Pollen Detection Gets a Tech Upgrade

Scientists have turned to a new imaging trick called evanescent wave scattering microscopy, or EWSM, to spot pollen in the air. This method lets them see tiny features on pollen grains—like spikes and holes—that regular bright‑field microscopes miss. Using this detailed view, researchers built a two‑step AI system to find and sort pollen. In the first step, they tested three computer models—YOLOv8n, YOLOv8l, and RT‑DETR—to locate pollen. The best performer was RT‑DETR, catching 82% of the true pollen grains. For the second step, they fed the images into an EfficientNetB0 network to decide which species or group each grain belonged to.

To improve accuracy, experts added a “human‑in‑the‑loop” process in two rounds. The first round lifted the overall score from 0. 28 to 0. 54 for 41 different pollen types, with an 80% recall rate. The second round combined those into 19 shape‑based groups, raising precision to 0. 65 and recall to 0. 74; the combined F1‑score hit 0. 69. By tightening detection thresholds, they cut false alarms by more than six‑tenths. This new framework shows that it’s possible to run automated pollen analysis in busy cities, giving health officials a sharper tool for allergy warnings as weather patterns shift.

Actions