The ability of measuring accurately airborne pollen concentration in the environment is an important goal for palynology. It has been unsatisfactory for agile usage to date. Huge volumes of airborne particles prevent palynologists from opportunely processing statistically suitable information. Additionally, measurements from stationary pollen monitors cannot be accurately associated to individuals. In the context of computer vision, this paper presents the outline for the structure of an image based pollen detection system, under the framework of the Personalized Pollen Profiling and Geospatial Mapping project based on individual information of allergic patient profile measured at multiple points from personal mobile wearable devices. With features from classical geometric and optical measures to specialised palynological information, it is feasible to characterise different pollen taxa images. Optimal selection of features allows a pattern detection system to split and recognise each taxon exactly. Gathering such accurate individual pollen concentration data in a geo-spatial context will benefit not only to patients through more precise medication but also to the improvement of pollen distribution models and forecasting.
