The inhalation of fine and ultra-fine dust particles at the workplace and from the environment can pose risks to human health. For recognizing, assessing and reducing such risks it is necessary to determine reliably the composition of dusts and the concentration of critical components. For this purpose, this research project uses computer-based methods to identify and quantify individual dust particles. This requires a combination of several imaging and chemical-analytical microscopy techniques that provide information on both particle shape and composition.
The starting point are electron-microscopic images of filter surfaces that collected airborne particles at workplaces and other indoor areas. Particles localized on filter images are characterised morphologically and spectroscopically with respect to shape, elemental composition and chemical properties in order to assign them to known particle groups. For this step also methods of machine learning are to be developed that will have previously been trained on reference dusts of known composition. An additional objective of the project is to enable identifying and quantifying selected types of particles and fibres of health relevance.
The developed algorithms will be published and implemented a software. Both automatic image generation as well as automatic image analysis and classification enable the evaluation of a large number of images and spectra. This is of particular interest to research institutions and analytical laboratories that would otherwise require considerable personnel for this task.
Unit 4.5 "Particulate Hazardous Substances, Advanced Materials"