Video segmentation based on scene can not only shorten the time of search but also improve the accuracy of re -trieval.It has become one of the hot researches in the video field .In order to cut apart video scenes rapidly and accurately , a multimodal segmentation algorithm under genetic algorithm was presented .In light of the idea of different features fusion , it ex-tracted physical characteristics of different modes from video key frame , integrated the similarity of same modal data and correla-tion of different modal data among shots to get the similarity , and the shot similarity matrix ...