First, you are unlikely to be able to use the command line tool for this.
The search for community will not be as easy as you think. The entire video is compressed and causes slight differences even in things that may look the same for the human eye.
If you know the main dimensions and area of the logo, you can limit your search to that part of the screen that should accelerate. After that, assuming that I know what a logo is, it subtracts one frame from the next. Do this by subtracting each color channel onto the corresponding color channel of the same pixel in the previous frame. If you have a lot of videos, some frames may be scattered, so that the slowly moving part of the image will not cause a false positive. As soon as you get this subtraction, everything that is the same should be close to zero. Everything should be much more valuable.
With this subtracted area, if there are a lot of pixels around zero, you probably have a logo, or at least something permanent. You may want to compose the results and then add them. This will exaggerate values that are not close to zero. Then select a threshold. If the values are lower than this, you probably have a logo.