Resumo

RODRIGUEZ-CRUZ, M.  e  MOCTEZUMA-FLORES, M.. Análisis bayesiano y fusión de datos para la clasificación de escenas urbanas del Distrito Federal. Ing. invest. y tecnol. [online]. 2006, vol.7, n.1, pp.17-28. ISSN 2594-0732.

In cities like Mexico, D.F., there exists in creasing administration difficulties, orignated by the lack of information on the urban structures and the dynamism of their evolution; therefore, the acquisition of aerial high resolution photographs of urban regions, can facilitate tasks in the update of cartography and urban planning through the segmentation and automatic extraction of interest regions. In this article, we present a method to segment and classify aerial photographs of Mexico City. In order to obtain it, first a reduction in the gray levels of the original image is made. Later its matrix of co-occurrence is calculated and from this ten tex tural features are obtained. The textural features that provide the sufficient information to segment to the image are chosen. Later, a homogenization process is made, and finally, the resultant texture matrix is taken into a multiclasses fusion process. In this last stage, we propose a method to fuse several classes. The final result shows the performance of the proposed method, providing urban scenes classified in three different classes.

Palavras-chave : Matrix of co-occurrence; textural features; segmentation; classification homogenization; multiclasses fusion.