A cognitive method for building detection from high-resolution satellite images

In recent years, high-resolution satellite (HRS) images have become an important source of data for extracting geo-spatial information. A deep understanding of human cognitive capabilities is required in order to automate the method of information retrieval from HRS images. The aim of this study is to emulate human cognitive processes by integrating cognitive task analysis for information extraction from HRS images. First, the preliminary knowledge about the cognitive processes which human beings acquire during the interpretation of satellite images is collected. Then, knowledge is represented in the form of rules which are based on the visual interpretation of the images by the human beings. During knowledge elicitation these rules are used to extract buildings from HRS images utilizing the mixture tuned matched filtering algorithm. Later, the method is tested using 14 HRS images of an urban area. The average of precision, recall and F-score is computed as 79.45%, 64.34% and 70.28% respectively.