Building Change Detection from Historical Aerial Photographs Using Dense Image Matching and Object-Based Image Analysis

A successful application of dense image matching algorithms to historical aerial photographs would offer a great potential for detailed reconstructions of historical landscapes in three dimensions, allowing for the efficient monitoring of various landscape changes over the last 50+ years. In this paper we propose the combination of image-based dense DSM (digital surface model) reconstruction from historical aerial imagery with object-based image analysis for the detection of individual buildings and the subsequent analysis of settlement change. Our proposed methodology is evaluated using historical greyscale and color aerial photographs and numerous reference data sets of Andermatt, a historical town and tourism destination in the Swiss Alps. In our paper, we first investigate the DSM generation performance of different sparse and dense image matching algorithms. They demonstrate the superiority of dense matching algorithms and of the resulting historical DSMs with root mean square error values of 1-1.5 GSD (ground sampling distance) and yield point densities comparable to those of recent airborne LiDAR DSMs. In the second part, we present an object-based building detection workflow mainly based on the historical DSMs and the historical imagery itself. Additional inputs are a current digital terrain model and a cadastral building database. For the case of densely matched DSMs, the evaluation yields building detection rates of 92% for grayscale and 94% for color imagery.