COMPARING DIFFERENT QUANTITATIVE PETROGRAPHY TECHNIQUES USING IMAGES FROM PARTIALLY DOLOMITIZED SAMPLES

During the Mio-Pliocene, there is an interesting phenomena of dolomite formation that is currently being investigated. The abundance of dolomite that occurs within the Neogene is being tested for its co-occurrence and predictability within red algal facies. Presently, the use of free tools such as an open-source imaging software (Fiji), specifically the Weka segmentation application, is easy to apply. In this project we aim to determine the percentage of dolomitized components within each sample using thin section images. Previous methods of quantitative petrography (point counting) were widely used and for this specific problem it is utilized to assess samples from Bonaire Island, in the Caribbean. While commonly used, point counting is subject to human error, is extremely time consuming, and analyzes the image on a large scale not on a pixel level. This can cause inaccuracies and miscalculation of the true representation of components in the sample. However, using Weka segmentation, individual classes can be assigned and the user chooses between three to ten features on the image for each class then runs the software’s algorithm to finish processing the image. Once the processing is complete, the image appears on the screen color coded for each class the user gave it. Depending on computational resources and image resolution, the entire process can take between ten to fifteen minutes to complete as compared to the hours it may take to process an image by point counting. Faster results allow the user to fix any errors immediately. The results using Weka segmentation will be compared with the previous method of point counting for accuracy. Problems may arise using Weka, but those will be assessed and dealt with as they appear. The method of point counting will be tested at three levels of expertise to determine the time it takes to process an image and the accuracy at each level. Results from this experiment will be compared to those of Weka to show differences in time and in accuracy. Weka is expected to be more accurate and less time consuming to process images during the experimentation. These results will help determine the percentage of dolomite within the given samples and the methods can be used for further investigation.