本文针对智能视频监控中的物体遗留事件检测进行了研究，给出了一套完整的检测方案。多高斯模型用于运动目标检测，其自适应性很好地解决了背景帧不断变化所带来的影响；MeanShift算法用于运动目标的跟踪，使得监控对象不再限于固定区域；目标的七阶不变矩可以很好地描述目标特征，利用这一特征通过支持向量机对目标进行识别。实验结果证明了本文方法的有效性。

The paper studies the detection of the event of the objects left behind in intelligent video surveillance systems,and gives a complete set of test schema. The multipleGaussian model is used for the detection of the moving objects,and its adaptability well solve the impact of changing background frame.The  MeanShift algorithm is used for tracking the moving targets,which expands the scope,not only limited scope.The feature of Sevenband Moment Invariants well describe the target, through which support Vector Machine can identify the objects. The experimental results prove the effectiveness of the method.