Aerial images are a key data source when it comes to analyzing changes of the environment over time. This has a number of reasons. While the oldest available imagery typically dates back many decades, today's competing wide-area surveying methods provide more recent data only. Furthermore, unlike maps, aerial imagery is uninterpreted, which avoids potential biases, and allows for a wider range of applications. For quantifying changes over time, the orientation of images in a common reference frame is a precondition, however. This talk presents recent endeavors at TU Wien to achieve this goal more accurately and efficiently, including aspects like robust bundle adjustment strategies, and the search for homologous points within and across image data sets from different points in time, and covering civil and reconnaissance imagery taken since the 1940ies, including a Romanian data set from 1971.