摘要
为提高视频检索的准确率,提出了一种基于多特征融合的视频运动对象识别算法。该算法首先使用基于背景帧构造及关键帧截取的方法提取视频中的运动对象的区域;然后分别提取运动对象的局部特征SURF描述子和全局特征如颜色直方图、边缘直方图等,并使两者融合为统一的特征向量;最后使用支持向量机对特征进行学习和识别,用以识别视频对象。实验证明该算法有效地提高了视频中运动对象识别的准确率。
In order to raise the recognition accuracy of video retrieval, a video motion object recognition algorithm using multi-features fusion is proposed. With background frame and key frame, firstly the area of motion object in video, then the local features SURF descriptors and global features, such as color histogram and edge histogram, are extracted from motion object, and then these two histograms fused into unified feature vector, and finally support vector machine is employed to train and recognize the features representing video object. Experimental result shows that this algorithm could effectively raise the recognition accuracy of motion object in video
出处
《信息安全与通信保密》
2012年第3期57-58,共2页
Information Security and Communications Privacy
关键词
对象识别
多特征融合
机器学习
支持向量机
object recognition
multi-features fusion
machine learning
support vector machine