Lira-YOLO: a lightweight model for ship detection in radar images

doi:10.23919/JSEE.2020.000063

Journal of Systems Engineering and Electronics ›› 2020, Vol. 31 ›› Issue (5): 950-956.doi: 10.23919/JSEE.2020.000063

• Defence Electronics Technology • Previous Articles Next Articles

Lira-YOLO: a lightweight model for ship detection in radar images

Long ZHOU^1,^2,*(), SuyuanX WEI²(), Zhongma CUI¹(), Jiaqi FANG¹(), Xiaoting YANG¹(), Wei DING^1,³()

¹ Beijing Institute of Remote Sensing Equipment, Beijing 100854, China
² Xi'an High-tech Research Institute, Xi'an 710025, China
³ School of Information and Communication Engineering, University of Electronic Science

Received:2019-05-13 Online:2020-10-30 Published:2020-10-30
Contact: Long ZHOU E-mail:1792563879@qq.com;weisuyuan2002@136.com;czmsy@sina.com;fangjiaqi123@hotmail.com;younger-bit@sina.com;screamdw@126.com
About author:ZHOU Long was born in 1995. He received his B.S. degrees from Xi'an University of Posts & Telecommunications in 2017. He is currently an M.S. degree candidate in Xi'an High-tech Research Institute. His current major research interests are lightweight convolutional neural network and object detection of radar images.E-mail: 1792563879@qq.com|WEI Suyuan was born in 1971. She received her Ph.D. degree from Xidian University in 2013. Her major research interests are pattern recognition and computer network & communication.E-mail: weisuyuan2002@136.com|CUI Zhongma was born in 1978. He received his B.S. degree from Anhui University in 2000, and M.S. degree from the Second Academy of China Aerospace Science and Industry Corporation Limited in 2003, respectively. His major research interests are radar system design and intelligent identification technology.E-mail: czmsy@sina.com|FANG Jiaqi was born in 1984. He received his B.S., M.S. and Ph.D. degrees from Xidian University in 2007, 2011 and 2016, respectively. His major research interests are radar signal processing and intelligent identification technology.E-mail: fangjiaqi123@hotmail.com|YANG Xiaoting was born in 1989. She received her B.S. and M.S. degrees from Hebei University of Science and Technology and Beijing Institute of Technology in 2012 and 2016, respectively. She is now an engineer and her major research interest is object detection based on radar signals.E-mail: younger-bit@sina.com|DING Wei was born in 1992. He received his B.S. degree in North China University of Water Resources and Electric Power in 2017. He is currently an M.S. degree candidate in University of Electronic Science and Technology of China. His current major research interest is lane detection.E-mail: screamdw@126.com
Supported by:
the Joint Fund of Equipment Pre-Research and Aerospace Science and Industry(6141B07090102);This work was supported by the Joint Fund of Equipment Pre-Research and Aerospace Science and Industry (6141B07090102)

Abstract

Abstract:

For the detection of marine ship objects in radar images, large-scale networks based on deep learning are difficult to be deployed on existing radar-equipped devices. This paper proposes a lightweight convolutional neural network, LiraNet, which combines the idea of dense connections, residual connections and group convolution, including stem blocks and extractor modules. The designed stem block uses a series of small convolutions to extract the input image features, and the extractor network adopts the designed two-way dense connection module, which further reduces the network operation complexity. Mounting LiraNet on the object detection framework Darknet, this paper proposes Lira-you only look once (Lira-YOLO), a lightweight model for ship detection in radar images, which can easily be deployed on the mobile devices. Lira-YOLO's prediction module uses a two-layer YOLO prediction layer and adds a residual module for better feature delivery. At the same time, in order to fully verify the performance of the model, mini-RD, a lightweight distance Doppler domain radar images dataset, is constructed. Experiments show that the network complexity of Lira-YOLO is low, being only 2.980 Bflops, and the parameter quantity is smaller, which is only 4.3 MB. The mean average precision (mAP) indicators on the mini-RD and SAR ship detection dataset (SSDD) reach 83.21{%} and 85.46{%}, respectively, which is comparable to the tiny-YOLOv3. Lira-YOLO has achieved a good detection accuracy with less memory and computational cost.

Key words: lightweight, radar images, ship detection, you only look once (YOLO)

Long ZHOU, SuyuanX WEI, Zhongma CUI, Jiaqi FANG, Xiaoting YANG, Wei DING. Lira-YOLO: a lightweight model for ship detection in radar images[J]. Journal of Systems Engineering and Electronics, 2020, 31(5): 950-956.

Figures/Tables 15

Table 1

Fig 1

Fig 2

Fig 3

Table 2

Fig 4

Fig 5

Table 3

Table 4

Table 5

Table 6

Fig 6

Fig 7

Fig 8

Fig 9

References 34

1	NOVAK L, OWIRKA G J, BROWER W S, et al. The automatic target-recognition system in SAIP. Lincoln Laboratory Journal, 1997, 10 (2): 187- 202.
2	VACHON P W, CAMPBELL J W M, BJERKELUND C A, et al. Ship detection by the RADARSAT SAR validation of detection model predictions. Canadian Journal of Remote Sensing, 1997, 23 (1): 48- 59.
3	JIANG Q, AITNOURI E, WANG S, et al. Automatic detection for ship target in SAR imagery using PNN-model. Canadian Journal of Remote Sensing, 2000, 26 (4): 297- 305. doi: 10.1080/07038992.2000.10874780
4	KRONAUGE M, ROHLING H. Fast two-dimensional CFAR procedure. IEEE Trans. on Aerospace and Electronic Systems, 2013, 49(3): 1817-1820.
5	HOU B, CHENG X Z, JIAO L C. Multilayer CFAR detection of ship resolution targets in very high resolution SAR images. IEEE Geoscience and Remote Sensing Letters, 2015, 12 (4): 811- 815. doi: 10.1109/LGRS.2014.2362955
6	NGAN I T H, ZHENG Y, ZHU C, et al. Multiple scale faster-RCNN approach to driver's cell-phone usage and hands on steering wheel detection. Proc. of the Computer Vision & Pattern Recognition Workshops, 2016, 46- 53.
7	LI Y X, LI J W, LIN W Y, et al. Tiny-DSOD: lightweight object detection for resource-restricted usages. http://arXiv.org/abs/1807.11013.
8	ZEILER M D, FERGUS R. Visualizing and understanding convolutional networks. Proc. of the European Conference on Computer Vision, 2014, 818- 833.
9	SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition. http://arXiv.org/abs/1409.1556.
10	HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.
11	FU C, LIU W, RANGA A, et al. DSSD: deconvolutional single shot detector. http://arXiv.org/abs/1701.06659.
12	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection. http://arXiv.org/abs/1708.02002.
13	LI J W, QU C W, PENG S J, et al. Ship detection in SAR images based on convolutional neural network. Systems Engineering and Electronics, 2018, 40 (9): 1953- 1959.
14	GIRSHICK R. Fast R-CNN. Proc. of the IEEE International Conference on Computer Vision, 2015, 1440- 1448.
15	QU C W, LIU C, ZHOU Q, et al. Ship targets detection method based on convolutional neural network for SAR image. Fire Control & Command Control, 2019, 44 (1): 40- 44.
16	HAN S, MAO H Z, DALLY W J. Deep compression: compressing deep neural networks with pruning, trained quantization and huffman coding. http://arXiv.org/abs/1510.00149.
17	YU X Y, LIU T L, WANG X C, et al. On compressing deep models by low rank and sparse decomposition. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, 7370- 7379.
18	LI Y X, LI J W, LIN W Y, et al. Tiny-DSOD: lightweight object detection for resource-restricted usages. http://arXiv.arg/abs/11013.
19	WONG A, SHAFIEE M J, LI F, et al. Tiny SSD: a tiny single-shot detection deep convolutional neural network for real-time embedded object detection. Proc. of the 15th Conference on Computer and Robot Vision, 2018, 95- 101.
20	WANG X, GIRSHICK R, GUPTA A, et al. Non-local neural networks. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, 7794- 7803.
21	SUN K, XIAO B, LIU D, et al. Deep high-resolution representation learning for human pose estimation. http://arXiv.org/abs/1902.09212.
22	CAO Y, XU J R, LIN S, et al. GCNet: non-local networks meet squeeze-excitation networks and beyond. http://arXiv.org/abs/1904.11492.
23	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector. http://arXiv.org/abs/1512.02325.
24	HOWARD A G, ZHU M, CHEN B, et al. MobileNets: efficient convolutional neural networks for mobile vision applications. http://arXiv.org/abs/1704.04861.
25	ZHANG X, ZHOU X, LIN M, et al. ShulfleNet: an extremely efficient convolutional neural network for mobile devices. http://arXiv.org/abs/1707.01083.
26	WANG R J, LI X, LING C X, et al. Pelee: a real-time object detection system on mobile devices. Proc. of the 32nd International Conference on Neural Information Processing Systems, 2018, 1963- 1992.
27	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, 779- 788.
28	JOSEPH R, ALI F. YOLO9000: better, faster, stronger. http://arXiv.org/abs/1612.08242.
29	REDMON J, FARHADI A. YOLOv3: an incremental improvement. http://arXiv.org/abs/1804.02767.
30	ZHOU L, WEI S Y, CUI Z M, et al. Multi-objective detection of complex background radar image based on deep learning. Systems Engineering and Electronics, 2019, 41 (6): 1258- 1264.
31	SHEN Z Q, LIU Z, LI J G, et al. DSOD: learning deeply supervised object detectors from scratch. Proc. of the IEEE International Conference on Computer Vision, 2017, 1937- 1945.
32	LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection. http://arXiv.org/abs/1612.03144.
33	HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition. Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, 770- 778.
34	HE K M, ZHANG X Y, REN S Q, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2015, 37(9): 1904-1916.

Dataset	Class	Training set	Test set hline
Mini-RD	3	1 280	267
SSDD	1	878	282

Stage	Layer	Output size hline
Input size		416×416×3
Stem block		104×04×6 hline
Extractor	Dense layer×3+ Transition layer	52×52×28
	Dense layer×4+ Transition layer	26×26×28
	Dense layer×8+ Transition layer	13×3×28
	Dense layer×6+ Conv 1×1/1	13×3×28

System	Ubuntu18.04 hline
CPU	Intel Core i5-8400 CPU @2.80 GHz×6
GPU	GeForce GTX 1070
Hardware acceleration	CUDA9.0; CUDNN7.5

Parameter	DSOD	#A1	#A2	#A3	The proposed model	PeleeNet
Weight/MB	13.6	13.6	13.6	14.4	13.5	10.1
Computation/ Bflops	13.716	14.181	13.282	17.973	13.028	6.105
mAP/%	88.16	87.86	87.70	88.05	87.93	86.93

Parameter	PeleeNet	#B1	#B2	Lira-YOLO.v1
Weight/MB	10.1	7.6	7.6	7.6
Computation/ Bflops	6.105	4.287	4.720	4.232
mAP/%	86.93	86.39	86.30	88.04

Lira-YOLO: a lightweight model for ship detection in radar images

RichHTML

PDF (PC)

Knowledge

Abstract

Cite this article

Share this article

Figures/Tables 15

References 34

Related Articles 1

Recommended Articles

Metrics

Comments

Parameter		YOLOv3	Tiny- YOLOv3	Lira- YOLO.v1	Lira- YOLO.v2
Weight/MB		246.3	34.7	7.6	4.3
Computation/Bflops		65.290	5.448	4.232	2.980
mAP/%	SSDD	89.72	88.69	88.04	86.30
mAP/%	Mini-RD	83.96	84.14	82.06	82.37