An information extraction model of roads from high-resolution remote sensing images based on improved Deeplabv3+
-
摘要: 针对传统的道路提取方法在高分辨率遥感影像中存在提取效果差和提取速度慢的问题,提出了改进Deeplabv3+的高分辨率遥感影像道路提取模型。采用MobileNetv2主干特征提取网络与Dice Loss函数相结合,较好地平衡了高分辨率遥感影像道路提取精度与速度的矛盾,实现较高提取精度的同时减少了模型参数,满足了时效性的要求。基于开源道路提取数据集的实验结果表明: ①该文提出的道路提取模型在高分辨率遥感影像上具有可行性,提取道路的整体精度达到98.71%,具有较高的提取精度; ②在提取道路的速度方面该方法平均帧数达到120.05,模型参数量仅为5.81 M,总体上比原模型更加轻量化,表明该方法满足了时效性的要求。该方法在大幅减少参数量、满足时效性的同时保证了提取的精确度,为提高基于高分辨率影像的道路提取精度和速度提供了一种新的改进思路和方法。
-
关键词:
- 遥感影像 /
- 道路提取 /
- 深度学习 /
- 语义分割 /
- DeepLabv3+模型
Abstract: Aiming at the problems of poor extraction effect and slow extraction speed of traditional road extraction methods in the information extraction of roads from high-resolution remote sensing images, this study proposed a new information extraction model based on improved Deeplabv3+. In the new model, the combination of the MobileNetv2 backbone feature extraction network with the Dice Loss function effectively balanced the contradiction between the precision and speed of road information extraction from high-resolution remote sensing images. As a result, high extraction precision was achieved while meeting timeliness requirements by reducing model parameters. The experimental results based on the open-source road information extraction dataset show that: ① The road information extraction model proposed in this study was feasible for high-resolution remote sensing images, with overall accuracy of up to 98.71%; ② In terms of the information extraction speed, the new model had an average frame number of 120.05 and parameter amount of only 5.81 M. Therefore, the new model was more lightweight lighter than original models, meeting the timeliness requirements. Therefore, the model proposed in this study meets the timeliness requirements by greatly reducing the parameter amount while ensuring high extraction accuracy. This study provides a new philosophy and method for improving the accuracy and speed of road information extraction from high-resolution images. -
-
[1] Herold M, Roberts D. Spectral characteristics of asphalt road aging and deterioration:Implications for remote-sensing applications[J]. Applied Optics, 2005, 44(20):4327-4334.
[2] Kass M, Witkin A, Terzopoulos D. Snakes:Active contour models[J]. International Journal of Computer Vision, 1988, 1(4):321-331.
[3] 罗庆洲, 尹球, 匡定波. 光谱与形状特征相结合的道路提取方法研究[J]. 遥感技术与应用, 2007, 22(3):339-344.
[4] Luo Q Z, Yin Q, Kuang D B. Research on extracting road based on its spectral feature and shape feature[J]. Remote Sensing Technolo-gy and Application, 2007, 22(3):339-344.
[5] Ghaziani M, Mohamadi Y, Koku A B, et al. Extraction of unstructured roads from satellite images using binary image segmentation[C]// 2013 21st Signal Processing and Communications Applications Conference (SIU).IEEE, 2013:1-4.
[6] Sirma?ek B, ünsalan C. Road network extraction using edge detection and spatial voting[C]// 2010 20th International Conference on Pattern Recognition.IEEE, 2010:3113-3116.
[7] 贾建鑫, 孙海彬, 蒋长辉, 等. 多源遥感数据的道路提取技术研究现状及展望[J]. 光学精密工程, 2021, 29(2):430-442.
[8] Jia J X, Sun H B, Jiang C H, et al. Road extraction technology based on multi-source remote sensing data: Review and prospects[J]. Optics and Precision Engineering, 2021, 29(2):430-442.
[9] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6):84-90.
[10] 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.IEEE, 2016:770-778.
[11] Mnih V, Hinton G E. Learning to detect roads in high-resolution aerial images[C]// European Conference on Computer Vision.Springer,Berlin,Heidelberg, 2010:210-223.
[12] 叶雪娜. 基于卷积神经网络的遥感图像道路提取研究[D]. 西安: 陕西师范大学, 2017.
[13] Ye X N. Research on remote sensing image road extraction based on convolutional neural network[D]. Xi’an: Shaanxi Normal University, 2017.
[14] Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]// Proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition.Washington, DC: IEEE Computer Society, 2015:3431-3440.
[15] Chen L C, Papandreou G, Kokkinos I, et al. Semantic image segmentation with deep convolutional nets and fully connected CRFs[J]. Computer Science, 2014(4):357-361.
[16] Chen L C, Zhu Y, Papandreou G, et al. Encoder-decoder with atrous separable convolution for semantic image segmentation[C]// Proceedings of the European Conference on Computer Vision, 2018:801-818.
[17] 魏云超, 赵耀. 基于DCNN的图像语义分割综述[J]. 北京交通大学学报, 2016, 40(4):82-91.
[18] Wei Y C, Zhao Y. A review on image semantic segmentation based on DCNN[J]. Journal of Beijing Jiaotong University, 2016, 40(4):82-91.
[19] Sandler M, Howard A, Zhu M, et al. Mobilenetv2:Inverted residuals and linear bottlenecks[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018:4510-4520.
[20] Mnih V. Machine learning for aerial image labeling[D]. Toronto: University of Toronto, 2013.
-
计量
- 文章访问数: 1478
- PDF下载数: 229
- 施引文献: 0
下载: