3D laser scanning application

3D scanners of the type used for architectural reconstruction don’t quite meet my definition of “machine vision” – there’s no extraction of information from the data points captured – but it’s still a fascinating technology.

A story on the BBC news website, “Road crash laser scanners to save millions of pounds,” (July 9^th, 2011,) explains how British police will be using scanner technology to capture road crash scenes. As the article is written for the lay reader it omits how the point cloud data will be used, but it’s always fun to speculate.

My guess is that the data will be merged to create 3D views of an accident scene. An investigator trying to determine how vehicles ended up in particular locations will be able to swoop in and around the scene, looking for clues from any and every direction. Clearly, this has the great advantage, besides getting the roads open quicker, of allowing investigators to go back and take a second look as hypotheses as to the accident cause emerge.

I do however see two problems. First, there’s the question of this evidence being admissible in court. It will probably take a test case or two to work that out. Second, there’s what Nassim Nicholas Taleb would call “the ice cube problem.”

What’s that, you ask. Well imagine we see a puddle of water and are told it came from ice cubes that have melted. Can we reconstruct the positions of the ice cubes before they melted? Only within very broad limits. And that will keep the lawyers employed for a long time.