Rock Mass Discontinuity Determination With Transfer Learning

Abstract

Abstract. Rock mass discontinuity and orientation are among the important rock mass features. They are conventionally determined with scan-line surveys by engineering geologists in field, which can be difficult or impossible depending on site accessibility. Photogrammetry and computer vision techniques can aid to automatically perform these measurements, although variations in size, shape and appearance of rock masses make the task challenging. Here we propose an automated approach for the detection of rock mass discontinuities using deep learning and photogrammetric image processing methods. Two deep convolutional neural network (DCNNs) were implemented for this purpose and applied to basalts in Kizilcahamam Guvem Geosite near Ankara, Türkiye. Red-green-blue (RGB) band images of the site were taken from an off-the-shelf camera with 1.7 mm resolution and a 3D digital surface model and orthophotos were produced by using photogrammetric software. The discontinuities were delineated manually on the orthophoto and converted to masks. The first DCNN model was based on the open-source crack dataset consisting of a total of 11,298 road and pavement images, which were used to train the Resnet-18 model (Model-1). The second model (Model-2) was based on fine-tuning of Model-1 using the study data from Kizilcahamam. After fine-tuning, Model-2 was able to achieve high performance with a Jaccard Score of 88% on the test data. The results show high potential of the methodology for transfer learning with fine-tuning of a small amount of data that can be applied to other sites and rock mass types as well.

A Novel Rock Mass Discontinuity Detection Approach with CNNs and Multi-View Image Augmentation

“…Yet, manual interpretation of the data from scan-line surveys to detect and measure discontinuities in rock masses has the major drawbacks of requiring expertise and being time-consuming. On the other hand, while deep learning methods, particularly CNNs for image segmentation and classification, provide promising results for discontinuity detection (see Yalcin et al [27,28]), they are also limited by the manual labeling required to obtain the necessary amount of data for learning the model parameters. Data augmentation techniques and transfer learning approaches can help overcome this obstacle.…”

“…On the other hand, using conventional edge operators such as Canny and Sobel for image-based discontinuity detection suffers from a high level of noise in identified edges due to the color variations and textural characteristics of rock surfaces (see Lee et al [25]). Deep learning models, especially CNNs, were used for this purpose by Yalcin et al [27,28]. The preliminary results presented by Yalcin et al [27] in the Kızılcahamam/Güvem Basalt Columns Geosite show that using orthophotos with data augmentation as the input to a U-Net architecture yielded an F1-score of 58%.…”

“…However, due to the nature of the problem, the identified discontinuities contain a high level of noise sourced from topographic variations and the presence of different textures on rock surfaces [25]. Convolutional neural network (CNN) architectures have been widely used for various image processing, feature extraction, and object segmentation tasks (e.g., see Qiu et al [26] and Yalcin et al [27][28][29][30]), including edge detection. However, they require large amounts of data and computational resources for model training.…”

“…However, they require large amounts of data and computational resources for model training. Yalcin et al [27,28] detected rock discontinuities on orthophotos using a CNN, and they emphasized that blurring was observed at the discontinuities due to the inferior quality of the digital surface model (DSM), especially at locations with poor illumination and shadow. On the other hand, detecting discontinuities in geometrically unprocessed (raw) images requires the use of multiple images with different viewing angles to obtain the 3D position information and ensure model completeness.…”

“…Domain adaptation, a specific type of transfer learning, occurs when related tasks have different data distributions [32]. We investigated the domain adaptation of CNNs pre-trained with crack data in our process as these approaches have recently proven to be successful in increasing model performance with a small amount of data being used for fine tuning (Yalcin et al [27,28,30]). A practical application was carried out on basalts in Bala town of Ankara using data acquired with an RPAS.…”