论文部分内容阅读
针对如何进一步提高镜头边界检测精度问题,提出了一种基于核的监督非保局(KSNLPP)投影视频镜头检测方法。在非保局投影(NLPP)中引入样本的类别信息,并利用核方法提升NLPP解决非线性问题的能力,提出了KSNLPP算法;在此基础上,将每个镜头视为一类,利用KSNLPP算法得到视频图像的投影矩阵,利用此投影矩阵对新的视频数据进行降维处理,实现有监督的视频特征提取;融合各帧的相邻帧特征构建中间特征,利用局部支持向量机(LSVM)分类器实现镜头边界检测。试验结果表明,提出的镜头边界检测方法能够显著提高镜头边界的检测精度。
Aiming at how to further improve the detection accuracy of the lens boundary, a kernel-based KSNLPP projection video camera detection method is proposed. Introducing the class information of NLPP and using the kernel method to improve the ability of NLPP to solve nonlinear problems, a KSNLPP algorithm is proposed. Based on this, each shot is considered as a class, and KSNLPP algorithm The projection matrix of the video image is obtained, and the new video data is dimensionally reduced by using the projection matrix to achieve supervised video feature extraction. The intermediate features of adjacent frame features are fused and classified by LSVM Realize the lens boundary detection. The experimental results show that the proposed method of lens boundary detection can significantly improve the detection accuracy of the lens boundary.