Abstract: Through the analysis of different kinds of shot boundaries,an algorithm for shot boundary detection based on a cascade classifier is proposed.The proposed algorithm first detects smooth intervals based on gray variance from the original frame sequence,then concatenates them to constructs a new frame sequence.After that,video features such as pixel-wise difference,HSV histogram difference and edge histogram difference are fed to a SVM classifier to decide the types of different shot boundaries.Experimental result shows that the proposed algorithm performs better than the accumulation and SVM-TMRA algorithm.