基於人物動作分析之影像日誌系統

Abstract

在現今工商業社會中，年輕人日間忙於工作無法陪伴家中年邁長輩，為了提供有效的協助，讓家屬能安心工作，本論文利用電腦視覺之技術，透過監視攝影機影像來分析被觀測者的活動狀態、高度變化、行為動作，並識別各肢體部位建立人體姿態模型，據以發展一個具隱私保護、以及兼顧各個使用者對被觀測者關心層面不同之需求的階層式、方便瀏覽的居家看護影像日誌系統。此系統除了可以減輕看護與醫療人員的工作負擔，同時可以讓家屬經由監視畫面及時掌握居家年長者的生活狀況與安全。另外由於監視畫面經隱私保護處理，被觀測者也可以比較自在安心。本階層式系統架構不僅可應用於居家看護系統，同時也可以應用於其他活動，例如球類運動，提供使用者不同的需求。

In today's society, young people are often too busy to accompany their family elders. In order to provide an effective assistance to family members, we propose a surveillance system based on computer vision technology to analyze people's active level, height changes, and action behavior, and recognize each limb to build a human body model for a display with privacy protection. Accordingly, a hierarchical home care image diary system is developed for convenient browsing, and to meet various concerns and demands. For example, the system can reduce nursing and medical staff's workload, while family members can use the monitoring screen to grasp the living conditions and safety of elders. Furthermore, since video images have been processed for privacy protection, people being looked after will feel more comfortable. In addition to home care system, this hierarchical idea of system architecture can be applied to videos of other activities, such as ball games, to serve various needs of the viewers.