🤖 AI Summary
Traditional pollen analysis is time-consuming (4–6 hours per sample) and highly subjective. This study proposes an automated, high-throughput microscopic analysis system that integrates brightfield imaging, $H_\infty$ robust mechanical control, and a deep learning pipeline to enable efficient, accurate counting, classification, and morphological characterization of pollen from the Bio Bío region of Chile. The approach combines U²-Net for salient object detection with a DINOv2 vision transformer classifier based on deep metric learning, augmented by a gradient-weighted attention mechanism to generate interpretable texture and diagnostic features. The system achieves a classification recall of 95.8% and processes samples six times faster than human experts.
📝 Abstract
Traditional melissopalynology is a time-consuming and subjective process, often taking 4-6 hours per sample. We present an automated, high-throughput microscopy system that integrates $H\infty$ robust mechanical control with advanced deep learning pipelines for the precise counting, classification, and morphological analysis of pollen grains from Bio Bio region in south central territory in Chile. Our system employs $U^{2}$-Net for salient object detection and a DINOv2 Vision Transformer backbone trained via Deep Metric Learning for classification. By integrating Gradient-Weighted Attention, the model provides human-interpretable texture and diagnostic feature annotations. The system achieves a 95.8$\%$ classification recall and a 6x processing speedup compared to manual expert analysis.