基于深度学习的高分遥感影像尾矿库提取

Extraction of tailings ponds based on deep learning from high-resolution remote sensing images

尾矿库检测和边界提取对于尾矿库的数字化管理和监测非常重要。利用改进U-Net模型进一步对目标区域准确地提取尾矿库,框架的整体精度达到98.12%。该方法能够从大面积高空间分辨率遥感影像中高精度、高速度地提取各种尾矿库,有效减少了尾矿库数字化管理的人力和财力成本,为政府部门快速获取尾矿库边界信息提供一定的参考。

Tailings pond detection and boundary extraction are important for the digital management and monitoring of tailings ponds.The tailings reservoir was further extracted accurately for the target area using the improved U-Net model,and the overall accuracy of the framework reached 98.12%.This method can extract various tailings ponds with high accuracy and speed from large area high spatial resolution remote sensing images,which effectively reduces the human and financial cost of tailings pond digital management and provides some reference for government departments to obtain tailings pond boundary information quickly.