As you move through a room looking for, say, your keys, you turn your head around at different angles and scan surfaces that might contain that object.

 

If the object has low contrast with the wall behind it, or is a certain distance away, or is located only in your peripheral vision, your probability of detecting the object decreases. There are many papers mapping these relationships, thanks to a field of study called psychometrics.

 

I modeled these human search processes (and more) in MatLab by finding a per-view probability of detection at many points in the room. Using this, we can tell which set of views give us the best probability of detecting the object no matter where the object is in the room, and generate the optimal path through a room. Based on how a searcher actually chooses to search the room, we can see how skilled that searcher is.