"Computer Vision for Artists and Designers": Reading Response

The four simple detection and tracking methods outlined in the article are:

1. Detecting motion – ‘frame differencing’

2. Detecting presence – ‘background subtraction’

3. Detection through brightness thresholding – controlled illumination (e.g. backlighting) and/or surface treatments (e.g. high-contrast paints)

4. Simple object tracking – tracking the single brightest pixel in every frame of video

One of the major challenges that face programmers when working with computer vision techniques is that computers are not completely accurate when trying to tell if a person is present in video information. Since video is made of pixels, computers have to distinguish between pixels that resemble a person and pixels that make up background objects. This differentiation is often difficult, especially if a person’s clothing is similar to the background in which the video is shot. Different filming techniques are needed for computers to fully tell apart a person from the background. Techniques like using backlit objects and stark contrasts in clothing to background are good ways to ensure that computers can better distinguish objects from one another. Another major challenge that occurs in computer vision techniques is accurately measuring distance in video. Until recently with Microsoft’s Kinect, measuring distance in video feeds has always been a challenge, as many cameras could not accurately measure depth of field. Even with Kinect, the technology is still new, and judging how far an object is from the camera has not been perfected yet. Even though we have come a long way with innovations in technology, the camera lenses being used today are not perfect and can only be improved upon in the future. One more challenge that occurs with computer vision techniques is that it is hard to observe and detect a person if they are stationary and not moving. Some cameras depend on movement to judge if a object is present in the video feed, for example, the Suicide Box that looked for vertical movement by the Golden Gate Bridge. This also ties into camera lenses not being the greatest, as well as higher end lenses have a higher price point and most often, it is cheaper to use low end cameras for manufacturing purposes.