During this study a solid method for the acquisition of digital micromorphological thin section images under PPL and XPL conditions was developped. Furthermore, it was proved that usual but time consuming stages of image processing and image segmentation can be omitted. The software Feature Analyst® is able to conduct supervised and unsupervised learning on soil thin sections without image processing techniques including batch processing for supervised learning. Nevertheless, training sets have to be digitized thoroughly as well as sample unit, sample size and sample method for the accuracy assessment. To answer the main research question: image recognition software is able to classify features in thin section images and the adaption of a remote sensing technique for micromorphology was successful. Quantitative analysis through image classification bares the potential to support micromorphological investigation especially in terms of area sizes and feature orientation. It is recommended to apply this technique to other thin sections to test the feasibility of other classifications e.g. for flint fragments, quartzes or other thin section features.
Moser, S. (2015): Towards automatic identification and quantification of features in thin sections - The application of an automated feature extraction model..
|Title||Towards automatic identification and quantification of features in thin sections - The application of an automated feature extraction model.|