Localization is one of the key challenges in building and deploying mobile robots. The quantitative measure of the robot position in a co-ordinate system is important in a variety of operational functions - delivery, path adherence, outdoor as well as indoor navigation, and many related activities. Compared to proximity sensors, cameras have several desirable properties in the quest for effective localization:

They are low-cost yet provide a tremendous amount of information.

They are passive. Vision-based navigation systems do not suffer from the interferences often observed when using active sound or light-based proximity sensors.

Moreover, if robots are deployed in populated environments, it makes sense to base the perceptional skills used for localization on vision like humans do.