论文部分内容阅读
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。
In order to make full use of the feature information contained in the reconstructed residuals in the sparse representation classification algorithm, the band information of the reconstructed residuals is fed back to the test samples to adaptively enhance the sparse feature extraction of the samples. In order to further improve the stability and classification accuracy of the algorithm, a close coupled set of pixels (CCSP) is proposed to smooth the feature distribution to solve the over-fits Finally, a close coupled set of pixels-based adaptive boosting class-wise sparse representation classifier (CCSP-ABCWSRC) is proposed based on tightly coupled pixels. The experimental results on three hyperspectral data sets of Indian Pines, University of Pavia and Salinas show that the proposed algorithm can classify hyperspectral images stably and efficiently and its classification accuracy is superior to that of similar algorithms.