[Editor's Note: Poster is from VideoIQ]
VideoIQ isn't targeting slip-n-fall style analytics.
Currently our primary focus analytics-wise is on object classification in a diverse set of environments. These means picking up all the people and vehicles in a scene, while ignoring dogs and foliage and shadows (to make a simplistic summary). We have made our biggest market in complex outdoor environments, and are now doing more in interior scenes (eg: for things like retail analytics).
Once an object is classified, we can determine if it is breaking one or more "rules", which I often describe as "macro-level behaviors", essentially the things that can be done with high repeatability. For example, a person entering a secure region, or travelling in a prohibited direction, or loitering near a store display. There isn't a lot of "grey" area here, and as long as you can pick out the people in the scene, you can do the rule/behavior analysis part reliably.
There is this whole other slew of behaviors that are closer (IMO) to micro-level behavior analysis. "Threatening or suspicious activity" "Slip and Fall" "Shoplifting team" and so forth. Yes, SOME of these things can be detected SOME of the time, but it becomes much harder to have a system that is both easy to install/maintain AND very trustworthy (an alarm is a real event, and the absence of alarms is truly the absence of events). From what I've seen of various products on the market, I have not come across anything that operates with high reliability and consistency over multiple environments such that I'd want to base MY business on it.