Multi-model and multi-scale scene recognition of shipbuilding enterprises based on convolutional neural network with spatial constraints
-
摘要: 船企场景识别对修复沿岸生态环境、保护水域环境以及促进船舶产业的协调发展具有现实意义,但传统方法基于中、低层次的特征难以实现卫星遥感图像中船企的自动识别。为此,提出了结合空间约束的卷积神经网络多模型多尺度船企场景识别方法。首先分别采用全局尺度的船企场景和局部尺度船坞(台)、厂房和船只样本训练多个卷积神经网络模型,并进行多模型多尺度检测; 进而对局部对象进行像素级定位并计算对象空间距离; 最终结合多尺度检测结果、对象标签组合方式、对象空间距离进行船企场景综合判别与提取。将此方法分别应用于中国江苏省、日本长崎县和爱媛县周边以及韩国木浦市和巨济市周边5个典型造船密集区。结果表明,江苏省整体识别精确度为87%,召回率为85%; 日本研究区整体识别精确度为91%,召回率为87%; 韩国研究区整体识别精确度为85%,召回率为92%。实验结果表明,此方法可以较好地实现遥感船企复杂场景的识别。Abstract: The scene recognition of shipbuilding enterprises is of practical significance for the restoration of the coastal ecological environment, the protection of water environment, and the promotion of the coordinated development of shipbuilding enterprises. However, it is difficult to realize the automatic recognition of shipbuilding enterprises from satellite remote sensing images based on traditional medium- and low-level features. Therefore, this paper proposes a multi-model multi-scale scene recognition method of shipbuilding enterprises based on a convolutional neural network with spatial constraints and the steps are as follows. Firstly, train multiple convolutional neural network models using the samples of global-scale shipbuilding enterprise scenes and local-scale docks (slipways), workshops, and ships individually, and conduct multi-model multi-scale detection. Then, locate local-scale objects at a pixel level and calculate the spatial distance of the objects. Finally, conduct comprehensive judgment and extraction of the shipbuilding enterprise scenes according to the multi-scale detection results, the combination method of object tags, and the spatial distance of objects. The method was applied to five typical shipbuilding intensive areas in Jiangsu Province, China, the surrounding areas of Nagasaki and Ehime prefectures, Japan, and Mokpo and Geoje cities, South Korea. As a result, the overall recognition accuracy and recall rate were 87% and 85%, respectively in Jiangsu Province, were 91% and 87%, respectively in the study area in Japanese, and were 85% and 92%, respectively in the study area in South Korean. The experimental results show that this method can realize the effective recognition of the complex scenes of shipbuilding enterprises based on remote sensing images.
-
-
[1] 杨青. 长三角船舶产业集群研究[D]. 上海:上海海事大学, 2007.
[2] Yang Q. Research on the shipbuilding industry cluster in the Yangtze River delta[D]. Shanghai:Shanghai Maritime University, 2007.
[3] 刘辉, 史雅娟, 曾春水. 中国船舶产业空间布局与发展策略[J]. 经济地理, 2017,37(8):99-107.
[4] Liu H, Shi Y J, Zeng C S. Spatial layout and development strategy of the ship-related industry in China[J]. Economic Geography, 2017,37(8):99-107.
[5] Zhou Z H, Zhang M L, Huang S J, et al. Multi-instance and multi-label learning[J]. Artificial Intelligence, 2012,176(1):2291-2320.
[6] Li Z L, Xu K, Xie J F, et al. Deep multiple instance convolutional neural networks for learning robust scene representations[J]. IEEE Transactions on Geoscience and Remote Sensing, 2020,58(5):3685-3702.
[7] Liu W H, Li Y D, Wu Q. An attribute-based high-level image representation for scene classification[J]. IEEE Access, 2019,7:4629-4640.
[8] Khan N, Chaudhuri U, Banerjee B, et al. Graph convolutional network for multi-label VHR remote sensing scene recognition[J]. Neurocomputing, 2019,357:36-46.
[9] Chen Z M, Wei X S, Wang P, et al. Multi-Label Image Recognition with Graph Convolutional Networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR).IEEE, 2019.
[10] Clement M, Kurtz C, Wendling L. Learning spatial relations and shapes for structural object description and scene recognition[J]. Pattern Recognition, 2018,84:197-210.
[11] Lazebnik S, Schmid C, Ponce J. Beyond bags of features:Spatial pyramid matching for recognizing natural scene categories[C]//Computer Vision and Pattern Recognition,2006 IEEE Computer Society Conference, 2006.
[12] Jiang Y, Yuan J, Gang Y. Randomized spatial partition for scene recognition[M]. Berlin:Springer Berlin Heidelberg, 2012:730-743.
[13] Weng C Q, Wang H X, Yuan J S, et al. Discovering Class-Specific Spatial Layouts for Scene Recognition[J]. IEEE Signal Processing Letters, 2017,24(8):1143-1147.
[14] He K M, Zhang X Y, Ren S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015,37(9):1904-1916.
[15] Xie L, Lee F F, Liu L, et al. Scene recognition:A comprehensive survey[J]. Pattern Recognition, 2020,102:107205.
[16] 船舶工业年鉴编委会. 中国船舶工业年鉴2018[Z].[s.l]:船舶工业年鉴编委会, 2018.
[17] Editorial Committee of Yearbook of Shipbuilding Industry. Yearbook of China shipbuilding industry 2018[Z].[s.l]:Editorial Committee of Yearbook of Shipbuilding Industry, 2018.
[18] He K M, Zhang X Y, Ren S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas:2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2016:770-778.
[19] 颜荔. 基于卷积神经网络的遥感图像飞机目标识别研究[D]. 合肥:中国科学技术大学, 2018.
[20] Yan L. Research on airplane target recognition in remote sensing images based on convolutional neural network[D]. Hefei:University of Science and Technology of China, 2018.
[21] Zhou Z H. A brief introduction to weakly supervised learning[J]. National Science Review, 2018,5(1):44-53.
[22] Selvaraju R R, Cogswell M, Das A, et al. Grad-CAM:Visual explanations from deep networks via gradient-based localization[J]. International Journal of Computer Vision, 2020,128(2):336-359.
[23] Xia G S, Hu J W, Hu F, et al. AID:A benchmark data set for performance evaluation of aerial scene classification[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017,55(7):3965-3981.
[24] Cheng G, Han J W, Lu X Q. Remote sensing image scene classification:Benchmark and state of the art[J]. Proceedings of the IEEE, 2017,105(10):1865-1883.
[25] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017,60(6):84-90.
[26] Karen S, Andrew Z. Very deep convolutional networks for large-scale image recognition[J]. arXiv, 2014.
-
计量
- 文章访问数: 1045
- PDF下载数: 38
- 施引文献: 0
下载: