Automated Pollen Detection with an Affordable Technology
Airborne pollen cause seasonal allergies and the number of people affected increases yearly due to global warming and urbanization. Governmental pollen sensing stations are sampling traps which require manual pollen identification and counting by trained personnel in the lab. In the past years, a number of researchers and startups started working towards automated pollen measurements by exploring a wide range of techniques. Many solutions reported in the literature are expensive or work for a limited number of pollen species. In this paper, we present the design of a prototype of an automated and affordable pollen detection device built from off-the-shelf components.
The design consists of three subsystems operating in the field and communicating the data to the backend server: (1) a particle trap with automatic filtering, (2) a particle concentration subsystem, and (3) a digital transmitted light microscope with layer-wise focus. The prototype shows effective particle gathering, filtering and concentration in a tiny-sized area. As a result, we reduce particle loss and improve image quality taken by the optical system when searching and focusing on pollen grains. The test results show that our device achieves high efficiency with up to 150 l/min air flow rates, evaluates over 90 % of captured pollen grains, and achieves 1 h measurement delay on average (2 h at maximum).
The prototype collects raw time-stamped microscopic images of pollen with 5-60 depth layers per sample depending on the number of objects contained in one sample. All images are transmitted to the backend server where we run a pollen detection algorithm to extract individual pollen grains from every image. We achieve 0.90 average precision and F1-score of 0.88 when detecting pollen in the images of individual layers taken in the field. Our prototype successfully operated in the wild for 115 days between April and August 2019, and shows high stability under a wide range of varying weather conditions, little maintenance need and low device-to-device variation.
N. Cao, M. Meyer, L. Thiele, O. Saukh, Automated Pollen Detection with an Affordable Technology, In Proc. of the International Conference on Embedded Wireless Systems and Networks (EWSN), 2020