Abstract: Shot boundary detection method based on Structure Perserving Projection（SPP） is proposed. In order to extract more effective video features, the SPP is presented. The adject matix of SPP is constructed based on mainflod structure not on space structure, and the singular value problem of SPP is resloved by QR decomposition method. Shot boundary is recognized with Support Vector Machine（SVM）. The experimental results show that the proposed method is powerful and effective for shot boundary detection.

Key words: shot boundary detction, structure preserving projection, QR decomposition, support vector machine

摘要： 提出了一种基于结构保留投影的镜头边界检测方法。为了提取更具判别力的视频特征，提出了结构保留投影方法，利用流形结构信息而非空间结构信息构建连接矩阵，利用QR分解方法解决了结构保留投影的奇异值问题；采用支持向量机检测镜头边界。实验结果表明该方法具有很好的检测效果。

关键词: 镜头边界检测, 结构保留投影, QR分解, 支持向量机