Change detection is a highly demanded field of research with multiple approaches. Most approaches have in common that changes are detected without describing information. Classification-based methods offer the opportunity to extract detailed information concerning the type of the changes. Conventionally, land-use / land-cover classification can be performed in supervised or unsupervised manner. Both strategies require comprehensive training and reference data. In this paper, an aspect of the approach for the classification of changes is presented, which does not require training or reference data. Furthermore, no expert knowledge is needed. In consequence, a high rate of practical utility for image recognition results.
