Abstarct: Identification of rock discontinuities is an essential part of studies and design of mining and construction projects. Conducting these studies by traditional methods is very difficult, time-consuming, and dangerous for human resources. Therefore, it does not take place in the full scope of the project. The use of unmanned aerial vehicle (UAV) has made it possible to access the whole area of the project in the shortest time, and without endangering human resources. In this thesis, using artificial intelligence and deep learning on the platform of Python program, the process of creating and training a neural network under the UNET architecture, with the aim of automatically identifying the discontinuities of the rock surface using the images taken by the UAV, has been carried out and the results have been stored as various models. The supervised learning process faced challenges such as overfitting. This challenge was solved using the data augmentation technique along with reducing about 6% of the model parameters. As a result, the value of compliance criterion increased from 53.27% in the initial model to 75.6%. Despite the significant lack of training data in the supervised learning process, the ability of the final model produced to identify rock discontinuities in the images has been acceptable and impressive.