Change detection is the process of taking two geo-registered images of the exact same area from two different dates and automatically identifying differences. It is really just another form of image classification. Just like our previous classification examples, it can range from trivial techniques like those used here, to highly-sophisticated algorithms that provide amazingly precise and accurate results.
For this example, we'll use two images from a coastal area. These images show a populated area before and after a major hurricane, so there are significant differences, many of which are easy to visually spot, making these samples good for learning change detection. Our technique is to simply subtract the first image from the second to get a simple image difference using NumPy. This is a valid and often used technique.
The advantages are it...