中文摘要：

为了从不同类型的视频中有效地提取关键帧，提出基于群体智能与K-均值相结合的关键帧提取方法。该方法首先提取视频的颜色特征向量，利用基于群体智能的聚类方法自组织地对颜色向量进行聚类，得到初始聚类；然后通过K-均值对初始聚类进行优化并加快算法收敛，得到最终聚类；最后，提取每类中距离聚类中心最近的向量对应帧并将其作为关键帧。实验表明，该算法能有效地提取出代表视频内容的关键帧，对镜头的相似性和连续性反映准确。

英文摘要：

In order to extract key frame efficiently from different type of video, this paper proposed an efficient method for key frame extraction based on swarm intelligence and K-means. Firstly, applied swarm intelligence clustering analysis to the histogram differences of video shot self-organized,and obtained an initial clustering result. Secondly,conducted K-means to optimize the in- itial clustering result and to accelerate the convergence of the algorithm, obtained a final clustering result. Finally, extracted the center frame of each clustering as the key frames. Experiment indicates that the proposed algorithm can extract the representative key frame of the video accurately and efficiently, and can reflect the similarities and continuity of the shot accuratly.