In this paper, we propose a face recognition approach-Structed Sparse Representation-based classification when the measurement of the test sample is less than the number training samples of each subject. When this con...In this paper, we propose a face recognition approach-Structed Sparse Representation-based classification when the measurement of the test sample is less than the number training samples of each subject. When this condition is not satisfied, we exploit Nearest Subspaee approach to classify the test sample. In order to adapt all the eases, we combine the two approaches to an adaptive classification method-Adaptive approach. The adaptive approach yields greater recognition accuracy than the SRC approach and CRC_RLS approach with low ~ample rate on the Extend Yale B dataset. And it is more efficient than other two approaches.展开更多
为了充分利用稀疏表示分类算法中重构残差包含的特征信息,将重构残差的波段信息反馈到测试样本中,自适应增强样本的稀疏特征提取。但反馈调整过程可能会出现特征过拟合的问题,为了进一步提高算法的稳定性和分类精度,提出了紧耦合像元生...为了充分利用稀疏表示分类算法中重构残差包含的特征信息,将重构残差的波段信息反馈到测试样本中,自适应增强样本的稀疏特征提取。但反馈调整过程可能会出现特征过拟合的问题,为了进一步提高算法的稳定性和分类精度,提出了紧耦合像元生成算法(close coupled set of pixels,CCSP)来平滑特征分布以解决过拟合问题,并最终提出了基于紧耦合像元的自适应增强类内稀疏表示高光谱图像分类方法(close coupled set of pixels-based adaptive boosting class-wise sparse representation classifier,CCSP-ABCWSRC)。在Indian Pines,University of Pavia,Salinas三个高光谱数据集上的实验结果表明,提出的算法对高光谱图像进行了稳定有效的分类并且其分类精度优于同类算法。展开更多
基金Supported by National Natural Science Foundation of China(No.61170324 and No.61100105)
文摘In this paper, we propose a face recognition approach-Structed Sparse Representation-based classification when the measurement of the test sample is less than the number training samples of each subject. When this condition is not satisfied, we exploit Nearest Subspaee approach to classify the test sample. In order to adapt all the eases, we combine the two approaches to an adaptive classification method-Adaptive approach. The adaptive approach yields greater recognition accuracy than the SRC approach and CRC_RLS approach with low ~ample rate on the Extend Yale B dataset. And it is more efficient than other two approaches.
文摘为了充分利用稀疏表示分类算法中重构残差包含的特征信息,将重构残差的波段信息反馈到测试样本中,自适应增强样本的稀疏特征提取。但反馈调整过程可能会出现特征过拟合的问题,为了进一步提高算法的稳定性和分类精度,提出了紧耦合像元生成算法(close coupled set of pixels,CCSP)来平滑特征分布以解决过拟合问题,并最终提出了基于紧耦合像元的自适应增强类内稀疏表示高光谱图像分类方法(close coupled set of pixels-based adaptive boosting class-wise sparse representation classifier,CCSP-ABCWSRC)。在Indian Pines,University of Pavia,Salinas三个高光谱数据集上的实验结果表明,提出的算法对高光谱图像进行了稳定有效的分类并且其分类精度优于同类算法。
