Determination of a Person's Height from Images Using a Known Object Size

This paper describes an image processing method for an automatic determination of a person's height from image captured by a single camera. It is supposed that at least one object with a known size which defines the metric calibration is present in the image. Our global algorithm consists of several image processing techniques such as camera calibration, edge line detection, Hough transform, vanishing points detection and Histograms of oriented gradients for human detection. Original methodology was developed and tested in C++ as a standalone process which is applicable only on static frames (not optimized for video). The accuracy of the method is about 2.25± 2.79 cm, preliminary results were obtained using a database which contained 19 images of standing persons inside a room.