Observing human behavior through worldwide network cameras.

This chapter presents an overview of the opportunities from video data available in worldwide network cameras and a research tool, called CAM² (Continuous Analysis of Many CAMeras), for obtaining and analyzing visual data at a large scale. It starts by providing an overview of computer technologies for analyzing human behaviors from video data, followed by a discussion of how various human behaviors are classified and the challenges in creating analysis software. The chapter uses several examples as sources of data (i.e., network cameras) from which it would be possible to observe human behaviors. It then introduces a software infrastructure built at Purdue University, called CAM², as a general-purpose computing platform for retrieving and analyzing visual data from many sources across the globe. Behavioral researchers use video cameras to capture contextual information, human facial and body information, and social interactions.