Automatic 3D Human Body Modelling

Abstract

Objective 3-layers adaptive system that detects, extracts, analyzes and models human bodies existed in video stream, it uses different detection methods in parallel and evaluates each detection result according to correctness measurements, and controls method’s work and priorities according to dynamic parameters; these parameters are updated continuously depending on the system’s statistics. The system gathers the information gradually; builds complete model based on already existed and trusted results, and waits for the missing information to update. Results introducing effective way to use different detection methods in parallel and control them adaptively, then make decision based on these unrelated sources, reducing the ambiguity in high complex environments by depending on all features of the human body, which leaded to more accurate and trustful results. Practical importance The automatic 3D human body modeling system is presented; this system takes RGB video stream input, and extract the human body inside the video, then model this body with no depth information. This system can be used as an electronic vision system for mobile robots.