A city wants to estimate how many people use a public plaza at different times. It is considering cameras capable of identifying individual faces. Which approach best follows data minimization?