中文摘要：

针对传统的视频拷贝检测算法计算量大、对视频变化敏感等问题，提出一种新的视频指纹算法来产生包含时空域信息的关键帧。采用奇异值分解的方法提取视频关键帧的特征向量，对视频特征序列进行降维。改进的视频拷贝检测算法对噪声、帧平移、帧旋转、丢帧等有很好的鲁棒性。与以往的拷贝算法相比，提高了视频拷贝检测的查全率和准确率，降低了误码率，与源视频具有较高的匹配率。

英文摘要：

According to the problems of heavy computation load and sensitivity to video variation in conventional video copy de tection algorithms, a novel video fingerprinting algorithm is presented to produce the key frame that contains the time and spa tial information. The singular value decomposition is used to extract feature vectors of key frames, and then to reduce the di mension of video feature sequence. The improved video copy detection algorithm has good robustness against noise, frame shift ing, framerate conversion and frame lost, etc. Compared with previous copy algorithms, the video copy detection recall and precision are improved with high matching rate to the source video, while the error rate is reduced.