The invention relates to a motion-sensing technology based kindergarten intelligent monitoring method, which comprises the steps of installing 3D motion-sensing equipment in a kindergarten, capturing 3D human body skeleton images of children in real time, extracting key joint feature vectors, inputting the feature vectors into a motion classifier so as to carry out motion classification, and finding out motions with the highest matching degree, wherein motion classification is trained in advance through a machine learning method. The method provided by the invention can realize detection, analysis and tracking for human body motion postures and figure motion trails through a three-dimensional visual perception technology, and motions of the children are captured dynamically in real time, so that dangerous behaviors such as fighting, tumbling and boundary crossing can be predicted, and a warning is given out in advance so as to reduce occurrence of kindergarten accidents.