2019 · 12 citations · 6 references
ComputingEarth SciencesGeologyComputer VisionEngineering GeologyFeature EngineeringImage ClassificationObject RecognitionRemote Sensing Image AnalysisRock ClassificationSupport Vector MachinesAutomatic ClassificationGeological Feature RecognitionRock Type ClassificationSupport Vector Machine ClassificationSvmsThin Section Analysis
For an efficient and flexible access to the rock classification with features extracted from rocks casting thin section images, a SVM based approach is proposed for automatic classification of sand rocks. A PCA procedure is used to enhance correlation between features and reduce dimensionality of feature space. The relationship between feature spaces and sand rock types can be accessed with SVM. 400 images from Ordos basin are separated into two sets consisted by 300 and 100 samples respectively for training and testing of SVM to prove its availability and reliability. Classification of testing set result shows that this SVM classifier provides over 97.0% classification accuracy, which presents a good prospect in practical usage and helps geologists to authenticate casting thin section.
6