期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - Part 1
出版社：Copernicus Publications
摘要：Building can be classified into many categories and/or classes which can be identified efficiently from images. The classes can include usage, types, shapes, or any other specific categories. The classification of buildings to predefined classes is called building recognition. This is usually based on the fact that building consumes region in the image and these region boundaries can be used to process and create building descriptors to be used in a building libraries to be utilized in the classification procedure. These building descriptors parameters are then used in creating the libraries to be used in the recognition technique which can be represented in both the spatial and frequency domains. This research paper will consider building boundary as building representative and use Monte Carlo Simulation (MCS) capabilities for investigating building descriptors. MCS is used to generate random variables with pre-specified distribution to identify the boundary of the region of interest through identifying the boundary of this building or area. The building boundary can be described by the ratio between its inside generated random values and total generated random values in the boundary image. This research concentrates on investigation the optimum number of generated random variables to be used as building descriptors for buildings recognition. This descriptor will identify the building regardless of its orientation or scale. This means that the descriptor will not be affected or sensitive to the direction or the scale of the building and therefore it can be applied to any type of images without any geometrical corrections. This non sensitivity of the proposed MCS descriptor is the motivation for the assessment of this technique for building recognition. In the paper two different classification techniques will be used for the assessment of this new descriptor for building recognition. The first one is the minimum distance classifier. The second is the optimum statistical classifier that will produce the Probability Density Function (PDF) of building recognition. The assessment methodology for testing this recognition technique will be conducted by using Sequential Monte Carlo Simulation (SMCS) based on Bayes theorem to get the probability density function. Conclusions and recommendations are given with respect to the suitability, accuracy, and efficiency of this method
关键词：Building Representation; Building Description; Building Recognition; MCS
