Support vector machines and convolutional neural networks applied to blasthole dislocation detection in open pit mines APCOM 2021

The knowledge of the final location of a drillhole is critical in obtaining the expected blasting results. This paper proposes a new methodology to control the drilling accuracy in open-pit mines using aerial images and machine learning. Specifically, it investigates how support vector machines (SVM) and convolutional neural networks (CNN) are useful for automatically detecting blastholes over drone images captured in blasting patterns. Several datasets were obtained from different mines in Nevada (USA) and were used to train and test models. We then processed those images to create orthomosaic and digital elevation representations of the patterns. Thousands of patches were then extracted and augmented to develop SVM and CNN models. The results demonstrate that machine learning is useful for drillhole detection and, in this way, to control the drilling accuracy and use the explosive energy more efficiently.