Frame by Frame Video analysis


    Jun 10 2006 | 4:28 pm
    Hi Im trying to figure out how to create a patch that analyses every 10 FPS for any color disturbaces coming from a stationary video camera input.
    Ex. Trying to monitor motion in a space with a ceiling mounted video camera where the patch first recognizes a presence by color disturbance, and then recognizes then that disturbance STOPS moving in the monitored environment.
    If anyone could help me with this it would be greatly appreciated. Thanks.

    • Jun 11 2006 | 10:53 am
      here is a simple one. it uses cv.jit and jit.m3
      it outputs center of mass, direction of general motion and on/off
      there is anybody in the area. it can be expanded.
      look into the cv.jit library as it is full of examples and possibilities.
      as before. computer vision happens between the environment (light and
      lens) and the code. a webcam is easy and cheap but lacks wide lenses,
      manual control over whitebalnace and infrared capabilities (search for
      "webcam infra" in google, which makes it easier to control light).
      simple vision solutions (like this one) are VERY site specific , don't
      expect a setup to work outofbox in diffrent locations. calibration and
      testing is a must. beware of changing light environment, windows and
      even projection can change your light conditions. (infra is blind to
      projected video but very sensitive to sunlight).
      goodluck and enjoy
      max v2;
    • Jun 11 2006 | 11:00 am
      there is nothing simple about the LK or centroids block, only in the
      patch wacking ;)