U^(2)-Net在建筑物提取中的边缘精度分析

Edge Accuracy Analysis of U^(2)-Net in Building Extraction

随着城市化的进程和遥感科学技术的发展,在高分辨遥感影像中进行建筑物提取一直是摄影测量与遥感领域的一个热点研究主题。针对遥感影像中提取建筑物存在边缘模糊的问题,本文运用U^(2)-Net网络算法提取建筑物,并与lr-aspp、fcn、deeplab_v3三种网络算法分别进行了建筑物提取对比实验;结果表明U^(2)-Net网络,在不损失预测精度的情况下,耗时较短,且准确率可提升至97.478%,可较好地解决建筑物提取中的边缘模糊问题。

With the process of urbanization and the development of remote sensing science and technology,building extraction from high-resolution remote sensing images has been a hot research topic in the field of photogrammetry and remote sensing.Aiming at the problem of blurred edges in building extraction from remote sensing images,this paper uses U^(2)-Net to extract buildings,and compares U^(2)-Net with lr-aspp,fcn and deeplab_v3 to extract buildings.The results show that U^(2)-Net,without loss of prediction accuracy,takes less time,and the comprehensive recognition accuracy can be increased to 97.478%,which can solve the problem of edge ambiguity in building extraction.