Abstract: China has entered the aging society.The elderly have become an increasing concern.The detection algorithm of indoor human activities based on 3-frame difference is discussed.The method of maximum classes square error（method of Ostu） and the 3-frame-differencing are applied to the image segmentation of an indoor elder moving slowly and the basic estimation for his/her activity posture.The experiments show that the human posture image can be segmented successfully.The real time requirement is satisfied，and it lays a good basis for the following identifying the elderly posture and determining his/her security.

Key words: three-frame-differencing, motion detection, the elderly aging-in-place

摘要： 中国已经进入老龄化社会，老人的安全日益受到关注。通过最大类间方差法和三帧时间差分法进行自然背景和光照条件下室内运动人体目标的分割及运动状态的检测。实验表明算法基本能检测到运动人体轮廓，满足实时性要求，并为以后的独居老人姿态识别和安全判断打好基础。

关键词: 三帧差分法, 运动检测, 独居老人