Graduate Degree Type
Applied Computer Science (M.S.)
School of Computing and Information Systems
Dr. Jonathan Leidig
Dr. Greg Wolffe
High-Performance Thin-Layer Chromatography (HPTLC) is a widely used and accepted technique for identification of botanicals. Current best practices involve subjective comparison of HPTLC- generated images between test samples and certified botanical reference materials. This research project was designed to evaluate the potential of cutting-edge computer vision-based machine learning techniques to automate the identification of botanicals using native HPTLC image data. As there is very little relevant chromatographic data available, the first step involved developing a deep conditional generative adversarial network and presenting it with a small set of HPTLC images from Ginger, its closely related species, and common adulterants in order to create a large, synthetic dataset. This synthetic dataset was then used to train and validate a deep convolutional neural network capable of automatically classifying new HPTLC image data. Performance of both neural networks was evaluated using appropriate loss functions as a measure of their performance during learning. Validation of the overall system was measured via the accuracy of the learned system against real, previously unseen HPTLC data. The resultant machine vision system achieves high- accuracy identification (>95%) of HPTLC images corresponding to Ginger and six related species.
Stern, Nathan, "Autonomous Machine Vision Software for Botanical Identification and Adulteration Detection" (2023). Culminating Experience Projects. 387.
Available for download on Monday, June 10, 2024