基于对抗注意力机制的水下遮挡目标检测算法

doi:10.19678/j.issn.1000-3428.0068279

摘要/Abstract

摘要：

水下环境复杂, 遮挡目标信息缺失严重而难以提取到足够的特征信息, 导致水下遮挡目标易被漏检。为解决该问题, 提出一种基于对抗注意力机制的水下目标检测算法。以Faster R-CNN算法为框架, 提出基于空间注意力机制的对抗生成遮挡样本网络(AOGN)。AOGN与Faster R-CNN网络相互竞争, 通过三阶段训练过程, 在不增加推理负担的情况下学习生成检测网络难以正确区分的样本, 提高Faster R-CNN网络对水下遮挡目标的检测精度。使用Focal loss增加困难样本的损失比重, 解决水下数据集难易样本不平衡的问题。在此基础上, 为获得更丰富的水下目标特征信息, 使用SE-ResNet50代替VGG16作为骨干网络, 通过残差网络和SE模块的结合获得更有效、更丰富的水下目标信息, 提高对检测目标的特征提取能力, 同时加入多条ROIpooling支路实现多尺度特征融合, 增加特征的丰富性。实验结果表明, 该算法在URPC数据集和水下垃圾数据集上分别取得了73.76%和86.85%的平均精度均值(mAP), 遮挡目标漏检率分别达到2%和7%, 相较于其他检测算法能够有效提升检测性能。

关键词: 机器视觉, 水下目标检测, 对抗样本, 损失函数, SE-ResNet50网络, 特征融合

Abstract:

The complexity of underwater environments and the severe lack of occluded target information, make the extraction of sufficient information difficult, resulting in a high omission factor for underwater occlusion targets. To solve this problem, the present study proposes an occluded underwater target detection algorithm based on an improved adversarial attention mechanism. Using Faster R-CNN as an adversary network, the Adversarial Occlusion sample Generation Network (AOGN), which has a competitive relationship with the Faster R-CNN is designed to improve the detection accuracy for occlusion targets. Through a three-stage learning process, AOGN learns how to generate samples that are difficult for the detection network to classify correctly, thereby improving the detection accuracy of the Faster R-CNN for underwater occlusion targets. Subsequently, the Focal loss function is used to increase the proportion of difficult samples in the loss. Finally, to solve the problem of low resolution of underwater images, SE-ResNet50 is used as the backbone in place of VGG16, thereby enhancing the feature extraction ability. Furthermore, multi-scale feature fusion is adopted based on multi-ROIpooling branches to increase the richness of features. The proposed algorithm achieves mean Average Precision (mAP) values of 73.76% and 86.85% and omission factor values of 2% and 7%, on the URPC and underwater common trash datasets, respectively. These results demonstrate that the algorithm effectively outperforms existing detection methods.

Key words: machine vision, underwater target detection, adversarial sample, loss function, SE-ResNet50 network, feature fusion

罗偲, 李凯扬, 吴吉花, 任鹏. 基于对抗注意力机制的水下遮挡目标检测算法[J]. 计算机工程, 2024, 50(10): 313-321.

LUO Cai, LI Kaiyang, WU Jihua, REN Peng. Underwater Occlusion Target Detection Algorithm Based on Adversarial Attention Mechanism[J]. Computer Engineering, 2024, 50(10): 313-321.

https://www.ecice06.com/CN/Y2024/V50/I10/313

图/表 12

图1 SE-ResNet模块结构

Fig.1 Structure of SE-ResNet module

图2 AOGN网络结构

Fig.2 AOGN network architecture

图3 随机掩码生成

Fig.3 Random mask generation

图4 预测掩码生成

Fig.4 Predicted mask generation

图5 SAM结构

Fig.5 Structure of SAM

图6 多尺度特征融合

Fig.6 Multi-scale feature fusion

图7 对抗Faster R-CNN模型结构

Fig.7 Adversarial Faster R-CNN model structure

图8 AOGN损失曲线

Fig.8 Loss curve of AOGN

图9 水下遮挡目标测试结果比较

Fig.9 Comparison of underwater occlusion target test results

参考文献 27

1	董金耐, 杨淼, 谢卓冉, 等. 水下图像目标检测数据集及检测算法综述. 海洋技术学报, 2022, 41(5): 60- 72.
	DONG J N, YANG M, XIE Z R, et al. Overview of underwater image object detection data set and detection algorithms. Journal of Ocean Technology, 2022, 41(5): 60- 72.
2	张鹏程, 王冰洁, 徐航, 等. 布尔混沌调制激光雷达水下目标成像实验研究. 半导体光电, 2022, 43(5): 999- 1004.
	ZHANG P C, WANG B J, XU H, et al. Experimental study on underwater target imaging of Boolean chaos modulated LiDAR. Semiconductor Optoelectronics, 2022, 43(5): 999- 1004.
3	张胜虎, 马惠敏. 遮挡对于目标检测的影响分析. 图学学报, 2020, 41(6): 891- 896.
	ZHANG S H, MA H M. An analysis of occlusion influence on object detection. Journal of Graphics, 2020, 41(6): 891- 896.
4	李柯泉, 陈燕, 刘佳晨, 等. 基于深度学习的目标检测算法综述. 计算机工程, 2022, 48(7): 1- 12. URL
	LI K Q, CHEN Y, LIU J C, et al. Survey of deep learning-based object detection algorithms. Computer Engineering, 2022, 48(7): 1- 12. URL
5	GIRSHICK R. Fast R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2015: 1440-1448.
6	REN S, HE K, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137- 1149. doi: 10.1109/TPAMI.2016.2577031
7	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2016: 779-788.
8	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[EB/OL]. [2023-07-05]. https://arxiv.org/abs/1512.02325.
9	SUN X, SHI J Y, LIU L P, et al. Transferring deep knowledge for object recognition in low-quality underwater videos. Neurocomputing, 2018, 275, 897- 908. doi: 10.1016/j.neucom.2017.09.044
10	XU F Q, WANG H B, PENG J J, et al. Scale-aware feature pyramid architecture for marine object detection. Neural Computing and Applications, 2021, 33(8): 3637- 3653. doi: 10.1007/s00521-020-05217-7
11	WANG X L, SHRIVASTAVA A, GUPTA A. A-fast-RCNN: hard positive generation via adversary for object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2017: 2606-2615.
12	王凤随, 王启胜, 陈金刚, 等. 基于注意力机制和Soft-NMS的改进Faster R-CNN目标检测算法. 激光与光电子学进展, 2021, 58(24): 405- 416.
	WANG F S, WANG Q S, CHEN J G, et al. Improved Faster R-CNN target detection algorithm based on attention mechanism and Soft-NMS. Laser & Optoelectronics Progress, 2021, 58(24): 405- 416.
13	赵晓飞, 于双和, 李清波, 等. 基于注意力机制的水下目标检测算法. 扬州大学学报(自然科学版), 2021, 24(1): 62- 67.
	ZHAO X F, YU S H, LI Q B, et al. Underwater object detection algorithm based on attention mechanism. Journal of Yangzhou University (Natural Science Edition), 2021, 24(1): 62- 67.
14	连泽宇, 田景文. 基于掩膜生成网络的遮挡人脸检测方法. 计算机工程, 2021, 47(11): 292-297, 304. URL
	LIAN Z Y, TIAN J W. Occluded face detection method based on mask generation network. Computer Engineering, 2021, 47(11): 292-297, 304. URL
15	史建柯, 乔美英, 李冰锋, 等. 基于注意力机制的水下遮挡目标检测算法. 电子科技, 2023, 36(5): 62- 70.
	SHI J K, QIAO M Y, LI B F, et al. Underwater occlusion target detection algorithm based on attention mechanism. Electronic Science and Technology, 2023, 36(5): 62- 70.
16	LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision. Washington D. C., USA: IEEE Press, 2017: 2980-2988.
17	HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2016: 770-778.
18	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Washington D. C., USA: IEEE Press, 2018: 7132-7141.
19	HSIAO T Y, CHANG Y C, CHOU H H, et al. Filter-based deep-compression with global average pooling for convolutional networks. Journal of Systems Architecture, 2019, 95, 9- 18. doi: 10.1016/j.sysarc.2019.02.008
20	SHU Y Y, YU B S, XU H M, et al. Improving fine-grained visual recognition in low data regimes via self-boosting attention mechanism[EB/OL]. [2023-07-05]. https://arxiv.org/abs/2208.00617.
21	ZUO X, LI J J, HUANG J, et al. Pedestrian detection based on one-stage YOLO algorithm. Journal of Physics: Conference Series, 2021, 1871(1): 012131.
22	YAN Z Y, YAN M L, SUN H, et al. Cloud and cloud shadow detection using multilevel feature fused segmentation network. IEEE Geoscience and Remote Sensing Letters, 2018, 15(10): 1600- 1604.
23	LEI F, TANG F F, LI S H. Underwater target detection algorithm based on improved YOLOv5. Journal of Marine Science and Engineering, 2022, 10(3): 310.
24	ISA I S, ROSLI M S A, YUSOF U K, et al. Optimizing the hyperparameter tuning of YOLOv5 for underwater detection. IEEE Access, 2022, 10, 52818- 52831.
25	XU Q Y, ZHANG X F, CHENG R S, et al. Occlusion problem-oriented adversarial Faster-RCNN scheme. IEEE Access, 2019, 7, 170362- 170373.
26	REDMON J, FARHADI A. YOLOv3: an incremental improvement[EB/OL]. [2023-07-05]. http://arxiv.org/abs/1804.02767v1.
27	LIN W H, ZHONG J X, LIU S, et al. RoIMiX: proposal-fusion among multiple images for underwater object detection[C]//Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing. Washington D. C., USA: IEEE Press, 2020: 2588-2592.

[1]	党小超, 刘涧, 董晓辉, 祝忠彦, 李芬芳. 面向不平衡数据的机械设备故障命名实体识别[J]. 计算机工程, 2024, 50(9): 104-112.
[2]	李俊仪, 李向阳, 龙朝勋, 李海燕, 李红松, 余鹏飞. 基于多级区域选择与跨层特征融合的野生菌分类[J]. 计算机工程, 2024, 50(9): 179-188.
[3]	张华青, 夏张涛, 陆晓庆, 童基均. 基于字形特征的血管外科命名实体识别[J]. 计算机工程, 2024, 50(8): 13-21.
[4]	李华昱, 张智康, 闫阳, 岳阳. 基于知识图谱增强的领域多模态实体识别[J]. 计算机工程, 2024, 50(8): 31-39.
[5]	刘锁兰, 王炎, 王洪元, 朱生升. 基于多流语义图卷积网络的人体行为识别[J]. 计算机工程, 2024, 50(8): 64-74.
[6]	赵婉秋, 张俊虎, 李海涛. 用于建筑物分割的平行结构特征融合网络[J]. 计算机工程, 2024, 50(8): 239-248.
[7]	赵宏, 王枭. 基于Swin-Transformer的黑色素瘤图像病灶分割研究[J]. 计算机工程, 2024, 50(8): 249-258.
[8]	王富平, 刘鸿玮, 张锲石, 段冠庄. 基于深度特征抑制的遮挡人脸识别网络[J]. 计算机工程, 2024, 50(8): 259-269.
[9]	闵莉, 董冰洁, 安冬. 基于多注意力机制与跨特征融合的语义分割算法[J]. 计算机工程, 2024, 50(8): 282-289.
[10]	高爽, 史轶伦, 徐巧枝, 于磊. 基于对比学习的非对称编解码结构的心脏MRI分割研究[J]. 计算机工程, 2024, 50(8): 290-300.
[11]	陈宇航, 杨勇, 先木斯亚·买买提明, 帕力旦·吐尔逊, 樊小超, 任鸽, 刁宇峰. 基于主题感知和语义增强的作文自动评分方法[J]. 计算机工程, 2024, 50(8): 363-371.
[12]	张诗婧, 莫绪涛, 赵行, 董杨林. 基于球面折反射成像和YOLOv7的内螺纹缺陷检测[J]. 计算机工程, 2024, 50(7): 282-292.
[13]	王晋涛, 秦昂, 张元, 陈一飞, 王廷凤, 谢承霖, 邹刚. 基于注意力增强与特征融合的中文医学实体识别[J]. 计算机工程, 2024, 50(7): 324-332.
[14]	谭巨全, 王然. 特征融合下田径录像3D人体动作DTW捕捉算法[J]. 计算机工程, 2024, 50(7): 71-78.
[15]	屠乃威, 焦猛, 阎馨. 复杂环境下输电线路鸟巢目标图像检测模型[J]. 计算机工程, 2024, 50(7): 216-226.

选择文件类型/文献管理软件名称

选择包含的内容