Results

Method 1:

Our first method (using edge detection and the distance between two edges to see if it could be the tank) didn't work as well due to the fact that we had lots of edges and we had problems telling whether the two edges we had was just two objects not too far apart or if it was actually the tank. We also had problems because it's hard to detect the tank when it was at an angle.

Method 2:

Our second method worked very well, however. Here are some of the results:

Duncan Original Image:

Duncan Output:

Brick Original Image:

Brick Output:

Dirt Original Image:

Dirt Output:

Problems:

The only real problem we had (which we expected to happen) was that if we had the tank hidden by an object (for example, a tree) we had no way to detect if the tank was there or not.

An example of a hidden tank:

But overall our method worked really well. Later on, it could be expanded to include an automatic background detector (since our code requires the user to input the background due to the threshold changes). Also, we could expand our code to recognize other objects (such as the helicopter, the tank barrel, etc).

Names of images tested:

We tested a total of 11 images, only one of which did not work (due to the hidden tank). We tested four dirt images (Dirt1.jpg-Dirt4.jpg), four Duncan hall images (Duncan1.jpg-Duncan4.jpg), and three brick images (Brick1.jpg-Brick4.jpg). We were extremely pleased with our results.