Abstract <p>The use of thermal imaging for the reconstruction of historical heritage sites is limited by the influence of present day cultivation and the absence of archaeological sites with contrasting thermal properties. Unlike other methods, the presented statistical analysis algorithm uses Haralick textural features at the feature extraction stage. Subsequent image segmentation makes it possible to identify occupation layer sections of varying thickness as areal landscape objects. Their properties are interpreted based on reference data from interdisciplinary studies—excavations, geophysics, and soil drilling. After assessing the trends in the relative position of segments of different classes and the ratio of their areas, we propose a hypothesis about the boundaries of the archaeological site. Analysis of the structure of a segmented thermal imaging image on the site of the medieval Kushmanskoye III settlement (Udmurtia) showed fundamental differences in the distribution of the humus layer of soil in the settlement and the adjacent territory, not subject to anthropogenic impact. The proposed algorithm can be considered as an effective tool for studying historical heritage sites</p>