Proof of Concept: Autonomous Machine Vision Software for Botanical Identification

Area of Science:

• Botany
• Computer Science
• Analytical Chemistry

Background:

• High-Performance Thin-Layer Chromatography (HPTLC) is standard for botanical identification.
• Current methods rely on subjective visual comparison of HPTLC images.
• This subjectivity can lead to inaccuracies in identifying plant species.

Purpose of the Study:

• To evaluate machine vision and machine learning for automated botanical identification.
• To develop an AI system using HPTLC image data.
• To overcome limitations of subjective human analysis in botanical identification.

Main Methods:

• Generated synthetic HPTLC datasets using a deep conditional generative adversarial network.
• Trained a deep convolutional neural network (CNN) on synthetic and real HPTLC data.
• Validated the CNN model's accuracy and speed in identifying botanical species.

Main Results:

• The AI system successfully generated realistic synthetic HPTLC images.
• The trained CNN achieved high accuracy in identifying Ginger and related species.
• The machine vision system demonstrated feasibility as a proof-of-concept.

Conclusions:

• An automated machine vision system for botanical identification using HPTLC is feasible.
• The developed AI model provides accurate and rapid identification, reducing subjectivity.
• This technology offers significant time and resource savings in botanical analysis.