基于感知向量的光学遥感图像舰船检测

doi:10.12305/j.issn.1001-506X.2022.12.06

摘要/Abstract

摘要：

针对光学遥感图像中近岸舰船目标检测干扰大、虚警率高的问题, 在基于包围框边缘感知向量(box boundary-aware vectors, BBAVectors)检测网络的基础上提出了改进方法。首先在特征融合网络后加入一个有监督的注意力模块来增强目标区域信息, 削弱无关背景信息干扰; 然后利用边界感知向量间的几何关系设计了一个自监督损失函数, 用以加强向量间的耦合关系, 防止向量独立性导致包围框出现不规则形状。实验结果显示, 在HRSC2016数据集L₂级检测任务中, 改进模型检测结果的平均精度相较于原网络提高了6.91%, 有效抑制了背景噪声的干扰, 降低了近岸舰船目标检测的虚警率, 证明了改进方法的有效性。

关键词: 光学遥感图像, 舰船目标检测, 包围框边缘感知向量, 监督, 注意力模块

Abstract:

To solve the problem of severe interference and high false positive rate in ship detection from remote sensing images, an enhanced method based on the box boundary-aware vectors (BBAVectors) detection network is proposed.Firstly, a supervised attention module is added to the feature fusion network to enhance the relevant information within the target region and reduce the interference of irrelevant background information. Then a self-supervised loss function is proposed based on the geometric relations among the boundary vectors to guarantee the coupling relation between vectors and prevent the irregular shape of the bounding boxes caused by vectors' independence. Experimental results in the L₂ level detection task on the HRSC2016 dataset show that the mean average precision of the detection results for the proposed model gets improved by 6.91% compared with the original network. The proposed model can effectively suppress the interference of background noise and reduce the false alarm rates in near-shore ship detection, which demonstrates its effectiveness.

Key words: optical remote sensing images, ship targets detection, box boundary-aware vectors (BBAVectors), supervised, attention module

中图分类号:

P237

潘超凡, 李润生, 许岩, 胡庆, 牛朝阳, 刘伟. 基于感知向量的光学遥感图像舰船检测[J]. 系统工程与电子技术, 2022, 44(12): 3631-3640.

Chaofan PAN, Runsheng LI, Yan XU, Qing HU, Chaoyang NIU, Wei LIU. Ship detection of optical remote sensing images based on aware vectors[J]. Systems Engineering and Electronics, 2022, 44(12): 3631-3640.

图/表 17

图1

图2

图3

图4

图5

图6

图7

表1

图8

图9

表2

图10

表3

图11

表4

表5

表6

参考文献 29

1	KRIZHEVSKYA , SUTSKEVER I , HINTON G E . Imagenet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60 (6): 84- 90. doi: 10.1145/3065386
2	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
3	GIRSHICKR. Fast R-CNN[C]//Proc. of the IEEE International Conference on Computer Vision, 2015: 1440-1448.
4	RENS Q , HE K M , GIRSHICK R , et al. FasterR-CNN: towards real-time object detection with region proposal networks[J]. IEEE Trans.on Pattern Analysis & Machine Intelligence, 2017, 39 (6): 1137- 1149.
5	LAW H , DENG J . CornerNet: detecting objectsas paired keypoints[J]. International Journal of Computer Vision, 2020, 128 (3): 642- 656. doi: 10.1007/s11263-019-01204-1
6	DUAN K W, BAI S, XIE L X, et al. CenterNet: keypoint triplets for object detection[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2019: 6569-6578.
7	ZHOU X Y, ZHUO J C, KRHENBVHL P. Bottom-up object detection by grouping extremeand center points[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 850-859.
8	陈冬, 句彦伟. 基于改进型YOLOv3的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2021, 43 (4): 937- 943.
	CHEN D , JU Y W . Ship detection in SAR imagebased on improved YOLOv3[J]. Systems Engineering and Electronics, 2021, 43 (4): 937- 943.
9	LIU Z K, HU J G, WENG L B, et al. Rotated region based CNN for ship detection[C]//Proc. of the IEEE International Conference on Image Processing, 2017: 900-904.
10	JIANG Y Y, ZHU X Y, WANG X B, et al. R2CNN: rotational region CNN for orientation robust scene text detection[EB/OL]. [2021-07-10]. https://arxiv.org/abs/1706.09579v2.
11	MA J Q , SHAO W Y , YE H , et al. Arbitrary-oriented scene text detection via rotation proposals[J]. IEEE Trans.on Multimedia, 2018, 20 (11): 3111- 3122. doi: 10.1109/TMM.2018.2818020
12	DING J, XUE N, LONG Y, et al. Learning RoI transformer for detecting oriented objects in aerial images[EB/OL]. [2021-07-10]. https://arxiv.org/abs/1812.00155.
13	LIAO M H, ZHU Z, SHI B G, et al. Rotation-sensitive regression for oriented scene text detection[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 5909-5918.
14	WEI H Y , ZHANG Y , CHANG Z H , et al. Orientedobjects as pairs of middle lines[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2020, 169, 268- 279. doi: 10.1016/j.isprsjprs.2020.09.022
15	ZHOU L , WEI H R , LI H , et al. Objects detection for remotesensing images based on polar coordinates[J]. IEEE Access, 2020, 8, 223373- 223384. doi: 10.1109/ACCESS.2020.3041025
16	ZHU Z C , DIAO W H , CHEN K Q , et al. Diamondnet: ship detection in remote sensing images by extractingand clustering keypoints in a diamond[J]. ISPRS Annals of Photogrammetry, Remote Sensing & Spatial Information Sciences, 2020, 5 (2): 625- 632.
17	ZHANG F, WANG X Y, ZHOU S L, et al. Arbitrary-oriented ship detection through center-headpoint extraction[EB/OL]. [2021-07-10]. https://arxiv.org/abs/2101.11189v2.
18	YI J R, WU P X, LIU B, et al. Oriented object detectionin aerial images with boxboundary-aware vectors[C]//Proc. of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2021: 2150-2159.
19	YANG X, YANG J R, YAN J C, et al. SCRDet: towards more robust detection for small, clutteredand rotated objects[C]//Proc. of the IEEE/CVF International Conference on Computer Vision, 2019: 8232-8241.
20	HU J, SHEN L, SUN G, et al. Squeeze-and-excitation networks[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
21	WOO S, PARK J, LEE J Y, et al. CBAM: convolutional block attention module[C]//Proc. of the European Conference on Computer Vision, 2018.
22	LIU Z K, YUAN L, WENG L B, et al. A high resolution optical satellite image dataset for ship recognition and some new baselines[C]//Proc. of the International Conference on Pattern Recognition Applications and Methods, 2017: 324-331.
23	仲伟峰, 郭峰, 向世明, 等. 旋转矩形区域的遥感图像舰船目标检测模型[J]. 计算机辅助设计与图形学学报, 2019, 31 (11): 1935- 1945.
	ZHONG W F , GUO F , XIANG S M , et al. Ship detectionin remote sensing based with rotated rectangular region[J]. Journal of Computer-Aided Design & ComputerGraphics, 2019, 31 (11): 1935- 1945.
24	LIU Z K, HU J G, WENG L B, et al. Rotatedregion based CNN for ship detection[C]//Proc. of the IEEE International Conference on Image Processing, 2017: 900-904.
25	LIN Y T, FENG P M, GUAN J, et al. IENet: interacting embranchment one stage anchor free detector for orientation aerial object detection[EB/OL]. [2021-07-10]. https://arxiv.org/abs/1912.00969.
26	ZHANG Z H , GUO W W , ZHU S N , et al. Toward arbitrary-oriented ship detection withrotated region proposal and discri-minationnetworks[J]. IEEE Geoscience and Remote Sensing Letters, 2018, 15 (11): 1745- 1749. doi: 10.1109/LGRS.2018.2856921
27	LIAO M H, ZHU Z, SHI B G, et al. Rotation-sensitive regression for oriented scene text detection[C]//Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2018: 5909-5918.
28	DING J, XUE N, LONG Y, et al. Learning ROI transformer for oriented object detection in aerial images[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 2849-2858.
29	PAN X J, REN Y Q, SHENG K K, et al. Dynamic refinement network for oriented and denselypacked object detection[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 11207-11216.

方法	mAP(07)	mAP(12)	推理速度/fps
CP^[24]	55.70	-	-
BL2^[24]	69.60	-	-
R²CNN^[10]	73.07	79.73	5.00
IENet^[25]	75.01	-	-
RC1^[24]	75.70	-	-
RC2^[24]	75.70	-	-
RRPN^[11]	79.08	85.64	1.50
R²PN^[26]	79.60	-	-
RRD^[27]	84.30	-	-
RoI Trans.^[28]	86.20	-	5.90
DRN^[29]	-	92.70	-
基线网络	88.83	90.75	13.40
本文方法	90.06	94.50	12.68

目标	方法
目标	基线网络	本文方法
航母	80.90	97.65
战舰	92.54	94.1
商船	64.66	68.84
潜艇	67.19	77.62
mAP(07)	74.2	83.05
mAP(12)	76.33	84.55

目标	方法
目标	基线网络	基线网络+注意力模块
航母	80.90	89.32
战舰	92.54	92.63
商船	64.66	63.19
潜艇	67.19	76.24
mAP(12)	76.33	80.35

目标	方法
目标	基线网络	基线网络+L_vec
航母	80.90	93.85
战舰	92.54	92.86
商船	64.66	64.84
潜艇	67.19	76.81
mAP(12)	76.33	82.09

weight₁	weight₂	航母	战舰	商船	潜艇	mAP(12)
1	0.3	89.99	93.43	69.15	75.28	81.96
1	0.5	93.83	93.74	71.53	79.28	84.59
1	0.8	91.19	94.81	70.49	79.02	83.88
1	1	97.65	94.10	68.84	77.62	84.55
1	2	95.64	94.82	71.17	71.00	83.15
1	3	87.17	93.29	67.89	75.54	80.97
1	4	93.27	94.56	66.78	69.27	80.97
2	1	88.07	92.99	66.52	70.02	79.40
3	1	93.62	93.16	68.07	65.09	79.99
4	1	93.27	94.25	70.25	65.61	80.84