Automated Palynological Analysis System: Integrating Deep Metric Learning and $U^{2}$-Net Detection in $H\infty$ bright field microscopy

📅 2026-04-17
📈 Citations: 0
Influential: 0
📄 PDF
🤖 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.

Technology Category

Application Category

📝 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.
Problem

Research questions and friction points this paper is trying to address.

melissopalynology
pollen analysis
automated microscopy
high-throughput
subjective identification
Innovation

Methods, ideas, or system contributions that make the work stand out.

Deep Metric Learning
U²-Net
DINOv2 Vision Transformer
Gradient-Weighted Attention
H∞ robust control
💼 Related Jobs
No related jobs found.
J
J. Staforelli-Vivanco
Departamento de Física, Facultad de Ciencias Físicas y Matemáticas, Universidad de Concepción, Concepción, Chile
R
R. Jofré
Departamento de Física, Facultad de Ciencias Físicas y Matemáticas, Universidad de Concepción, Concepción, Chile
B
B. Muñoz
Departamento de Física, Facultad de Ciencias Físicas y Matemáticas, Universidad de Concepción, Concepción, Chile
V
V. Salamanca
Departamento de Física, Facultad de Ciencias Físicas y Matemáticas, Universidad de Concepción, Concepción, Chile
P
P. Coelho
Facultad de Ingeniería, Universidad San Sebastián, Concepción, 4080871, Chile
I
I. Sanhueza
Facultad de Ingeniería, Universidad San Sebastián, Concepción, 4080871, Chile
L
L. Viafora
Facultad de Ingeniería, Universidad San Sebastián, Concepción, 4080871, Chile
C
C. Toro
Facultad de Ingeniería, Universidad Andrés Bello, Talcahuano, CCP-THNO 7100 Chile
J
J. Troncoso
Universidad Arturo Prat, Sede Victoria, Chile
M
M. Rondanelli-Reyes
Laboratory of Palynology and Plant Ecology, School of Sciences and Technologies, University of Concepción, Los Angeles Campus
I
I. Lamas
Laboratory of Palynology and Plant Ecology, School of Sciences and Technologies, University of Concepción, Los Angeles Campus