针对传统的支持向量机学习算法(SVM)在对视频关键帧提取过程中普遍存在学习参数不易确定,准确度偏低的问题,提出一种自主扰动变异差分SVM算法用来对视频进行关键帧提取.首先,对差分进化算法的生物学机理进行研究,提出一种改进的自主扰动变异方式.其次,结合改进形式的自主扰动差分进化算法对SVM参数进行选取优化,设计了基于该改进差分SVM算法的视频关键帧提取算法.通过在标准测试函数及视频测试数据库中的实验表明,改进的自主扰动变异差分视频关键帧提取算法能够更加有效地优化支持向量机参数,从而有助于改善视频检索的查全(准)率两个算法性能评价标准.
In video key frames extraction, the accuracy of traditional support vector machine (SVM) algorithm is low and it is difficult to determine the learning parameters. To avoid these drawbacks, a self perturbation mutation differential evolution based SVM algorithm was designed. After studying the biological mechanism of difference evolu- tionary algorithm, a new advanced variation of self mutation differential method was proposed. The improved method was used to optimize the parameters of SVM algorithm, based on which a new video key frame extraction algorithm was designed. The experiments results of standard functions test and video database test have shown that the im- proved self perturbation mutation differential video key frame extraction algorithm can optimize the SVM parameters more effectively, which improves the two important indicators of recall and precision in video retrieval.