After generating An Average Day on the Bibbulmun Track I started wondering what a similar process would generate on more random images; say, photographs of all the visitors to my house.
Obviously, it's a huge jump from a very regulated set of portraits that were manually aligned and therefore produced a much firmer image – to a possibly very randomly positioned set of images (but in my mind that was one of the interesting beauties of this project: averaging different people and their mannerisms would produce what?): so I needed to do a wee test to see what actually came out, if it was possible.
I fixed the security camera pointing down our hall to the back door (which is our main doorway). The camera is programmed to record a video when it detects motion within the door frame. These videos are stored on a NAS. Most of videos are deleted as they're most often just of us coming and going, but the ones of people coming to visit, or of us coming home from being out, are saved for later processing.
The appropriate frame is exported from the video as a snapshot. It would have been nice to use the camera's snapshot function, but the timing of the actual shots was not accurate enough and the subjects were not always located at the same point.
The aim is to get an image of each visitor at approximately the same location (in this case, just inside the threshold).
The snapshot images are sorted into three groups:
and an AHK script uses ImageMagick to generate 5 images at each processing:
and uploads the web versions to the web server (which you're viewing now).