深度学习的半监督遥感图像检索

发布时间：2018-05-29 10:31

本文选题：遥感图像检索 + 深度学习　；参考：《遥感学报》2017年03期

【摘要】：遥感图像数据的海量性、多样性和复杂性等特点对遥感图像检索的速度和精度提出了更高的要求,其中特征提取是影响遥感图像检索效果的关键。本文方法首先对遥感图像进行预处理,然后基于稀疏自动编码的方法在大量未标注的遥感图像上进行特征学习得到特征字典,基于卷积神经网络的思想,使用训练出来的特征字典对遥感图像进行卷积和池化得到每幅图像的特征图;接下来使用特征图训练Softmax分类器;最后对待检索图像分类,在同一类别中计算特征间的距离,进而实现遥感图像的检索。实验结果表明,该方法能够有效提高遥感图像检索的速度和准确度。
[Abstract]:The characteristics of mass, diversity and complexity of remote sensing image data put forward higher requirements for the speed and accuracy of remote sensing image retrieval, among which feature extraction is the key to the effect of remote sensing image retrieval. In this paper, the remote sensing image is preprocessed first, then the feature dictionary is obtained by feature learning based on sparse automatic coding on a large number of unlabeled remote sensing images, and based on the idea of convolution neural network. We use the trained feature dictionary to convolution and pool the remote sensing image to get the feature map of each image. Then we use the feature map to train the Softmax classifier. Finally, we deal with the classification of the retrieval image and calculate the distance between the features in the same category. Then the retrieval of remote sensing image is realized. Experimental results show that this method can effectively improve the speed and accuracy of remote sensing image retrieval.
【作者单位】：中国科学院遥感与数字地球研究所;中国科学院大学;
【基金】：国家自然科学基金(编号:61501460)~~
【分类号】：TP751

【相似文献】