面向合成孔径雷达图像任意方向舰船检测的改进YOLOv3模型

doi:10.3969/j.issn.1000-1093.2021.08.014

摘要/Abstract

摘要： 为实现合成孔径雷达(SAR)图像舰船检测能同时输出目标位置和方位角估计信息，提出基于改进YOLOv3的任意方向舰船目标检测模型。定义有利于模型参数回归稳定性的角度范围，根据垂直框和旋转框预测结果定义多任务损失函数。通过融合垂直框和旋转框预测结果进行目标方位角估计校正，以进一步提高检测性能。采用SAR舰船目标检测数据集(SSDD+)和高分辨率SAR图像数据集(HRSID)分别进行改进模型的性能测试和迁移测试。实验结果表明：对于SSDD+，当交并比为0.5时平均精度均值mAP_0.5达到了0.841；对HRSID进行迁移测试时，mAP_0.5能够达到0.530；当网络输入尺寸为416×416时处理一帧图像耗时约为25 ms；采用高分辨率可见光舰船数据集(HRSC2016)进行改进模型的适用性测试，mAP_0.5为0.888，超过了部分已知模型的精度；该改进模型适用于纯海洋背景下SAR图像中的舰船目标检测，能够满足舰船目标检测的实时性需求。

关键词: 合成孔径雷达图像, 舰船目标检测, YOLOv3模型, 方位角估计, 多任务损失

Abstract: An improved YOLOv3 model for arbitrary-oriented ship detection is proposed to realize the simultaneous output of both position and aspect angle estimation information for synthetic aperture radar (SAR) ship detection. The scope of target’s aspect angle which is beneficial for the stability of model parameter regression is defined, and the multi-task loss function is defined based on the predictions of both vertical and rotated bounding boxes. The combinations of prediction results of both vertical and rotated bounding boxes are used to rectify target’s aspect angle estimation for improving the detection performance. The SAR ship detection dataset plus (SSDD+) and high resolution SAR images dataset (HRSID) are used to do performance and transferability tests for the proposed model. The experimental results demonstrate that, for SSDD+dataset, the mean average precision reaches 0.841 when intersection over union equals 0.5 (mAP_0.5); mAP_0.5 can reach 0.530 when HRSID is used to perform transferability tests; the proposed model takes about 25 milliseconds to process one frame when the input resolution of the model is 416×416. High resolution ship collection 2016 (HRSC2016) which is a visual ship recognition dataset is also used to verify the adaptability of the proposed model, and mAP_0.5 reaches 0.888, which is superior to some known models. The proposed model can be applied to detect SAR ships in pure sea background, and can meet real-time requirement of ship detection.

Key words: syntheticapertureradarimage, shipdetection, YOLOv3model, aspectangleestimation, multi-taskloss

中图分类号:

TP753

徐英，谷雨，彭冬亮，刘俊，陈华杰. 面向合成孔径雷达图像任意方向舰船检测的改进YOLOv3模型[J]. 兵工学报, 2021, 42(8): 1698-1707.

XU Ying， GU Yu， PENG Dongliang， LIU Jun， CHEN Huajie. An Improved YOLOv3 Model for Arbitrary-oriented Ship Detection in SAR Image[J]. Acta Armamentarii, 2021, 42(8): 1698-1707.

参考文献

［1］马啸, 邵利民, 金鑫, 等. 舰船目标识别技术研究进展[J]. 科技导报, 2019, 37(24): 65-78.
MA X, SHAO L M, JIN X, et al. Advances in ship target recognition technology[J]. Science & Technology Review, 2019, 37(24): 65-78. (in Chinese)
[2］吴良斌. SAR图像处理与目标识别[M]. 北京: 航空工业出版社, 2013.
WU L B. SAR image processing and target recognition[M]. Beijing: Aviation Industry Press, 2013. (in Chinese)
[3］王超, 张红, 吴樊, 等. 高分辨率SAR图像船舶目标检测与分类[M]. 北京: 科学出版社, 2013.
WANG C, ZHANG H, WU F, et al. High-resolution SAR image ship detection and classification[M]. Beijing: Science Press, 2013. (in Chinese)
[4］高贵, 周蝶飞, 蒋咏梅, 等. SAR图像目标检测研究综述[J]. 信号处理, 2008, 24(6): 971-981.
GAO G, ZHOU D F, JIANG Y M, et al. Study on target detection in SAR image: a survey[J]. Signal Processing, 2008, 24(6): 971-981. (in Chinese)
[5］ El-DARYMLI K, MCGUIRE P, POWER D, et al. Target detection in synthetic aperture radar imagery: a state-of-the-art survey[J]. Journal of Applied Remote Sensing, 2013, 7(1): 071598.
[6］杜兰, 王兆成, 王燕, 等. 复杂场景下单通道SAR目标检测及鉴别研究进展综述[J]. 雷达学报, 2020, 9(1): 34-54.
DU L, WANG Z C, WANG Y, et al. Survey of research progress on target detection and discrimination of single-channel SAR images for complex scenes[J]. Journal of Radars, 2020, 9(1): 34-54. (in Chinese)
[7］ LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521: 436-444.
[8］ LIU L, OUYANG W L, WANG X G, et al. Deep learning for generic object detection: a survey[J]. International Journal of Computer Vision, 2020, 128: 261-318.
[9］ LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]∥Proceedings of European Conference on Computer Vision. Amsterdam, The Netherlands:Springer, 2016: 21-37.
[10］ REDMON J, FARHADI A. YOLOv3: An incremental improvement: arXiv: 1804. 02767v1[R/OL]. Ithaca, NY, US: Cornell University, (2018-04-08) [2020-08-06]. https:∥arxiv.org/abs/1804.02767.
[11］ REN S Q, HE K M, 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.
[12］李健伟, 曲长文, 彭书娟, 等. 基于卷积神经网络的SAR图像舰船目标检测[J]. 系统工程与电子技术, 2018, 40(9): 1953-1959.
LI J W, QU C W, PENG S J, et al. Ship detection in SAR images based on convolutional neural network[J]. Systems Engineering and Electronics, 2018, 40(9): 1953-1959. (in Chinese)
[13］ HUANG L Q, LIU B, LI B Y, et al. OpenSARShip: a dataset dedicated to Sentinel-1 ship interpretation[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018, 11(1): 195-208.
[14］ WANG Y Y, WANG C, ZHANG H, et al. A SAR dataset of ship detection for deep learning under complex backgrounds [J]. Remote Sensing, 2019, 11(7): 765.
[15］孙显, 王智睿, 孙元睿, 等. AIR-SARShip-1.0：高分辨率SAR舰船检测数据集[J]. 雷达学报, 2019, 8(6): 852-862.
SUN X, WANG Z R, SUN Y R, et al. AIR-SARShip-1.0: high-resolution SAR ship detection dataset[J]. Journal of Radars, 2019, 8(6): 852-862. (in Chinese)
[16］ WEI S J, ZENG X F, QU Q Z, et al. HRSID: a high-resolution SAR images dataset for ship detection and instance segmentation[J]. IEEE Access, 2020, 8: 120234-120254.
[17］赵保军, 李珍珍, 赵博雅, 等. 基于低复杂度卷积神经网络的星载SAR舰船检测[J]. 北京交通大学学报, 2017, 41(6): 1-7.
ZHAO B J, LI Z Z, ZHAO B Y, et al. Spaceborne SAR ship detection based on low complexity convolution neural network［J］. Journal of Beijing Jiaotong University, 2017, 41(6):1-7. (in Chinese)
[18］李健伟, 曲长文, 彭书娟, 等. 基于生成对抗网络和线上难例挖掘的SAR图像舰船目标检测[J]. 电子与信息学报, 2019, 41(1): 143-149.
LI J W, QU C W, PENG S J, et al. Ship detection in SAR images based on generative adversarial network and online hard examples mining[J]. Journal of Electronics and Information Technology, 2019, 41(1): 143-149. (in Chinese)
[19］ JIAO J, ZHANG Y, SUN H, et al. A densely connected end-to-end neural network for multiscale and multiscene SAR ship detection[J]. IEEE Access, 2018, 6:20881-20892.
[20］ ZHAO J P, ZHANG Z H, YU W X, et al. A cascade coupled convolutional neural network guided visual attention method for ship detection from SAR images[J]. IEEE Access, 2018, 6: 50693-50708.
[21］ CHEN C, HE C, HU C H, et al. A deep neural network based on an attention mechanism for SAR ship detection in multiscale and complex scenarios[J]. IEEE Access, 2019, 7: 104848-104863.
[22］ ZHANG X H, WANG H P, XU C A, et al. A lightweight feature optimizing network for ship detection in SAR image[J]. IEEE Access, 2019, 7: 141662-141678.
[23］杨龙, 苏娟, 李响. 基于深度卷积神经网络的SAR舰船目标检测[J]. 系统工程与电子技术, 2019, 41(9): 1990-1997.
YANG L, SU J, LI X. Ship detection in SAR images based on deep convolutional neural network[J]. Systems Engineering and Electronics, 2019, 41(9): 1990-1997. (in Chinese)
[24］ CUI Z Y, LI Q, CAO Z J, et al. Dense attention pyramid networks for multi-scale ship detection in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11): 8983-8997.
[25］ CHEN C, HE C, HU C H, et al. MSARN: a deep neural network based on an adaptive recalibration mechanism for multiscale and arbitrary-oriented SAR ship detection[J]. IEEE Access, 2019, 7: 159262-159283.
[26］ AN Q Z, PAN Z X, LIU L, et al. DRBox-v2: an improved detector with rotatable boxes for target detection in SAR images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11): 8333-8349.
[27］ 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]∥Proceedings of the 6th International Conference on Pattern Recognition Applications and Methods. Porto, Portugal: SciTePress, 2017: 324-331.
[28］ LIN T Y, DOLLR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]∥Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Honolulu: IEEE, 2017: 936-944.
[29］ ZHANG H Y, CISSE M, DAUPHIN Y N, et al. mixup: beyond empirical risk minimization[C]∥Proceedings of International Conference on Learning Representaitons. Vancouver, BC, Canada: the Association for the Advancement of Artificial Intelligence, 2018.
[30］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 detection in remote sensing based with rotated rectangular region[J]. Journal of Computer-Aided Design & Computer Graphics, 2019, 31(11):1935-1945. (in Chinese)

[1]	李响，苏娟，杨龙. 基于改进YOLOv3的合成孔径雷达图像中建筑物检测算法[J]. 兵工学报, 2020, 41(7): 1347-1359.
[2]	伍薇, 苏娟. 一种基于双向匹配融合的合成孔径雷达景象匹配算法[J]. 兵工学报, 2019, 40(1): 124-133.
[3]	庞海洋，王岩飞，张健. 基于合成孔径雷达图像的伪装遮障目标特性建模技术[J]. 兵工学报, 2016, 37(3): 559-564.
[4]	袁礼海，宋建林，沈涛. 合成孔径雷达围像特征关键度分析与分类算法研究[J]. 兵工学报, 2007, 28(10): 1186-1190.