Using deep learning to detect small targets in infrared oversampling images

doi:10.21629/JSEE.2018.05.07

Journal of Systems Engineering and Electronics ›› 2018, Vol. 29 ›› Issue (5): 947-952.doi: 10.21629/JSEE.2018.05.07

收稿日期:2017-11-17 出版日期:2018-10-26 发布日期:2018-11-14

Using deep learning to detect small targets in infrared oversampling images

Liangkui LIN*(), Shaoyou WANG(), Zhongxing TANG()

Shanghai Institute of Satellite Engineering, Shanghai 201109, China

Received:2017-11-17 Online:2018-10-26 Published:2018-11-14
Contact: Liangkui LIN E-mail:gxckk1980@sina.com;wong926@sina.com;yujishenlan@163.com
About author:LIN Liangkui was born in 1980. He received his Ph.D. degree in communication and information system from National University of Defense Technology in 2011. He is currently an engineer in the Shanghai Institute of Satellite Engineering. His research interests include machine learning, multisensor multi-target tracking, and image processing. E-mail: gxckk1980@sina.com|WANG Shaoyou was born in 1982. He received his M.S. degree in optical engineering from Harbin Institute of Technology in 2009. He is currently a senior engineer in the Shanghai Institute of Satellite Engineering. His research interests include design and analysis of space optical sensor, and infrared information processing. E-mail: wong926@sina.com|TANG Zhongxing was born in 1985. He received his Ph.D. degree in spacecraft design from Beihang University in 2014. He is currently an engineer in the Shanghai Institute of Satellite Engineering. His research interests include spacecraft dynamics and control, imaging task planning and scheduling. E-mail: yujishenlan@163.com
Supported by:
the National Key Research and Development Program of China(2016YFB0500901);the Natural Science Foundation of Shanghai(18ZR1437200);the Satellite Mapping Technology and Application National Key Laboratory of Geographical Information Bureau(KLSMTA-201709);This work was supported by the National Key Research and Development Program of China (2016YFB0500901), the Natural Science Foundation of Shanghai (18ZR1437200), and the Satellite Mapping Technology and Application National Key Laboratory of Geographical Information Bureau (KLSMTA-201709)

摘要/Abstract

Abstract:

According to the oversampling imaging characteristics, an infrared small target detection method based on deep learning is proposed. A 7-layer deep convolutional neural network (CNN) is designed to automatically extract small target features and suppress clutters in an end-to-end manner. The input of CNN is an original oversampling image while the output is a cluttersuppressed feature map. The CNN contains only convolution and non-linear operations, and the resolution of the output feature map is the same as that of the input image. The L1-norm loss function is used, and a mass of training data is generated to train the network effectively. Results show that compared with several baseline methods, the proposed method improves the signal clutter ratio gain and background suppression factor by 3 – 4 orders of magnitude, and has more powerful target detection performance.

Key words: infrared small target detection, oversampling, deep learning, convolutional neural network (CNN)

. [J]. Journal of Systems Engineering and Electronics, 2018, 29(5): 947-952.

Liangkui LIN, Shaoyou WANG, Zhongxing TANG. Using deep learning to detect small targets in infrared oversampling images[J]. Journal of Systems Engineering and Electronics, 2018, 29(5): 947-952.

图/表 7

参考文献 26

1	LATRY C. In-flight commissioning of SPOT5 THR quincunx sampling mode. Proc. of the International Symposium on Remote Sensing, 2003: 189-199.
2	LARSEN M F, JR J J T, SORENSON G O, et al. Impact of the SPIRIT III sensor design on algorithms for background removal, object detection, and point-source extraction. Signal and Data Processing of Small Targets, 1996, 2759, 194- 204.
3	LOMHEIM T S, MILNE E L, KWOK J D, et al. Performance/sizing relationships for a short-wave/mid-wave infrared scanning point-source detection space sensor. Proc. of the IEEE Aerospace Conference, 1999: 113-138.
4	LAWRIE D G, LOMHEIM T S. Advanced electro-optical space-based systems for missile surveillance, TR-2001(8556)- 1. El Segundo CA: National Technical Information Service, 2011.
5	WANG S T, ZHANG W, JIN L H, et al. Point target detection based on temporal-spatial over-sampling system. Journal of Infrared & Millimeter Waves, 2013, 32 (1): 68- 72.
6	LIN L, WANG S, WANG T. Simulation and analysis of point target detection performance for infrared scanning oversampling system. Acta Optica Sinica, 2016, 36 (5): 0528001. doi: 10.3788/AOS201636.0528001
7	WANG T B, LI M, LIN Z P, et al. Comparative performance analysis of over-sampling scanning. Journal of Infrared & Millimeter Waves, 2015, 34 (1): 87- 91.
8	ZHANG W Y. Point target imaging simulation and experiment based on temporal-spatial oversampling system. Infrared and Laser Engineering, 2015, 44 (10): 3124- 3129.
9	DESHPANDE S D, MENG H E, VENKATESWARLU R, et al. Max-mean and max-median filters for detection of small targets. Proc. of the Signal and Data Processing of Small Targets, 1999: 74-83.
10	DRUMMOND O E. Morphology-based algorithm for point target detection in infrared backgrounds. Proc. of the Signal and Data Processing of Small Targets, 1993: 2-11.
11	CAO Y, LIU R M, YANG J. Small target detection using twodimensional least mean square (TDLMS) filter based on neighborhood analysis. International Journal of Infrared & Millimeter Waves, 2008, 29 (2): 188- 200.
12	ZHENG C Y, LI H. Small infrared target detection based on low-rank and sparse matrix decomposition. Applied Mechanics & Materials, 2013, 239-240 (6): 214- 218.
13	LI L, LI H, LI T, et al. Infrared small target detection in compressive domain. Electronics Letters, 2014, 50 (7): 510- 512. doi: 10.1049/el.2014.0180
14	CHEN C L P, LI H, WEI Y, et al. A local contrast method for small infrared target detection. IEEE Trans. on Geoscience & Remote Sensing, 2013, 52 (1): 574- 581.
15	HAN J, LIANG K, ZHOU B, et al. Infrared small target detection utilizing the multiscale relative local contrast measure. IEEE Geoscience & Remote Sensing Letters, 2018, 15 (4): 612- 616.
16	SHI Y, WEI Y, YAO H, et al. High-boost-based multiscale local contrast measure for infrared small target detection. IEEE Geoscience & Remote Sensing Letters, 2017, 15 (1): 33- 37.
17	GIRSHICK R, DONAHUE J, DARRELL T, et al. Regionbased convolutional networks for accurate object detection and segmentation. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2016, 38 (1): 142- 158. doi: 10.1109/TPAMI.2015.2437384
18	GIRSHICK R. Fast R-CNN. Proc. of the IEEE International Conference on Computer Vision, 2015: 1440-1448.
19	REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks. Proc. of the International Conference on Neural Information Processing Systems, 2015: 91-99.
20	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection. Proc. of the Computer Vision and Pattern Recognition, 2016: 779-788.
21	REDMON J, FARHADI A. YOLO9000: better, faster, stronger. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 6517-6525.
22	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector. Proc. of the European Conference on Computer Vision, 2016: 21-37.
23	HE K, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN. Proc. of the IEEE International Conference on Computer Vision, 2017: 2980-2988.
24	IAN G, YOSHUA B, AARON C. Deep learning. Cambridge, MA: MIT Press, 2017.
25	ANDREA V, KAREL L, ANKUSH G. MatConvNet convolutional neural networks. [2017-11-02]. http://www.vlfeat.org/matconvnet/matconvnet-manual.pdf.
26	GEOFF F. Sbirs-two heads are better than one. [2017-11-02]. http://www.armscontrolwonk.com/archive/302137/sbirstwoheads-are-better-than-one/.

Target	GSCR				BSF
Target	Max-Median	Top-Hat	LRSMD	CNN	Max-Median	Top-Hat	LRSMD	CNN
T1	1.706 5	5.089 0	6.605 0	1.215 9×10⁴	3.462 1	6.492 7	8.988 9	1.807 9×10⁴
T2	0.851 0	3.504 6	6.889 9	1.318 4×10⁴	2.490 2	4.958 8	8.930 2	1.791 7×10⁴
T3	1.446 1	4.370 7	6.284 0	7.438 4×10³	2.754 0	4.690 9	7.555 7	9.627 1×10³
T4	4.271 1	10.213 8	7.261 6	1.126 5×10⁴	6.005 5	10.640 0	10.819 0	1.532 3×10⁴
T5	2.264 4	9.900 9	9.097 2	1.081 2×103⁴	5.754 5	13.115 2	14.231 6	1.202 0×10³

Using deep learning to detect small targets in infrared oversampling images

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 7

参考文献 26

相关文章 0

编辑推荐

Metrics

本文评价