Object Detection Research of SAR Image Using Improved Faster Region-Based Convolutional Neural Network

doi:10.11947/j.JGGS.2020.0302

Abstract

Abstract:

Target detection technology of synthetic aperture radar (SAR) imageis widely used in the field of military reconnaissance and surveillance. The traditional SAR image target detection methods need to be provided a lot of empirical knowledge because the characteristics of SAR images in different configurations (attitude, pitch angle, imaging parameters, etc.) will change greatly,resulting in high generalization error. Currently, deep learning method has achieved great success in the field of image processing. Research shows that deep learning can achieve a more intrinsic description of the data, while the model has a stronger ability of modeling and generalization. In order to solve the problem of insufficient data in SAR data sets, an experimental system for acquiring SAR image data in real scenes was built. Then the transfer learning method and the improved convolution neural network algorithm (PCA+Faster R-CNN) are applied to improve the target detection precision. Finally, experimental results demonstrate the significant effectiveness of the proposed method.

Key words: target detection; SAR image; deep learning; transfer learning; PCA+Faster R-CNN

Long SUN, Tao WU, Guangcai SUN, Dazheng FENG, Lieshu TONG, Mengdao XING. Object Detection Research of SAR Image Using Improved Faster Region-Based Convolutional Neural Network[J]. Journal of Geodesy and Geoinformation Science, 2020, 3(3): 18-28.

Figures/Tables 20

Fig.1

Fig.2

Fig.3

Fig.4

Tab.1

Tab.2

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Tab.3

Tab.4

Fig.10

Fig.11

Tab.5

Fig.12

Fig.13

Tab.6

Fig.14

References 21

[1]	MOREIRA A, PRATS-IRAOLA P, YOUNIS M , et al. A tutorial on synthetic aperture radar[J]. IEEE Geoscience and Remote Sensing Magazine, 2013,1(1):6-43.
[2]	EL-DARYMLI K, GILL E W, MCGUIRE P , et al. Automatic target recognition in synthetic aperture radar imagery: a state-of-the-art review[J]. IEEE Access, 2017,4:6014-6058.
[3]	BRIECHLE K, HANEBECK U D. Template matching using fast normalized cross correlation [C]∥Proceeding of the SPIE 4387, Optical Pattern RecognitionXII. Orlando: SPIE, 2001: 95-102.
[4]	RALAIVOLA L, D’ALCHÉ-BUC F. Incremental support vector machine learning: a local approach[C]∥Proceedings of International Conference on Artificial Neural Networks. Austria: Springer, 2001.
[5]	WANG Lu, ZHANG Fan, LI Wei , et al. A method of SAR target recognition based on Gabor filter and local texture feature extraction[J]. Journal of Radars, 2015,4(6):658-665.
[6]	QI Huijiao, WANG Yinghua, DING Jun , et al. SAR target recognition based on multi-information dictionary learning and sparse representation[J]. Systems Engineering and Electronics, 2015,37(6):1280-1287.
[7]	GALLJ, YAO A, RAZAVI N , et al. Hough forests for object detection, tracking, and action recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2011,33(11):2188-2202.
[8]	NGUYEND T, LI Wanqing, OGUNBONA P O . Human detection from images and videos: asurvey[J]. Pattern Recognition, 2016,51:148-175.
[9]	SCHMIDHUBER J . Deep learning in neural networks: an overview[J]. Neural networks, 2015,61:85-117.
[10]	LITJENSG, KOOI T, EHTESHAMI BEJNORDI B , et al. A survey on deep learning in medical image analysis[J]. Medical Image Analysis, 2017,42:60-88.
[11]	HAN Zhaoying, CHONG Jinsong. A review of ship detection algorithms in polarimetric SAR images [C]∥Proceedings of the 7th International Conference on Signal Processing. Beijing, China: IEEE, 2004.
[12]	LIU Shuaishi, CHENG Xi, GUO Wenyan , et al. Progress report on new research in deep learning[J]. CAAI Transactions on Intelligent Systems, 2016,11(5):567-577.
[13]	REN Shaoqing, HE Kaiming, GIRSHICK R , et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017,39(6):1137-1149.
[14]	DING Shifei, JIA Weikuan, SU Chunyang, et al. PCA-based Elman neural network algorithm [C]∥Proceedings of the 3rd International Symposium on Intelligence Computation and Applications. Wuhan, China: Springer, 2008: 315-321.
[15]	SPULAK D, OTREBSKI R, KUBINGER W . Evaluation of PCA, LDA and fisherfacesin appearance-based object detection in thermal infra-red images with incomplete data[J]. Procedia Engineering, 2015,100:1167-1173.
[16]	CHEN Fubing, CHEN Xiuhong, WANG Wensheng , et al. The generalization of PCA for human face recognition[J]. Computer Engineering and Applications, 2005,41(34):34-38.
[17]	DELAC K, GRGIC M, GRGIC S . Independent comparative study of PCA, ICA, and LDA on the FERET data set[J]. International Journal of Imaging Systems and Technology, 2005,15(5):252-260.
[18]	LIU Chen, QU Changwen, ZHOU Qiang , et al. SAR image target classification based on convolutional neural network transfer learning[J]. Modern Radar, 2018,40(3):38-42.
[19]	SHIN H C, ROTH H R, GAO Mingchen , et al. Deep convolutional neural networks for computer-aided detection: CNN architectures, dataset characteristics and transfer learning[J]. IEEE Transactions on Medical Imaging, 2016,35(5):1285-1298.
[20]	ZHANG Yimei, DING Xiangqian, HE Ying , et al. The research on analysis and improvement of logistic model tree[J]. Microcomputer & Its Applications, 2014(23):25-28.
[21]	LIU Jianwei, LIU Yuan, LUO Xionglin . Research and development on deep learning[J]. Application Research of Computers, 2014,31(7):1921-1930, 1942.

Parameter name	Parameter content
Flight platform	Yun-7 aircraft
Flight altitude	3400m
Flight speed	110-140m/s
Working band	X
Resolving power	X band 0.3m
Polarization mode	X band full polarization(HH、HV、VH、VV)

Target category	Target name	Quantity	Detailed parameters
1	Truck	9	Length 8.80m Width 2.50m Height 4.01m
2	Bus	9	Length 10.49m Width 2.48m Height 3.58m
3	Crane	6	Length 12.79m Width 2.5m Height 3.46m Angle of elevation -2°~80°
4	Car	6	Length 5.1m Width 1.83m Height 1.97m

Test image	Fig.9(a)	Fig.9(b)	Fig.9(c)	Fig.9(d)	Total
Target number(NP)	20	20	20	20	80
Number of targets detected(TP)	19	19	17	18	73
Number of false alarm targets(FP)	0	0	0	0	0
Accuracy	1.0	1.0	1.0	1.0	1.0
Recall	0.95	0.95	0.85	0.90	0.91
F1-Score	0.9744	0.9744	0.9189	0.9474	0.9542

Test image	Fig.10(a)	Fig.10(b)	Fig.10(c)	Fig.10(d)	Total
Target number(NP)	14	13	15	14	56
Number of targets detected(TP)	10	12	12	10	44
Number of false alarm targets(FP)	0	0	0	0	0
Accuracy	1.0	1.0	1.0	1.0	1.0
Recall	0.71	0.92	0.80	0.71	0.79
F1-Score	0.8333	0.9600	0.8889	0.8333	0.8800

Test image	Fig.12(a)	Fig.12(b)	Fig.12(c)	Fig.12(d)	Total
Target number(NP)	20	20	20	20	80
Number of targets detected(TP)	20	19	19	20	78
Number of false alarm targets(FP)	1	0	0	1	2
Accuracy	0.95	1.0	1.0	0.95	0.97
Recall	1.00	0.95	0.95	1.00	0.98
F1-Score	0.9756	0.9744	0.9744	0.9756	0.9750