A convolutional neural network for semi-automated lineament detection and vectorisation of remote sensing data using probabilistic clustering: A method and a challenge

Authors: Amin Aghaee (a) Pejman Shamsipour (a) Shawn Hood (a) Rasmus Haugaard (b)
(a) GoldSpot Discoveries Corp., 69 Yonge Street, Suite 1010, M5E1K3, Toronto, Ontario, CanadaMetal Earth, Mineral Exploration Research Centre,  (b) Harquail School of Earth Sciences, Laurentian University, Sudbury, P3E2C6, Ontario, Canada 
Click here to be directed to the publication.
Abstract
In this paper we present the framework and open-source software code to train and apply Deep Learning Convolutional Neural Networks (CNNs) to the prediction of geological lineaments using topographic, magnetic, and gravity raster data. Many important applications relate to the recognition of linear geological structures from remote sensing data, such as thrust faults, bedrock fault and shear zones, lithological contacts, fractures and fold structures. The digitization of fault lineaments is conventionally performed by geologists or geophysicists with working knowledge of the relevant data, e.g., topographic Digital Elevation Model, magnetic data, and gravity data. Visual inspection and extraction is simple but subjective; the process is also time-expensive with efficiency and accuracy depending on the individual's knowledge, experience, and skill. For decades there has been interest in ways to automate this process. Our CNN approach is trained using publicly available lineament GIS data from the Quest BC project in British Columbia, Canada, and the Loch Lilly-Kars area of New South Wales, Australia. The datasets used to train the prediction models resulted in interesting predictions proximal to the training areas: some major lineaments are indicated, some are missed, and potential new (valid) lineaments are indicated. The results indicate potential for use as a semi-automated lineament detection solution. In contrast, but as anticipated, application of the model to the blind-test area of the Swayze greenstone belt, Ontario, produced poor lineament prediction results (as compared to publicly available interpretations). We interpret this result as related to insufficiently large training data inputs. However, it is inferred that results could be improved through feature engineering (e.g., use of topographic slope, rather than simply elevation) without the need to simply create larger training datasets. We hope that, by making the code open-source, the geoscience community will use this platform to gradually improve an open source fault prediction model.

Datasets available for download

Additional Info

Field Value
Last Updated March 25, 2026, 02:06 (UTC)
Created March 25, 2026, 02:06 (UTC)
Domain / Topic
Domain or topic of the dataset being cataloged.
Title
Title for the Dataset.
A convolutional neural network for semi-automated lineament detection and vectorisation of remote sensing data using probabilistic clustering: A method and a challenge
Description
A description of the dataset.

Authors: Amin Aghaee (a) Pejman Shamsipour (a) Shawn Hood (a) Rasmus Haugaard (b)
(a) GoldSpot Discoveries Corp., 69 Yonge Street, Suite 1010, M5E1K3, Toronto, Ontario, CanadaMetal Earth, Mineral Exploration Research Centre,  (b) Harquail School of Earth Sciences, Laurentian University, Sudbury, P3E2C6, Ontario, Canada 
Click here to be directed to the publication.
Abstract
In this paper we present the framework and open-source software code to train and apply Deep Learning Convolutional Neural Networks (CNNs) to the prediction of geological lineaments using topographic, magnetic, and gravity raster data. Many important applications relate to the recognition of linear geological structures from remote sensing data, such as thrust faults, bedrock fault and shear zones, lithological contacts, fractures and fold structures. The digitization of fault lineaments is conventionally performed by geologists or geophysicists with working knowledge of the relevant data, e.g., topographic Digital Elevation Model, magnetic data, and gravity data. Visual inspection and extraction is simple but subjective; the process is also time-expensive with efficiency and accuracy depending on the individual's knowledge, experience, and skill. For decades there has been interest in ways to automate this process. Our CNN approach is trained using publicly available lineament GIS data from the Quest BC project in British Columbia, Canada, and the Loch Lilly-Kars area of New South Wales, Australia. The datasets used to train the prediction models resulted in interesting predictions proximal to the training areas: some major lineaments are indicated, some are missed, and potential new (valid) lineaments are indicated. The results indicate potential for use as a semi-automated lineament detection solution. In contrast, but as anticipated, application of the model to the blind-test area of the Swayze greenstone belt, Ontario, produced poor lineament prediction results (as compared to publicly available interpretations). We interpret this result as related to insufficiently large training data inputs. However, it is inferred that results could be improved through feature engineering (e.g., use of topographic slope, rather than simply elevation) without the need to simply create larger training datasets. We hope that, by making the code open-source, the geoscience community will use this platform to gradually improve an open source fault prediction model.
Tags / Keywords
Keywords/tags categorizing the dataset.
Format (CSV, XLS, TXT, PDF, etc)
File format of the dataset.
Dataset Size
Dataset size in megabytes.
0.07
Metadata Identifier
Metadata identifier – can be used as the unique identifier for catalogue entry
Published Date
Published date of the dataset.
2021-03-30
Time Period Data Span (start date)
Start date of the data in the dataset.
Time Period Data Span (end date)
End date of time data in the dataset.
GeoSpatial Area Data Span
A spatial region or named place the dataset covers.
Field Value
Access category
Type of access granted for the dataset (open, closed, service, etc).
public
License
License used to access the dataset.
License not specified
Limits on use
Limits on use of data.
Location
Location of the dataset.
https://metalearth.geohub.laurentian.ca/datasets/fca6d09c0abc4c8eab3572005353d499
Data Service
Data service for accessing a dataset.
Owner
Owner of the dataset.
MetalEarth
Contact Point
Who to contact regarding access?
Contact Point Email
The email to contact regarding access?
Publisher
Publisher of the dataset.
Publisher Email
Email of the publisher.
Author
Author of the dataset.
MetalEarth
Author Email
Email of the author.
Accessed At
Date the data and metadata was accessed.
2021-03-30
Field Value
Identifier
Unique identifier for the dataset.
fca6d09c0abc4c8eab3572005353d499
Language
Language(s) of the dataset
English
Link to dataset description
A URL to an external document describing the dataset.
https://metalearth.geohub.laurentian.ca/datasets/fca6d09c0abc4c8eab3572005353d499
Persistent Identifier
Data is identified by a persistent identifier.
Globally Unique Identifier
Data is identified by a persistent and globally unique identifier.
Contains data about individuals
Does the data hold data about individuals?
Contains data about identifiable individuals
Does the data hold identifiable data about individual?
Contains Indigenous Data
Does the data hold data about Indigenous communities?
Portal Type
Platform type of the source portal.
Field Value
Version
Version of the datatset
None
Source
Source of the dataset.
None
Version notes
Version notes about the dataset.
Is version of another dataset
Link to dataset that it is a version of.
Other versions
Link to datasets that are versions of it.
Provenance Text
Provenance Text of the data.
Provenance URL
Provenance URL of the data.
Temporal resolution
Describes how granular the date/time data in the dataset is.
GeoSpatial resolution in meters
Describes how granular (in meters) geospatial data is in the dataset.
GeoSpatial resolution (in regions)
Describes how granular (in regions) geospatial data is in the dataset.
Field Value
Indigenous Community Permission
Who holds the Indigenous Community Permission. Who to contact regarding access to a dataset that has data about Indigenous communities.
Community Permission
Community permission (who gave permission).
The Indigenous communities the dataset is about
Indigenous communities from which data is derived.
Field Value
Number of data rows
If tabular dataset, total number of rows.
Number of data columns
If tabular dataset, total number of unique columns.
Number of data cells
If tabular dataset, total number of cells with data.
Number of data relations
If RDF dataset, total number of triples.
Number of entities
If RDF dataset, total number of entities.
Number of data properties
If RDF dataset, total number of unique properties used by the triples.
Data quality
Describes the quality of the data in the dataset.
Metric for data quality
A metric used to measure the quality of the data, such as missing values or invalid formats.

0 Comments

Please login or register to comment.