From Habr (Discussions on Crime Tracking System):

The most interesting thing is how she will recognize crimes. Let’s say the banal hop-stop “uncle, let me smoke.”

YYY: I think one of the main models of crime detection:
- transition to running (it is not natural for the city)
- acceptance by a person of a horizontal position (a person to stand out on a constant background - elementary, well, and the fact that he fell to the ground - will also not be difficult)
Not a naturally loud noise for a particular point of observation.
- detection of weapons in the hands (both firearms and cold, well, if at night in the hands of the beat - obviously not for the game.)
This is what first came to mind, and there and before the statistical analysis is not far away - as holds a particular object, the frequent flashing of one person in the field of observation. and etc. and etc.

zzz: Falling baseball player who runs home from the stadium in the evening with a beat behind his shoulders will be sentenced to 10 years :)