The authors discuss the features of determining the contours of technogenically disturbed territories based on open satellite imagery data using computer vision technologies. They will automate the formation and updating of a retrospective information and analytical database of solid minerals open mining sites for the subsequent protection of environmental components, strategic and operational risk management associated with the mining sector. We identified artificial neural convolutional networks as the main tool for segmenting raster data and selected one of the popular implementation options in the form of the YOLOv8 architecture. The set used consists of fragments of Sentinel-2 satellite data and markings in the form of vector polygonal objects for the territory of Novosibirsk oblast. Tools for marking, preparing and generating a data set are described. The results of a comparison of several variants of pre-trained networks are presented in terms of final accuracy and training time, as well as conclusions on setting up hyper-parameters for similar tasks. The model is built into a data processing pipeline based on Prefect software