本文研究了视频网站内容监播、监控的规律，并且依据新的规律改进了现有的K-MEANS聚类算法。受互视频内容数据非结构化因素的影响，K-MEANS聚类算法进行聚类识别和计算效率上存在明显的缺陷。K-MEANS聚类改进算法的优势在考虑到了视频帧序列关系，并保持了可接受的检索速度。

In this paper, we study the rules of video website content monitoring, and improve the existing K-MEANS clustering algorithm according to the new rule. The K-MEANS clustering algorithm has obvious defects in clustering recognition and computing efficiency due to the influence of unstructured factors of the video content data. The advantage of the improved K-MEANS clustering algorithm takes into account the sequence relations of video frames and maintains the acceptable retrieval speed.