准确提取钢铁厂对去产能监测和环境保护具有重要意义。传统的人工目视解译方法效率低、成本高，无法满足开展大区域钢铁厂监测的需求。以深度学习目标检测网络SSD为基础，构建面向遥感影像钢铁厂提取的深度学习目标检测网络，提出maxout模块，将负样本通路优化为多分支结构，突出难分负样本特征并提升网络对无用特征的抵制效果。利用国产GF-1数据对京津冀地区的钢铁厂进行快速自动提取实验。与人工解译的钢铁厂点位数据的对比表明，该目标检测方法的提取精度达到80%以上。
It is important to accurately detect steel plants for capacity reduction monitoring and environmental protection. The traditional method is time-consuming and laborious, and can not be used to monitor the steel plants in large areas. We propose a stable and accurate method by adding a maxout module to SSD, namely, transforming the negative sample path into a multi-branch structure. The neural network learns abundant features of hard negative samples, and thereby increases resistance to the useless features. Meanwhile, we used the well-trained model to detect steel plants in the Jing-Jin-Ji area based on GF-1 data. The results were compared with the data of the steel plants obtained from visual interpretation. Our method detects steel plants in the Jing-Jin-Ji area with an accuracy of more than 80%.
[1] Hinton G E, Osindero S, Teh Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7):1527-1554.
[2] 张兵. 遥感大数据时代与智能信息提取[J]. 武汉大学学报,2018, 43(7):1861-1871.
[3] LeCun Y, Touresky D, Hinton G, et al. A theoretical framework for back-propagation[C]//Proceedings of the 1988 Connectionist Models Summer School. CMU, Pittsburgh, Pa:Morgan Kaufmann, 1998, 1:21-28.
[4] Girshick R, Donahue J, Darrell T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2014:580-587.
[5] He K, Zhang X, Ren S, et al. Pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1904-1916.
[6] Girshick R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision, 2015:1440-1448.
[7] Ren S, He K, Girshick R, et al. Faster R-CNN:towards real-time object detection with region proposal networks[C]//Advances in Neural Information Processing Systems, 2015:91-99.
[8] Redmon J, Divvala S, Girshick R, et al. You only look once:Unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016:779-788.
[9] Liu W, Anguelov D, Erhan D, et al. SSD:single shot multibox detector[C]//European Conference on Computer Vision. Springer, Cham, 2016:21-37.
[10] Wang S, Quan D, Liang X, et al. A deep learning framework for remote sensing image registration[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 145:148-164.
[11] Xu Z, Xu X, Wang L, et al. Deformable convnet with aspect ratio constrained nms for object detection in remote sensing imagery[J]. Remote Sensing, 2017, 9(12):1312.
[12] Xu K F, Lei B, Zhang Y T. Retrieval of land surface emissivity using spectral and texture features based on neural network[J]. Journal of University of Chinese Academy of Sciences, 2018,35(1):102-108.
[13] 李松,魏中浩,张冰尘,等. 深度卷积神经网络在迁移学习模式下的SAR目标识别[J]. 中国科学院大学学报,2018,35(1):75-83.
[14] 朱思捷,雷斌,吴一戎. 基于稠密卷积神经网络的遥感图像自动色彩校正[J]. 中国科学院大学学报,2019,36(1):93-100.
[15] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. Eprint Arxiv, 2014, 30:1409-1556.
