首页 >  , Vol. , Issue () : -

摘要

全文摘要次数: 406 全文下载次数: 644
引用本文:

DOI:

10.11834/jrs.20200124

收稿日期:

2020-04-24

修改日期:

2020-09-04

PDF Free   EndNote   BibTeX
结合E-YOLO和水体指数约束的大面幅影像水利设施检测
许泽宇1, 沈占峰1, 李杨1, 李均力2, 李硕1, 王浩宇1, 焦淑慧3, 李苓苓4
1.中国科学院空天信息创新研究院;2.中国科学院新疆生态与地理研究所;3.武汉大学;4.应急管理部国家减灾中心
摘要:

水利设施对水资源与水量调度、自然湿地生态保护与修复、资源和生态功能的利用及经济效益发展有重要作用,传统方法统计水利设施位置、数量等依赖于汇编资料,存在耗时长、资料更新不够及时以及具体地理位置不详等缺点,遥感为大规模监测水利设施提供了新的可能。本文以YOLO v3网络为基础,结合水利设施的特点,提出了一种基于大面幅影像快速检测水利设施的算法,主要分为2个方面:1)改进的YOLO算法(E-YOLO)。E-YOLO提出PPA特征融合方法和等比预测框与四特征图交叉预测方法,对小样本等问题进行优化;改进损失函数,突出置信度损失;同时使用迁移学习的方法,读取特征提取部分的预训练模型参数。2)基于E-YOLO算法和水体指数约束的大面幅水利设施检测算法。通过水体指数约束滑动步幅来解决影像面幅大、目标尺度小的问题,同时降低漏检率和误检率,再结合轮廓合并方法,优化检测结果。本研究中采用高分二号影像数据实现大面幅影像水利设施检测,实验结果表明:E-YOLO算法可以明显提高水利设施检测效果,相比平均F2精度相比YOLO v3提高了1.25%。且有更好的稳定性;水体指数约束的大面幅检测方法可以在保证效率的情况下提高检测精度,其F2精度相比大步幅和小步幅方法分别提高了3.72%和2.70%,为遥感水利设施检测提供了良好方案。

Detection of water conservancy facilities in large-format image combining E-YOLO algorithm and NDWI constraint
Abstract:

Water conservancy facilities play an important role in water resources and water volume scheduling, natural wetland ecological protection and restoration, utilization of resources and ecological functions, and economic benefits. The traditional method of statistics on the location and number of water conservancy facilities depends on compiled data. Due to the length of time and the shortcomings of insufficient data update and unclear specific geographic location, remote sensing provides new possibilities for large-scale monitoring of water conservancy facilities. Aiming at the problem of detection of water conservancy facilities with remote sensing images, based on the YOLO v3 network and the characteristics of water conservancy facilities, this paper proposes a large-scale image detection algorithm for water conservancy facilities, which is mainly divided into two aspects: 1) Improving the YOLO algorithm and get E-YOLO algorithm. E-YOLO proposes a PPA feature fusion method and a four-feature map cross prediction method with proportional prediction box to optimize the problems of small samples. Besides, we improve the loss function by highlighting the loss of confidence. What’s more, we use the transfer learning method to read the feature extraction part parameters of the pre-trained model. 2) With the improved E-YOLO algorithm as the core, a large-area water conservancy facility detection algorithm combined with the water body index constraint is obtained. Aiming at the problem of large image size with the small target scale, we use the water body index to constrain the sliding step to reduce the missed detection rate and false detection rate at the same time. Then we combine it with the contour merging method to optimize the detection results. We use the GF-2 data and the experimental results show that: the E-YOLO algorithm can significantly improve the detection effect of water conservancy facilities. Compared with YOLO v3, the average F2 score of E-YOLO is increased by 1.25% and the E-YOLO algorithm has better stability. The large-area detection method constrained by the water index can improve the detection accuracy while ensuring efficiency. Compared with the large-step and small-step methods, its F2 accuracy is increased by 3.72% and 2.70%. Our method provides a good solution for the inspection of water conservancy facilities.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮