3D密集全卷积的高光谱地物分类算法研究

Research on Hyperspectral Ground Object Classification Algorithm Based on 3D Dense Full Convolution

针对高光谱遥感图像训练样本较少、光谱维度较高、空间特征与频谱特征存在差异性而导致高光谱地物分类的特征提取不合理、分类精度不稳定和训练时间长等问题,提出了基于3D密集全卷积(3D-DSFCN)的高光谱图像(HSI)分类算法。算法通过密集模块中的3D卷积核分别提取光谱特征和空间特征,采用特征映射模块替换传统网络中的池化层和全连接层,最后通过softmax分类器进行分类。实验结果表明,基于3D-DSFCN的HSI分类方法提高了地物分类的准确率、增强了低频标签的分类稳定性。

In view of the problems that hyperspectral remote sensing images have few training samples, high spectral dimension, and differences between spatial and spectral features, which lead to unreasonable feature extraction, unstable classification accuracy and long training time, a hyperspectral image(HSI)classification algorithm based on 3D dense full convolution(3D-DSFCN)is proposed. The algorithm extracts spectral features and spatial features through 3D convolution kernel in dense module, replaces pool layer and full connection layer in traditional network with feature mapping module,and finally classifies by softmax classifier. Experimental results show that the HSI classification method based on 3D-DSFCN improves the accuracy of ground object classification and enhances the classification stability of low frequency tags.