Research On The Method Of Object-oriented Residential Areas Change Detection

Residential area is one of the key features of the fundamental geographic information. With the development of the urbanization, the residential area data change frequently. Residential areas change detection is very important for disaster assessment, urban expansion, environmental change and spatial database updating. Therefore, the residential area change detection methods from multi-temporal images are studied and a new method is proposed in this paper. The main research work is as follows:(1) The theory of multi-resolution image segmentation is explored. Different spatial resolution images are used for segmentation. The comparison of different scale parameters is made in the process of image segmentation using experiment.(2) An automatic residential area extraction method is proposed based on structural similarity. In this method, the characteristics of residential area samples are stored in MySQL database, and the similar samples can be avoided duplicate storage by means of structural similarity index, thus the correlation between the samples in the database is very small. Then, by comparing the gray and texture characteristics of divided image objects with that of samples in the database according to their structural similarity, the residential areas can be obtained. Furthermore, since the samples exist in the database, which can be repeated use, the efficiency of selecting samples is improved.(3) A residential area change detection method is proposed based on overlapping area ratio. In this method, the overlay operation between the map spots of residential area extracted from two temporal images is conducted. Then through the size of overlapping area ratio, the changed map spots and type of change (including no change, partly change, appearance, disappearance) can be determined.(4) A prototype system of residential area change detection is developed using visual studio2008, Arcgis Engine9.3and ENVI/IDL to examine the method presented in this paper. Then the proposed method is compared with the other two change detection methods. Experimental results show the method can extract changed map spots of residential area well and the detection accuracy is high. Moreover, it is superior to the other two change detection methods.