14 Aug 2023
 | 14 Aug 2023
Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

Real-Time Pollen Identification using Holographic Imaging and Fluorescence Measurement

Sophie Erb, Elias Graf, Yanick Zeder, Simone Lionetti, Alexis Berne, Bernard Clot, Gian Lieberherr, Fiona Tummon, Pascal Wullschleger, and Benoît Crouzy

Abstract. Over the past few years, a diverse range of automatic real-time instruments has been developed to respond to the needs of end users in terms of information about atmospheric bioaerosols. One of them, the SwisensPoleno Jupiter, is an airflow cytometer used for operational automatic bioaerosol monitoring. The instrument records holographic images and fluorescence information for single aerosol particles, which can be used for identification of several aerosol types, in particular different pollen taxa. To improve the pollen identification algorithm applied to the SwisensPoleno Jupiter and currently based only on the holography data, we explore the impact of merging fluorescence spectra measurements with holographic images. We demonstrate that combining information from these two sources results in a considerable improvement in the classification performance compared to using only a single source (balanced accuracy of 0.992 vs. 0.968 and 0.878). This increase in performance can be ascribed to the fact that often classes which are difficult to resolve using holography alone can be well identified using fluorescence and vice versa. We also present a detailed statistical analysis of the features of the pollen grains that are measured and provide a robust, physically-based insight into the algorithm’s identification process. The results are expected to have a direct impact on operational pollen identification models, particularly improving the recognition of taxa responsible for respiratory allergies.

Sophie Erb et al.

Status: open (extended)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse
  • RC1: 'Comment on egusphere-2023-1572', Anonymous Referee #1, 04 Sep 2023 reply

Sophie Erb et al.

Sophie Erb et al.


Total article views: 416 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
267 138 11 416 10 5
  • HTML: 267
  • PDF: 138
  • XML: 11
  • Total: 416
  • BibTeX: 10
  • EndNote: 5
Views and downloads (calculated since 14 Aug 2023)
Cumulative views and downloads (calculated since 14 Aug 2023)

Viewed (geographical distribution)

Total article views: 392 (including HTML, PDF, and XML) Thereof 392 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
Latest update: 30 Sep 2023
Short summary
In this study, we focus on an automatic bioaerosol measurement instrument and investigate the impact of using its fluorescence measurement for pollen identification. The fluorescence signal is used together with a pair of images, from the same instrument, to identify single pollen grains via neural networks. We test whether considering fluorescence as a supplementary input improves the pollen identification performance by comparing three different neural networks.