摘要
相较于普通图像,遥感图像蕴含更丰富的地物细节和空间信息,但也伴随更显著的噪声干扰,这为高精度图像分割带来了新的挑战。在道路提取任务中,遥感影像的道路区域分割尤其注重边界细节的准确刻画。为此,构建了一种基于边界损失的多尺度识别网络(Multi-Scale Recognition Network,MSRNet),用于从遥感图像中自动提取道路区域。在不同数据集上的对比实验表明,与传统遥感道路提取方法相比,所提出的MSRNet在分割精度方面均有显著提升。
Compared with ordinary images,remote sensing images contain richer ground object details and spatial information,but they also suffer from more significant noise interference,which poses new challenges for high-precision image segmentation.In road extraction tasks,the segmentation of road areas from remote sensing images places particular emphasis on the accurate delineation of boundary details.To this end,a multi-scale recognition network integrated with boundary loss(Multi-Scale Recognition Network,MSRNet)was constructed to automatically extract road areas from remote sensing images.Comparative experiments on different datasets show that the proposed MSRNet achieves a significant improvement in segmentation accuracy compared with traditional remote sensing road extraction methods.
作者
宋梦醒
裴新宇
SONG Mengxing;PEI Xinyu(Shangqiu Institute of Technology,Shangqiu 476000,Henan,China)
出处
《农业装备与车辆工程》
2026年第2期112-117,共6页
Agricultural Equipment & Vehicle Engineering
基金
商丘工学院2023年度校级科研项目“基于机器视觉的颜色识别系统研究”(2023KYXM27)。
关键词
遥感图像
边界损失
多尺度识别
remote sensing images
boundary loss
multi-scale recognition
