A rapid the elderly sudden disability detection algorithm based on vision sensor is proposed for solving the problem of failure to detect sudden disability for sitting and lying the elderly. The static elderly is detected by using bounding box selective update policy. In the paper, video grid technology is proposed for dividing video into a plurality of sub-areas. Different frequency and duration of entering the sub-region are counted for establishing a Gaussian model based on duration. Experiments show that the proposed method has smooth running, real time, and without monitoring dead angle, besides it has the ability of fast and stable in detecting sudden disability for sitting and lying live-aloned old people.