基于语义特征的遥感舰船关重部位检测网络

张冬冬; 王春平; 付强

基于语义特征的遥感舰船关重部位检测网络

陆军工程大学石家庄校区电子与光学工程系，河北石家庄 050003

基金项目: 军内某科研项目（LJ20191A040155）

详细信息

作者简介:
张冬冬（1993—），男，硕士研究生，主要从事光电侦察情报处理、目标识别技术研究。E-mail：1042911849@qq.com

通讯作者:
付强（1981—），男，讲师，博士，主要从事智能视觉与目标检测技术研究。E-mail：1418748495@qq.com

中图分类号: TP753
计量
- 文章访问数: 64
- HTML全文浏览量: 18
- PDF下载量: 5
- 被引次数: 0
出版历程
- 网络出版日期: 2022-09-17

Remote sensing ship critical part detection network based on semantic feature

Department of electronic and optical engineering, People Liberation Army Engineering University-Shijiazhuang, Shijiazhuang 050003, China

摘要

摘要: 在近岸场景中，受背景影响，舰船关重部位误检概率高、检测精度低。针对以上问题，提出了一种基于语义特征的舰船关重部位检测网络，并命名为CPDNet（critical part detection network）。通过优化网络结构及引入注意力机制，提升网络的特征表达能力以及对关重部位的感知能力；基于语义信息，设计了语义掩膜模块，以降低背景对检测精度的影响；增加角度参数，使网络适用于具有方向性的目标；构建了舰船关重部位数据集，命名为CP-Ship，以验证所提网络的有效性。在CP-Ship数据集上的实验结果表明：所提网络的平均精度比RetinaNet提高了11.35%，与其他网络模型相比，在检测精度和速度均表现优异。
- 深度学习 /
- 遥感图像 /
- 关重部位检测 /
- 语义特征
Abstract: In near-shore scenes, under the influence of background, the probability of false detection and low detection accuracy of ship critical parts are high. To address the above problems, this paper proposes a detection network for ship critical parts based on semantic features, which named CPDNet (critical part detection network). Firstly, by optimizing the network structure and adopting attention mechanism, the feature expression ability and the perception ability of the ship's critical parts are improved. Secondly, based on semantic information, a semantic mask module is designed to reduce the impact of background on detection accuracy. In addition, the angle parameter is added to make the network applicable to targets with orientation. Finally, a ship's critical parts dataset, named CP-Ship, is constructed to verify the effectiveness of the proposed network. The experimental results on the CP-Ship dataset show that the average accuracy of the proposed network is 11.35% higher than RetinaNet. Compared with other network models, the proposed network performs well in both detection accuracy and speed.
- deep learning /
- remote sensing /
- critical part detection /
- semantic feature

HTML全文

图 1 CPDNet的整体架构

Fig. 1 Overall architecture of the CPDNet

下载: 全尺寸图片幻灯片

图 2 FE-FPN结构图

Fig. 2 The structure of FE-FPN

下载: 全尺寸图片幻灯片

图 3 CIEM结构图

Fig. 3 The structure of CIEM

下载: 全尺寸图片幻灯片

图 4 SFEM结构图

Fig. 4 The structure of SFEM

下载: 全尺寸图片幻灯片

图 5 标注示意图

Fig. 5 Schematic diagram of labeling

下载: 全尺寸图片幻灯片

图 6 特征图可视化示例

Fig. 6 Visualization examples of feature map

下载: 全尺寸图片幻灯片

图 7 不同网络模型的PR曲线

Fig. 7 The PR curves of different network models

下载: 全尺寸图片幻灯片

图 8 不同网络模型的检测结果可视化示例

Fig. 8 Visualization examples of detection results of different network models

下载: 全尺寸图片幻灯片

表 1 SFEM各层的详细参数

Table 1 Detailed parameters of each SFEM layer

卷积层	通道数（输入）	通道数（输出）	大小（输入）	大小（输出）
Deconv 1	4 096	2 048	W×H	2W×2H
Conv 1	2 048	1 024	2W×2H	2W×2H
Deconv 2	2 048	1 024	2W×2H	4W×4H
Conv 2	1 024	512	4W×4H	4W×4H
Conv 3	512	128	4W×4H	4W×4H
Conv 4	128	64	4W×4H	4W×4H
Conv 5	64	2	4W×4H	4W×4H

下载: 导出CSV

表 2 CP-Ship数据集详细信息

Table 2 Details of CP-Ship dataset

数据集	CP-Ship
数据集	图像数量	关重部位个数
训练集	812	2 295
测试集	203	561
总计	1 015	2 856

下载: 导出CSV

表 3 膨胀结构元素尺寸选取

Table 3 Selection of expansion structure element size

结构元素尺寸	AP/%	结构元素尺寸	AP/%
3×3	68.67	13×13	71.38
5×5	69.69	15×15	70.97
7×7	70.28	17×17	69.94
9×9	71.25	19×19	68.32
11×11	71.76	21×21	66.46

下载: 导出CSV

表 4 消融实验结果

Table 4 Experiment results of ablation

RetinaNet	角度参数	FE-FPN	SMN	AP/%
√				60.41
√	√			62.64
√	√	√		66.45
√	√	√	√	71.76

下载: 导出CSV

表 5 不同网络模型的定量结果

Table 5 Quantiative results of different network models

类别	方法	主干网络	锚框类型	TP	FP	FN	AP/%	FPS
双阶段网络	Faster R-CNN^[9]	ResNet-50	水平框	411	152	150	65.78	19.2
	ReDet^[25]	ResNet-50	旋转框	454	103	107	75.24	3.1
	Dynamic R-CNN^[26]	ResNet-50	水平框	400	130	161	63.85	20.5
	Oriented R-CNN^[27]	ResNet-50	旋转框	423	169	138	68.39	11.8
单阶段网络	YOLOv3^[28]	DarkNet-53	水平框	413	84	148	66.49	51.8
	R3Det^[15]	ResNet-50	旋转框	393	146	168	64.28	11.3
	RetinaNet^[13]	ResNet-50	水平框	387	229	174	60.41	23.2
	CFA^[29]	ResNet-50	旋转框	414	225	306	63.61	19.3
	CPDNet	ResNet-50	旋转框	433	150	128	71.76	12.5

下载: 导出CSV

参考文献(29)

[1]	HUANG Z. A sea-land segmentation algorithm based on graph theory[C]//2nd ISPRS International Conference on Computer Vision in Remote Sensing (CVRS 2015), Xiamen, China: ISPRS, 2016: 990111.
[2]	XU J, FU K, SUN X. An invariant generalized Hough transform based method of inshore ships detection[C]//2011 International Symposium on Image and Data Fusion. Tengchong, China: IEEE, 2011: 1-4.
[3]	牛戈, 陈小前, 季明江, 等. 基于注意力机制特征重建网络的舰船目标检测[J]. 上海航天,2021,38(4):128-136. doi: 10.19328/j.cnki.2096-8655.2021.04.017 NIU G, CHEN X Q, JI M J, et al. Ship detection with feature reconstruction network based on attention mechanism[J]. Aerospace Shanghai,2021,38(4):128-136. doi: 10.19328/j.cnki.2096-8655.2021.04.017
[4]	李晨瑄, 顾佼佼, 王磊, 等. 多尺度特征融合的Anchor-Free轻量化舰船要害检测算法[J/OL]. 北京航空航天大学学报: 1-16 2-08-31]. DOI: 10.13700/j. bh. 1001-5965.2021. 0050. LI C X, GU J J, WANG L, et al. Warship’s vital parts detection algorithm based on lightweight Anchor-Free network with multi-scale feature fusion[J/OL]. Journal of Beijing University of Aeronautics and Astronautics: 1-16[2022-08-31]. DOI:10.13700/j.bh.1001-5965.2021.0050.
[5]	李晨瑄, 钱坤, 胥辉旗. 基于深浅层特征融合的舰船要害关键点检测算法[J]. 系统工程与电子技术,2021,43(11):3239-3249. doi: 10.12305/j.issn.1001-506X.2021.11.24 LI C X, QIAN K, XU H Q. Key-points detection algorithm based on fusion of deep and shallow features for warship’s vital part[J]. Systems Engineering and Electronics,2021,43(11):3239-3249. doi: 10.12305/j.issn.1001-506X.2021.11.24
[6]	李晨瑄, 李湉雨, 李梓正, 等. 智能化舰船要害检测、轨迹预测与位姿估计算法[J/OL]. 北京航空航天大学学报: 1-18 2-08-31]. DOI: 10.13700/j. bh. 1001-5965.2021. 0253. LI C X, LI T Y, LI Z Z, et al. Intelligent algorithm of warship’s vital parts detection, trajectory prediction and pose estimation[J/OL] Journal of Beijing University of Aeronautics and Astronautics: 1-18[2022-08-31]. DOI:10.13700/j.bh.1001-5965.2021.0253.
[7]	赵微, 惠斌, 张玉晓. 红外舰船目标的要害点检测算法[J]. 红外与激光工程. 2014, 43(1): 48-52. ZHAO W, HUI B, ZHANG Y X. Detection algorithms of aim points of IR warship images[J]. Infrared and Laser Engineering. 2014, 43(1): 48-52.
[8]	GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile: IEEE, 2015: 1440-1448.
[9]	REN S, HE K, GIRSHICK R, et al. Faster r-cnn: towards real-time object detection with region proposal networks[J]. Advances in Neural Information Processing Systems,2015,28:91-99.
[10]	HE K, GKIOXARI G, DOLL A R P, et al. Mask r-cnn[C]//Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy: IEEE, 2017: 2961-2969.
[11]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]// ECCV 2016: Computer Vision – ECCV 2016. Amsterdam, The Netherlands: Springer C, 2016: 21-3 7.
[12]	REDMON J, FARHADI A. YOLO9000: Better, Faster, Stronger[J]. arXiv preprint arXiv,2016:1612.0824v1.
[13]	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceeding of the IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 2980-2988.
[14]	LIU Z, HU J, WENG L, et al. Rotated region based CNN for ship detection[C] //2017 IEEE International Conference on Image Processing (ICIP), Beijing, China: IEEE, 2017: 900-904.
[15]	DING J, XUE N, LONG Y, et al. Learning RoI transformer for oriented object detection in aerial images[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, CA, USA: IEEE, 2019: 2844-2853.
[16]	DANG X, LIU Q, YAN J, et al. R3det: refined single-stage detector with feature refinement for rotating object[J]. arXiv preprint arXiv, 2019: 1908.05612. 文献类型不明, 标注不规范
[17]	韩子硕, 王春平, 付强, 等. 基于超密集特征金字塔网络的SAR图像舰船检测[J]. 系统工程与电子技术. 2020, 42(10): 2214-2222. HAN Z S, WANG C P, FU Q, et al. Ship detection in SAR images based on super dense feature pyramid networks[J]. Systems Engineering and Electronics. 2020, 42(10): 2214-2222.
[18]	WOO S, PARK J, LEE J, et al. Cbam: convolutional block attention module[C]//ECCV 2018: Computer Vision – ECCV 2018, Munich, Germany: Springer C, 2018: 3-19.
[19]	HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA: IEEE, 2016: 770-778.
[20]	ZEILER M D, TAYLOR G W, FERGUS R. Adaptive deconvolutional networks for mid and high level feature learning[C]//2011 International Conference on Computer Vision. Barcelona, Spain: IEEE, 2011: 2018-2025.
[21]	LIU Z, YUAN L, WENG L, et al. A high resolution optical satellite image dataset for ship recognition and some new baselines[C]//International conference on pattern recognition applications and methods, Porto, Portugal : SCITEPRESS, 2017, 2: 324-331. 缺出版地
[22]	LI K, WAN G, CHENG G, et al. Object detection in optical remote sensing images: a survey and a new benchmark[J]. ISPRS Journal of Photogrammetry and Remote Sensing,2020,159:296-307. doi: 10.1016/j.isprsjprs.2019.11.023
[23]	XIA G, BAI X, DING J, et al. DOTA: a large-scale dataset for object detection in aerial images[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA: IEEE, 2018: 3974-3983.
[24]	DENG J, DONG W, SOCHER R, et al. ImageNet: a large-scale hierarchical image database[C]//2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, USA: IEEE, 2009: 248-255.
[25]	HAN J, DING J, XUE N, et al. ReDet: A rotation-equivariant detector for aerial object detection[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA: IEEE, 2021: 2785-2794.
[26]	ZHANG H, CHANG H, MA B, et al. Dynamic R-CNN: towards high quality object detection via dynamic training[C]//ECCV 2020: Computer Vision – ECCV 2020, Glasgow, UK: Springer C, 2020: 260-275.
[27]	XIE X, CHENG G, WANG J, et al. Oriented R-CNN for object detection[C]//2021 IEEE/CVF International Conference on Computer Vision (ICCV), Montreal, QC, Canada: IEEE, 2021: 3500-3509.
[28]	REDMON J, FARHADI A. Yolov3: an incremental improvement[J]. arXiv preprint arXiv, 2018: 1804.02767. 文献类型不明, 标注不规范
[29]	GUO Z, LIU C, ZHANG X, et al. Beyond bounding-box: convex-hull feature adaptation for oriented and densely packed object detection[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA: IEEE, 2021: 8788-8797.