Human height analysis using multiple uncalibrated cameras

Abstract

This paper presents a human feature retrieval algorithm using multiple uncalibrated cameras. In order to automatically synchronize multiple cameras, the proposed human height analysis algorithm consists of three steps: i) line extraction from the background, ii) estimation of the vanishing points using the extracted lines, and iii) camera calibration for human height estimation. The proposed method can robustly calibrate multiple cameras based on the background and pedestrians. In multiples frame of video acquired from the multiple uncalibrated cameras, the normalized information of pedestrian is calculated by using the camera calibration camera parameters.