The monitoring of building activity (erection of new buildings, demolishing buildings and especially the change of building heights) by manual inspection of space and aerial images is time consuming and a source of errors. A detection of building changes based on the comparison of digital surface models (DSMs) is more reliable. For this study DSMs have been generated based on aerial images, an IKONOS and a GeoEye-1 stereo pair taken from 2007 up to 2009. By pixel based matching with dynamic programming, semiglobal matching and least squares matching the visible surface has been determined. Semiglobal matching leads to sharp building shapes, while the area based least squares matching smoothes the height model and has more problems in areas with little or without contrast. As is shown in the investigation building changes and building height changes in the range of one floor can in many cases be determined with all methods, however building shapes are better determined using semiglobal matching.