The research presented is based on data acquired by the integrate measuring system of ADS40 digital camera. The data were collected from a lowland area, mostly agricultural, partly covered with forest, of which about 20% was urbanized area. Base on the ADS40 imagery, stereoscopic models were built and used to manual and autocorrelation measurements of three-dimensional objects. Before the correct orientation of stereoscopic models was obtained, the images were submitted to geometrical correction, and next to orientation in the process of aerotriangulation with the method of bundle adjustment. Next, a semi-automatic measurement was conducted which was connected to interpolation, filtration and classification, obtaining a three-dimensional model of the surface of roofs, and also performing a manual measurement of the skeletal line of buildings. It was found that the autocorrelation algorithm of the program Match-T could not realize the measurement of points on the lateral faces of the building because of a too small base relation of images. For autocorrelation measurements on digital images that included an urbanized area with high degree of vegetation cover the extraction of the building elements amounted to about 85%.