistical and structuraltexture models for detecting high-level structures in remotely sensed images. Thehigh-level structures correspond to complex geospatial objects with characteris-tic spatial layouts in a region. As opposed to the existing approaches that arebased on classifying images using pixel level methods, we propose to use simplegeospatial objects as textural primitives and exploit their spatial patterns. Thisrepresentation can be viewed as a ?generalized texture? measure where the imageelements of interest are urban primitives instead of the traditional case of pixels.The spatial patterns we are interested in correspond to the regular and irregulararrangements of these primitives within neighborhoods.The methodology we propose in this thesis has two steps. First, the primitivesof interest are detected using spectral, textural and morphological features withone-class classifiers. Then, the spatial patterns of these primitives are modeled.At this step, either a statistical or a structural approach can be followed. Inthe statistical approach, analysis of the spatial arrangement of the primitives isdone by co-occurrence-based spatial domain features and Fourier spectrum-basedfrequency domain features. These features are used to quantify the likelihood ofpresence of the focused object in the image region being analyzed. In the struc-tural approach, a graph-theoretic representation is proposed where the primitivesform the nodes of a graph and the neighborhood information is obtained throughiiiivVoronoi tessellation of the image scene. Next, the graph is clustered by threshold-ing its minimum spanning tree and the resulting clusters are classified as regularor irregular by examining the distributions of the angles between neighboringnodes.The algorithms proposed in this thesis are illustrated with the detection of twogeospatial objects: settlement areas and harbors. The first step in the modelingof these objects is the detection of primitives such as buildings for settlementareas, and boats and water for harbors. In the second step, both statisticaland structural approaches are illustrated for the modeling of the spatial patternsof these objects. Results of the experiments on high-resolution Ikonos satelliteimagery and DOQQ aerial imagery show that the proposed techniques can beused for detecting the presence of geospatial objects in large remote sensing imagedatasets.Keywords: Pattern recognition, one-class classification, geospatial object detec-tion, co-occurrence texture analysis, Fourier texture analysis, graph-based textureanalysis.