摘要
基于多示例学习框架的图像分类算法以其特有的多义性对象表示能力在图像分类中表现出较好的分类效果。但传统的包级空间多示例学习算法在特征选择过程中存在忽略小目标概念区域且包含大量冗余信息的问题,造成部分训练包信息损失,影响分类性能。为此,基于多示例学习与稀疏编码理论提出1种改进的多示例图像分类算法。该算法首先根据同类样本示例聚为一簇的特性,应用聚类算法构造每类图像的视觉词汇,并利用负包中所有示例都为负的特性,对视觉词汇进行约束,消除冗余信息;依据训练样本示例与视觉词汇的相似度,获得每类训练样本的包特征向量。然后,基于稀疏编码理论,对训练包中的包特征向量进行稀疏编码,获得每1类训练样本的字典矩阵。最后,对待分类样本特征进行稀疏线性组合,预测待分类样本的类别标签。通过对COREL数据集图像进行测试,结果表明,与其他多示例学习算法相比,文中提出的方法能较好地解决图像分类问题,具有较高的分类精度。
The classification algorithm based on multiple instance learning（MIL） has a good performance MIL has disambiguate ability. However, the bag-level space multiple instance learning algorithms always small target region and contains a large amount of redundant information during feature selection, which the information loss for partial bags and can affect the performance of classification. In this paper, we p due to the ignore the may cause roposed improved multiple instance learning classification algorithm based on the framework of muhiple instance learning and the sparse coding. Firstly, according to the characteristics of similar samples can cluster into one class, k- means algorithm is used to construct the visual vocabulary for each class of images. To eliminate redundant informa- tion, the negative characteristic of negative samples in negative bags is used to constrain the visual vocabulary. The bag feature vectors for each class of training samples are achieved by computing the similarity between the training sample and the visual vocabulary. Then, sparse coding is used to achieve the dictionary matrix for each class of the training samples. Finally, the labels for test images are predicted by linear combination of the dictionary and coeffi- cient, which is achieved in training data, to represent the bag-level features for test images. Experimental results on COREL image data show that the proposed algorithm can better solve the problems in multiple instance learning ira- age cl.assification and achieve higher classification based image classification algorithms. accuracy compared with the other multiple instance learning
作者
杨红红
曲仕茹
金红霞
Yang Honghong Qu Shim Jin Hongxia(College of Automation, Northwestern Polytechnical University, Xi＇an 710072, China)
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2017年第4期690-697,共8页
Journal of Northwestern Polytechnical University
基金
航天科技创新基金(CASC201104)
航空科学基金(2012ZC53043)资助
关键词
包特征向量
稀疏表示
多示例学习
图像分类
bag feature
sparse representation
multiple instance learning
image classification
作者简介
杨红红（1988-），女，西北工业大学博士研究生，主要从事图像处理与模型识别研究。
