The Fall Incident Surveillance System Design Based on Human Contour Features

• Main Authors: Zong-Sian Li, 李宗賢
• Other Authors: Ming-Hwa Sheu
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/70645501134946364463

碩士 === 國立雲林科技大學 === 電子與光電工程研究所碩士班 === 101 === Fall incident is going to be a crucial problem to an elder today because it usually causes serious injury. In many countries, the unintentional injury is one of main reasons for elderly deaths. Facing with the growing of aging population, many developed countries including Taiwan need to establish new healthcare systems to ensure the safety of elderly people at home. Computer vision provides a promising solution to analyze personal behavior and detect certain unusual events such as falls. In this thesis, a new method is proposed to detect falls by analyzing convex feature of human contour and motion detection of human corner during a video sequence. The convex of human contour technique is used to track the distribution information of human shape. The motion detection is used to track the vector similarity of corners over two adjacent frames. Finally, falls are detected from quick action and inaction period by using convex feature and motion feature. In this thesis has been conducted on a realistic data set of daily activities and simulated falls, and gives better results with 4% error rate compared to other common image processing methods for “Multiple cameras fall dataset”.