Document Type : Original Research Paper
Authors
1 Faculty of Mining, Petroleum & Geophysics Eng., Shahrood University of Technology, Shahrood, Iran
2 Shahrood University of Technology
3 Faculty of Electrical Engineering, Shahrood University of Technology, Shahrood, Iran
Abstract
The identification of rock discontinuities is a critical factor in the field of mining and construction projects. Traditional methods for conducting this task is often difficult, time-consuming, poses risks to the human safety, and lead to incomplete evaluations. With introduction of unmanned aerial vehicles (UAV) has changed this process and has allowed to cover all the area in a short time without endangering employees. The aim of this paper is to employ deep learning using python programming language to develop and train a neural network based on the UNET++ architecture in order to identify rock surface discontinuities automatically by means of UAV-captured imagery. It is also addresses challenges associated with supervised learning, particularly overfitting, by implementing data augmentation techniques and reducing model parameters by approximately 6%. Consequently, the pixel-wise precision criterion improved significantly from 53.27% to 75.6%. Especially, this work stands out from other studies by focusing specifically on UAV imagery for geological assessments, employing a dual strategy to overcome overfitting, and demonstrating effective performance despite the limited training data. The result showed that the model is capable to identify rock discontinuities accurately and is a suitable method for the mining and construction industries.
Keywords
Main Subjects
)