Overview on automatic detection of human body

Abstract

Automatic detection of human in a video sequence is a canonical instance of object detection. It's considered as a nonrigid object; it has many appearances at different perspectives. Different approaches are used by several methods to combine what is specific to the pedestrian detection, and what is common to the object recognition. A robust solution to this problem would have numerous applications to design smart systems, including video surveillance, power management, vehicle driver assistance systems, and evaluation of advertising policies. In this article we will discuss the main techniques of human (pedestian) detection on video sequence. Especially those how are learning-based methods. Then we describe in general terms the type of information that can be extracted once the detection is performed.