下载中心
优秀审稿专家
优秀论文
相关链接
首页 > , Vol. , Issue () : -
摘要

在高光谱解混的过程中考虑影像的空间信息,能够有效提高解混精度。而超像素分割能够划分空间同质区域,为此本文提出一种考虑光谱信息和超像素分割的解混网络(SSUNet)。首先需对原始影像进行超像素分割处理,获得具有空间特征的超像素分割数据,然后采用SSUNet对原始高光谱数据和超像素分割数据进行训练和解混。在线性和非线性混合模型生成的模拟数据集和两个真实数据集上的实验表明,与SUnSAL、SUnSAL-TV、SCLRSU 、MTAEU、EGU-Net-pw和1DCNN的解混结果相比,所提网络具有更高的解混精度和较好的鲁棒性。
The basic unit of remote sensing image is the pixel. If a single pixel contains multiple types of covering ground objects, it is called a mixed pixel. Hyperspectral unmixing is to decompose the mixed pixels into several basic component units (endmembers) and obtain the proportion (abundance) of each endmember, which can improve the accuracy of remote sensing image classification and subpixel level target detection, and its research promotes the development of hyperspectral remote sensing technology. Studies show that considering the spatial information in the process of hyperspectral unmixing can effectively improve the unmixing accuracy. However, most of the nonlinear unmixing networks based on deep learning only use the spectral information of images. In order to make full use of the spectral and spatial information of images, a hyperspectral unmixing network considering spectral information and superpixel segmentation(SSUNet) is proposed based on the supervised unmixing idea and one-dimensional convolutional neural network. Firstly, the original hyperspectral data should be processed by superpixel segmentation to obtain the superpixel segmentation data with spatial characteristics. Then, SSUNet is used to train and unmix the original hyperspectral data and superpixel segmentation data. The loss function adds l_1 regularization constraint term on the basis of the root mean square error to promote the sparsity of the unmixing abundance and make the unmixing result closer to the real value. The activation function of the network output layer is softmax, which makes the output value of each output node within the range of [0,1], and constrains the sum of output values of each node to be 1, thus satisfying the two constraints of unmixing: the abundance nonnegative constraint (ANC) and abundance sum-to-one constraint (ASC). Experiments on simulated datasets generated by the linear mixed model and nonlinear mixed model and two real datasets show that the proposed network has higher unmixing accuracy and better robustness than the unmixing results of SUnSAL, SUnSAL-TV, SCLRSU, MTAEU, Endmember-Guided Unmixing Network (EGU-Net-pw) and 1DCNN. Three Gaussian noises with different SNR levels (20dB, 30dB, 40dB) are added to the simulated dataset. The proposed network can achieve the best unmixing results at all SNR levels, and with the increase of SNR, the network also achieves higher unmixing accuracy. In addition, the influence of the change of w value on the unmixing result of the simulated datasets under different SNR is verified. The experimental results show that when the value range of w is [3,13], the RMSE value does not change much, and the best value of w is 5. Experiments on real datasets show that SSUNet can still achieve the best unmixing results in complex real scenes. The SSUNet network uses the dual-branch structure to mine the features of the original image data and the superpixel segmentation data with spatial features, and uses the fusion layer to fuse the features, to improve the unmixing accuracy of the model. Experiments on simulated datasets and real hyperspectral datasets show that the proposed network has higher accuracy.