We present the first validation of the Swisens Poleno, currently theonly operational automatic pollen monitoring system based on digitalholography. The device provides in-flight images of all coarseaerosols, and here we develop a two-step classification algorithm thatuses these images to identify a range of pollen taxa. Deterministiccriteria based on the shape of the particle are applied to initiallydistinguish between intact pollen grains and other coarse particulatematter. This first level of discrimination identifies pollen with anaccuracy of 96 %. Thereafter, individual pollen taxa arerecognized using supervised learning techniques. The algorithm istrained using data obtained by inserting known pollen types into thedevice, and out of eight pollen taxa six can be identified with anaccuracy of above 90 %. In addition to the ability tocorrectly identify aerosols, an automatic pollen monitoring systemneeds to be able to correctly determine particle concentrations. Tofurther verify the device, controlled chamber experiments usingpolystyrene latex beads were performed. This provided referenceaerosols with traceable particle size and number concentrations in order toensure particle size and sampling volume were correctly characterized.
展开▼
