For the identification complexity of rock image classification, based on a certain cut-off grade, an automatic classification recognition of rock images method is proposed in this dissertation, which is a part of the MSc in Data Science at the International Hellenic University. The main topic of this research was the identification of Rock image Classification through images in order to use the results for further decision making. Basis of specific approach in face mapping images taken by mine geologists for traditional rock-mass characterization. Digital grey image processing of rock face mapping images is used for features extraction. Then features contains the process of knowledge extraction from images in Matlab and furthermore as the classification procedure in Weka, in which different classifiers have been trained and tested in order to classify Ore or Waste. Finally, the model output is the rock image classification. Hellas Gold Company, subsidiary of Eldorado Gold provided a sample of 600 images for the case study. Specifically face mapping images are from Olympias mine located in North-East Chalcidice Prefecture, Greece. For the experiment, the dataset is divided into 300 images as a training dataset in order the algorithm to classify the >15% Ore and 300 images for classify the <15% of Waste. As an outcome the optimum classifier reached 98.5% accuracy for automatic identification of rock face mapping image. Therefore, the proposed method for improving geological pattern is effective and can result in accepted identification performance for rock image classification quickly and accurately. Nevertheless, the target is to reach an autonomous level such that no human intervention will be necessary.
Collections
Show Collections
