基于改进Faster R-CNN的SAR图像飞机检测算法

李广帅; 苏娟; 李义红

doi:10.13700/j.bh.1001-5965.2020.0004

基于改进Faster R-CNN的SAR图像飞机检测算法

doi: 10.13700/j.bh.1001-5965.2020.0004

火箭军工程大学核工程学院, 西安 710025

详细信息

作者简介:
李广帅男, 硕士研究生。主要研究方向:基于深度学习的目标检测

苏娟女, 博士, 副教授, 硕士生导师。主要研究方向:遥感图像处理、模式识别等

通讯作者:
苏娟, E-mail: suj04@mails.tsinghua.edu.cn

中图分类号: TP751
计量
- 文章访问数: 1070
- HTML全文浏览量: 206
- PDF下载量: 392
- 被引次数: 0
出版历程
- 收稿日期: 2020-01-04
- 录用日期: 2020-05-22
- 网络出版日期: 2021-01-20

An aircraft detection algorithm in SAR image based on improved Faster R-CNN

College of Nuclear Engineering, Rocket Force University of Engineering, Xi'an 710025, China

摘要

摘要:
在合成孔径雷达（SAR）图像分析领域，飞机作为一种重要目标，对其的检测越来越受到重视。针对传统SAR图像飞机检测算法需要人工设计特征且鲁棒性较差的问题，提出了一种基于改进Faster R-CNN的SAR图像飞机检测算法。制作了一个SAR图像飞机数据集（SAD），以Faster R-CNN为检测框架，利用改进k-means算法设计更合理的先验锚点框，以适应飞机目标的形状特点；借鉴inception模块思想，设计多路不同尺寸卷积核以扩展网络宽度，增强对浅层特征的表达；分析残差网络Layer5层的特征输出具有更大的感受野，对其上采样后进行特征融合以利用更多的上下文信息；同时引入Mask R-CNN算法中提出的RoI Align单元，消除特征图与原始图像的映射偏差。实验结果表明：相比原始的Faster R-CNN算法，所提改进的Faster R-CNN算法在SAR图像飞机数据集上平均检测精度提高了7.4%，同时保持了较快的检测速度。
- 飞机检测 /
- Faster R-CNN /
- 浅层特征增强 /
- 上下文信息 /
- RoI Align
Abstract:
In the field of Synthetic Aperture Radar (SAR) image analysis, as an important target, aircraft detection has attracted more and more attention. In order to solve the problem that traditional aircraft detection algorithms need to design hand-crafted features and have poor robustness, this paper proposes an aircraft detection algorithm based on improved Faster R-CNN. In this paper, a SAR Aircraft Dataset (SAD) is made. With Faster R-CNN as the detection framework, the improved k-means algorithm is used to design a more reasonable prior anchor frame to adapt to the characteristic of aircraft size. Based on the idea of inception module, multiple convolution kernels of different sizes are designed to expand the network width and enhance the expression of shallow features. By analyzing the residual network, the feature-map of Layer5 has a larger receptive field, and feature fusion is carried out after upsampling to make use of more context information. Meanwhile, the RoI Align unit proposed in Mask R-CNN algorithm is introduced to eliminate the mapping deviation between the feature-map and the original image. The experimental results show that, compared with the original Faster R-CNN algorithm, the proposed algorithm improves the average detection accuracy by 7.4% on the SAD, while maintaining a fast detection speed.
- aircraft detection /
- Faster R-CNN /
- shallow feature enhancement /
- context information /
- RoI Align

HTML全文

图 1 Faster R-CNN检测流程

Figure 1. Faster R-CNN detection flowchart

下载: 全尺寸图片幻灯片

图 2 锚点框示意图

Figure 2. Schematic diagram of anchor box

下载: 全尺寸图片幻灯片

图 3 VGG16、ResNet101结构示意图

Figure 3. Schematic diagram of VGG16 and ResNet101 structure

下载: 全尺寸图片幻灯片

图 4 bottleneck结构示意图

Figure 4. Schematic diagram of bottleneck structure

下载: 全尺寸图片幻灯片

图 5 改进的Faster R-CNN结构

Figure 5. Structure of improved Faster R-CNN

下载: 全尺寸图片幻灯片

图 6 预设锚点框范围示例

Figure 6. Example range of preset anchor box

下载: 全尺寸图片幻灯片

图 7 RoI池化、RoI Align示意图

Figure 7. Schematic diagram of RoI Pooling and RoI Align

下载: 全尺寸图片幻灯片

图 8 VOC数据集中各目标图像数量

Figure 8. Number of various target images of VOC dataset

下载: 全尺寸图片幻灯片

图 9 飞机目标信息统计直方图

Figure 9. Statistical histogram of aircraft target information

下载: 全尺寸图片幻灯片

图 10 SAD数据集中部分SAR飞机图像

Figure 10. Some SAR aircraft images in SAD dataset

下载: 全尺寸图片幻灯片

图 11 不同算法P-R曲线比较

Figure 11. P-R curves comparison among different algorithms

下载: 全尺寸图片幻灯片

图 12 检测结果对比

Figure 12. Comparison of detection results

下载: 全尺寸图片幻灯片

表 1 飞机目标统计参数

Table 1. Statistical parameters of aircraft target

统计参数	最大值	最小值	平均值
宽度/像素	459	27	87
宽度占比	0.510	0.030	0.097
高度/像素	378	42	81
高度占比	0.630	0.070	0.135
面积/像素	173 502	1 134	7 047
面积占比	0.321	0.002	0.013

下载: 导出CSV

表 2 实验结果对比

Table 2. Comparison of experimental results

算法	特征提取网络	AP/%	R/%	P/%	检测速度/ fps
Faster R-CNN	VGG16	81.1	78.1	77.0	16.6
Faster R-CNN	ResNet101	82.3	80.2	79.4	13.2
Faster R-CNN+ k-means	ResNet101	83.5	81.7	80.8	13.5
本文算法	ResNet101	88.5	89.5	85.9	12.7
注:fps为帧/s。

下载: 导出CSV

参考文献(21)

[1]	黎薇萍, 李渝, 段崇棣, 等.一种新的鲁棒CFAR检测器设计方法[J].空间电子技术, 2018(3):61-64. doi: 10.3969/j.issn.1674-7135.2018.03.013 LI W P, LI Y, DUAN C L, et al.A new robust CFAR detector design method[J].Space Electronic Technology, 2018(3):61-64(in Chinese). doi: 10.3969/j.issn.1674-7135.2018.03.013
[2]	CUI Y, ZHOU G, YANG J, et al.On the iterative censoring for target detection in SAR images[J].IEEE Geoscience and Remote Sensing Letters, 2011, 8(4):641-645. doi: 10.1109/LGRS.2010.2098434
[3]	BRUSCH S, LEHNER S, FRITZ T, et al.Ship surveillance with TerraSAR-X[J].IEEE Transactions on Geoscience and Remote Sensing, 2011, 49(3):1092-1103. doi: 10.1109/TGRS.2010.2071879
[4]	王彦华, 陈维, 王军福, 等.基于级联CFAR的SAR图像目标快速检测方法[J].现代雷达, 2019, 41(2):21-25. http://www.cnki.com.cn/Article/CJFDTotal-XDLD201902005.htm WANG Y H, CHEN W, WANG J F, et al.Fast target detection method of SAR image based on cascaded CFAR[J].Modern Radar, 2019, 41(2):21-25(in Chinese). http://www.cnki.com.cn/Article/CJFDTotal-XDLD201902005.htm
[5]	CHEN S, LI X.A new CFAR algorithm based on variable window for ship target detection in SAR images[J].Signal, Image and Video Processing, 2019, 13:779-786. doi: 10.1007/s11760-018-1408-4
[6]	GAO F, XUE X, WANG J, et al.Visual attention model with a novel learning strategy and its application to target detection from SAR images[C]//International Conference on Brain Inspired Cognitive Systems.Berlin: Springer, 2016: 149-160.
[7]	TU S, SU Y.Fast and accurate target detection based on multiscale saliency and active contour model for high-resolution SAR images[J].IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(10):5729-5744. doi: 10.1109/TGRS.2016.2571309
[8]	BENGIO Y, COURVILLE A, VINCENT P.Representation learning:A review and new perspectives[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8):1798-1828. doi: 10.1109/TPAMI.2013.50
[9]	LIU W, ANGUELOV D, ERHAN D, et al.SSD: Single shot multibox detector[C]//European Conference on Computer Vision.Berlin: Springer, 2016: 21-37.
[10]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection[C]//IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2016: 779-788.
[11]	REDMON J, FARHADI A.YOLO9000: Better, faster, stronger[C]//IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2017: 6517-6525.
[12]	REDMON J, FARHADI A.YOLOv3: An incremental improvement[EB/OL].(2018-04-08)[2020-01-01].http://arxiv.org/abs/1804.02767.
[13]	GIRSHICK R, DONAHUE J, DARRELL T, et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition.Piscataway: IEEE Press, 2014: 580-587.
[14]	GIRSHICK R.Fast R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Piscataway: IEEE Press, 2015: 1440-1448.
[15]	REN S, HE K M, GIRSHICK R, et al.Faster R-CNN: Towards real-time object detection with region proposal networks[C]//Proceedings of Advances in Neural Information Processing Systems, 2015: 91-99.
[16]	常鹏飞, 段云龙.Faster R-CNN模型在遥感图像飞机目标检测中的应用[J].无线电工程, 2019, 49(10):925-929. doi: 10.3969/j.issn.1003-3106.2019.10.016 CHANG P F, DUAN Y L.Application of Faster R-CNN model in aircraft target detection in remote sensing image[J].Radio Engineering, 2019, 49(10):925-929(in Chinese). doi: 10.3969/j.issn.1003-3106.2019.10.016
[17]	余东行, 郭海涛, 张保明, 等.级联卷积神经网络的遥感影像飞机目标检测[J].测绘学报, 2019, 48(8):1046-1058. http://www.cnki.com.cn/Article/CJFDTotal-CHXB201908012.htm YU D X, GUO H T, ZHANG B M, et al.Aircraft target detection in remote sensing image using cascaded convolutional neural network[J].Journal of Surveying and Mapping, 2019, 48(8):1046-1058(in Chinese). http://www.cnki.com.cn/Article/CJFDTotal-CHXB201908012.htm
[18]	王思雨, 高鑫, 孙皓, 等.基于卷积神经网络的高分辨率SAR图像飞机目标检测方法[J].雷达学报, 2017, 6(2):195-203. doi: 10.12000/JR17009 WANG S Y, GAO X, SUN H, et al.Method of aircraft target detection in high resolution SAR image based on convolutional neural network[J].Journal of Radar, 2017, 6(2):195-203(in Chinese). doi: 10.12000/JR17009
[19]	LECUN Y, BOTTOU L, BENGIO Y, et al.Gradient-based learning applied to document recognition[J].Proceedings of the IEEE, 1998, 86(11):2278-2324. doi: 10.1109/5.726791
[20]	IOFFE S, SZEGEDY C.Batch normalization: Accelerating deep network training by reducing internal covariate shift[C]//International Conference on Machine Learning, 2015: 448-456.
[21]	HE K M, GKIOXARI G, DOLLAR P, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Piscataway: IEEE Press, 2017: 2980-2988.