Multi-Input Convolutional Neural Networks for Automatic Pollen Classification

Mihai Boldeanu,Corneliu Burileanu,Luminița Mărmureanu,Horia Cucu

doi:10.3390/app112411707

Abstract

Pollen allergies are a cause of much suffering for an increasing number of individuals. Current pollen monitoring techniques are lacking due to their reliance on manual counting and classification of pollen by human technicians. In this study, we present a neural network architecture capable of distinguishing pollen species using data from an automated particle measurement device. This work presents an improvement over the current state of the art in the task of automated pollen classification, using fluorescence spectrum data of aerosol particles. We obtained a relative reduction in the error rate of over 48%, from 27% to 14%, for one of the datasets, with similar improvements for the other analyzed datasets. We also use a novel approach for doing hyperparameter tuning for multiple input networks.

Highlights

Pollen is a powdery substance produced by plants during their reproductive season
In respect to SAUSRB, we reduced the error rate from 26% to 23% and, compared to SAU-CH, from 20% to 15%
With the fine tuned architecture, compared to the initially proposed architecture, we obtained a relative reduction in the error rate of 13%, from 23% to 20%, for SAU-SRB; of 12%, from 16% to 14%, for SAU-LI; of 13%, from 15% to 13%, for SAU-CH; and of 20%, from 24% to 19%, for Magurele center for Atmosphere and Radiation Studies (MARS)

Summary

Introduction

Pollen is a powdery substance produced by plants during their reproductive season. While pollen is essential for plant survival and reproduction, certain species, especially those that use wind for pollen dispersal, are the cause of most seasonal allergies in humans [1].The first reports of pollen as a possible cause for several illnesses date back to the beginning of the XIX century with researchers such as Bostock and Blackley [2]. Pollen is a powdery substance produced by plants during their reproductive season. Blackley experimented with several instruments to sample the air for pollen. He exposed glass slides coated in a glycerin mixture for a period of 24 h after which he analyzed them using a microscope. This was performed daily from April to August to record the number of pollen grains captured. These first experiments are the basis for what the field of palynology has researched for the following 150 years

Methods

Results

Conclusion