Computer Vision for Artists and Designers: Pedagogic Tools and Techniques for Novice Programmers

The four simple detection and tracking methods outlined by Levin are:
Directing motion
Detecting presence
Detection through brightness thresholding
Simple object tracking

One of the major challengers that novice programmers face when working with computer vision techniques is pattern and feature recognition. This can include the fact that computers can struggle to detect a person; maybe due to poor lighting, perhaps it needs to identify a certain person and it is confused. Recently playing Kinect, we found that if there is one person trying to navigate they were unable to do so if a person was standing closely behind them – the sensor is confused on who they are getting directions from and will not work. A way to fix this would be to have an empty background with no distractions for Kinect, in the larger sense however a neutral background that would not be able to be confused with people, clothing or anything of that matter. Another challenge programmers face is distance. The computer has trouble judging the depth of field and so struggles to give an accurate depiction of the scene. However technology is improving. Once again with the Kinect, my roommate and I were racing at a track and field game; we had to run on spot, about 8-10 feet from the television – so the sensor could sense both of us and our movement. Naturally your body starts to move forward as you run on the spot which resulted in the sensor losing sight of us and our avatars slowing down. A third and final challenge is speed and quality. With older webcams, a person was able to look away, then turn back to the screen and would then see themselves looking away on the playback. The pixels were also brutal, your face or the person you were talking to looked like a digital blob. Now webcams are a lot more accurate in regards to playback, as well the quality is a lot clearer.