摘要
场景分类中使用了许多种类的图像特征,但通常情况下,一种特征很难对许多不同场景都得到不错的分类结果,故而对特征融合方面做了很多研究工作。但特征融合的方法存在一个问题,即一般维数会很高,这个高维的特征向量可能包含冗余信息和噪声,从而降低最终的分类准确率。因此提出了使用PCA对融合的特征进行降维以去除冗余信息和噪声,经实验验证,该方法提高了分类的准确率。
Many features are used in scene classification,but a single feature usually can't achieve good performance for different scenes ,so there are huge research of feature fusion. There exists one problem,if gets a merged feature of high dimension,it may includes redundancy information and noise. It leads to reduction of accuracy of scene classification. So PCA is used to reduce the redundancy information and noise in the merged feature, and the re- sult show that it improves the accuracy.
出处
《电视技术》
北大核心
2013年第13期10-12,25,共4页
Video Engineering
基金
国家自然科学基金项目(61071155)
关键词
场景分类
融合
冗余信息
噪声
scene classification
fusion
redundancy information
noise