基于U-Net网络和GF-6影像的尾矿库空间范围识别

Recognition of the spatial scopes of tailing ponds based on U-Net and GF-6 images

利用遥感手段实现尾矿库空间范围的快速识别对我国尾矿库监测监管具有重要意义。以U-Net网络框架为基础,提出了基于深度学习的尾矿库空间范围遥感智能识别方法,利用国产高分六号影像在云南省红河哈尼族彝族自治州开展了应用验证。结果表明,该方法对尾矿库空间范围识别的精确率(Precision)、召回率(Recall)、F1-score值分别达到0.874,0.843和0.858,显著优于随机森林、支持向量机、最大似然法;尾矿库空间范围识别的耗时与上述3种方法保持相同的数量级水平。该方法在全国尾矿库空间范围变化的遥感快速监测中具有广阔的应用前景。

It is of great significance for the monitoring and supervision of tailing ponds in China to realize the rapid recognition of the spatial scopes of tailing ponds using the remote sensing technique.Based on the U-Net framework,this paper proposes a deep learning-based intelligent recognition method of the spatial ranges of tailing ponds using the remote sensing technique.The method proposed was verified in Honghe Hani and Yi Autonomous Prefecture in Yunnan Province using Chinese GF-6 satellite images.The results show that the precision,recall rate,and F1-score of the method were 0.874,0.843,and 0.858,respectively,which were significantly better than those obtained using the methods of random forest,support vector machine,and maximum likelihood.Furthermore,the time consumption of the new method kept the same order of magnitude as that of the three methods.Therefore,the method proposed in this study has a broad application prospect in the rapid monitoring of the spatial scopes of tailing ponds in China.