| Abstract |
In Ireland, 13,000+ skin cancer cases are diagnosed yearly, making it the country’s most common cancer, with 90% caused by UV exposure. UV radiation leads to short and long term effects such as sunburn, eye damage and cataracts. To help people protect themselves, the UV index measures the intensity of ultraviolet radiation from the Sun at a specific time and place. We investigated whether this could be improved and found that the current formula only considers cloud cover percentage as a cloud factor. However, research shows that cloud types classified by height, shape, and thickness affect the severity of UV radiation. We hypothesised that incorporating cloud classification into the UV index would yield a more accurate measure of UV levels, providing better protection guidance for individuals than the legacy UVI model. To test our hypothesis, we created a cloud type classification model using photos we took from a Raspberry Pi on the ISS. We also created a ground based camera model using Raspberry Pi HQ camera with a coral ML accelerator. Finally, we used multiple UV sensors on our ground based raspberry pi to measure the real world UVI value.
Using a Python Flask Web App, we graphed the UVI sensor value, our satellite model value, ground based model value and external UVI forecast from Meteo.com which uses the legacy UVI prediction. We found our cloud classification satellite model to be constantly closer to the UVI sensor value and thus a better model for predicting UV exposure.
|