Detecting Microbursts with Machine Learning
The Project
A few years ago I woke up and trees and powerlines were down in my neighborhood. No alarms went off and no one knew what happened. We found out later it was caused by a microburst. Microbursts are hard to detect because they can occur between radar scans. I wanted to use machine learning to detect microbursts from the damage caused to trees, powerlines and streetlamps. I found images of damage caused by microbursts and trained a model. Then I tested the model with other images I found online. Two microburst images were identified as no damage. So I retrained the model by adding more images of microburst damage. I made a machine learning program to detect microburst damage and installed it on a Raspberry Pi. I tested the program and found 100% success in detecting microbursts. I added a camera to the Raspberry Pi and added the code to capture images and run it through the machine learning program to detect microburst I tested it and found it can also detect microburst damage. I found I can use machine learning to detect microbursts!
