Pedestrians generally tend to visit public places to perform activities such as buying items, meeting friends, buying food at food outlets and many more. All these pedestrian activities tend to exhibit considerable continuity of motion. The corresponding motion patterns are similar for the same activity irrespective of the pedestrian. For example, pedestrians who are going to meet come closer irrespective of the pedestrians who are going to meet. The motion patterns are represented by the underlying motion parameter variations such as distance of travel, direction of travel and many more. Such motion parameter variations could be learned to predict pedestrian activity. This research work aims at to learn the motion parameter variations and identify methods to predict potential customer-approach to food outlets and to predict of possible future pedestrian groups. Supervised machine learning techniques such as Support Vector Machines are utilized to learn the motion parameter variations which lead to the pedestrian activity.

More information on this research work will be added after the corresponding publications is published.