Machine learning and fire line detection and vegetation classification

Can algorithms and clusters of pixels be used on imagery to automatically draw a fire line or other interesting features during an incident?

1. What image processing technologies would be used?
2. As an alternative to labeling thousands of images to train a traditional machine learning model, can you develop an effective ML model coupled with an efficient imagery processing workflow for fire detection?
3. Other interesting features could consist of dead, or drought stressed trees, vegetation types, and more.
4. Bonus: integrate with mapping technologies like ArcGIS. CAL FIRE heavily uses enterprise ArcGIS so any technologies that consider connecting with ArcGIS is a plus.