EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Topographic lineaments extraction in digital elevation model

Pedro Rossi, Julio Cesar Horta de Almeida, Antônio de Pádua Pires, Mario Neto Araujo

Boletim de Geociências da Petrobras · 2025

DOI: 10.70369/188dww73 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Boletim de Geociências da Petrobras, 2025. Verfasst von Pedro Rossi und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Topographic lineaments seen on digital elevation models (DEM) provide good insights into geological structures and lithotypes. To extract useful information from these images, interpretation is a crucial step. Interpreting DEM images is conventionally done by trained scientists drawing line segments over the image, in an often repetitive, time-consuming, and undoubtedly subjective process. Recently, machine learning techniques have been employed for recognizing patterns in DEM data and extracting lineaments in semi-automated mode. This technique has the potential to help scientists reduce subjectivity and reduce the life cycle of interpreting lineaments on remote sensed images. In this paper we take on the task of extracting lineaments from Shuttle Radar Topography Mission (SRTM) DEM data in the central Ribeira fold belt in SE Brazil, a deep eroded Neoproterozoic-Cambrian orogen. We analyze how human interpretation varied over five rounds of lineaments annotations in the same area and compare them with the output of a convolutional neural network (CNN) with U-net architecture trained to classify images at pixel level into lineaments or non-lineaments. Results show that the criteria for manual lineament extraction varied substantially throughout the interpretation rounds. Lineaments became fewer, and repeatability increased towards the later versions. The CNN hyperparameters had to be set through a series of trial-and-error tests but eventually yielded results that mimic the ones obtained manually. CNNs are powerful tools to process and analyze data, but require time and effort to build, label, parametrize, train and test. In this sense it is justifiable to apply CNNs over large datasets, when it is unfeasible to interpret lineaments manually, otherwise, it is more practical to have a human draw by hand with two or more interpretation versions to reduce the subjectivity.

Zitieren

APA

Rossi, P., Almeida, J. C. H. D., Pires, A. D. P., & Araujo, M. N. (2025). Topographic lineaments extraction in digital elevation model. Boletim de Geociências da Petrobras. https://doi.org/10.70369/188dww73

Vancouver

Rossi P, Almeida JCHD, Pires ADP, Araujo MN. Topographic lineaments extraction in digital elevation model. Boletim de Geociências da Petrobras. 2025. doi:10.70369/188dww73

BibTeX
@article{Rossi2025,
  author = {Rossi, Pedro and Almeida, Julio Cesar Horta de and Pires, Antônio de Pádua and Araujo, Mario Neto},
  title = {Topographic lineaments extraction in digital elevation model},
  journal = {Boletim de Geociências da Petrobras},
  year = {2025},
  doi = {10.70369/188dww73}
}
Kostenlos starten →