Ship target detection algorithm based on context augmentation information with switchable dilated convolution

doi:10.16411/j.cnki.issn1006-7736.2023.04.013

PDF(13903 KB)

Journal of Dalian Maritime University ›› 2023, Vol. 49 ›› Issue (4) : 116-125. DOI: 10.16411/j.cnki.issn1006-7736.2023.04.013

Ship target detection algorithm based on context augmentation information with switchable dilated convolution

QIU Ruicong^1a,1b,ZHOU Haifeng^*1a,1b, CHEN Ying^1a,1b, ZHANG Xingjie^1c,HUANG Jinman², WENG Weizheng³

Author information +

History +

Abstract

Aiming at the low target detection accuracy of small ships in inland river environment and the problem that ship target detection is easily affected by complex background information near shore, a ship target detection algorithm SACAM-YOLOv5 was proposed based on improved YOLOv5s. Firstly, a context augmentation module was introduced to obtain context information by using the dilated convolution property, and to enhance the detection capability of the model on small ships. Secondly, C3_SAC module was designed to replace C3 modules at specific positions in the original backbone network by using switchable dilated convolutions' feature. By expanding the receptive field of the feature map extracted by the backbone network, the information contained in the feature map was enhanced, the influence of complex background was reduced, thereby the feature extraction capability of the network was improved. Finally, the loss function was improved, and NWD was introduced and together with CIoU_Loss as a new bounding box loss calculation function to reduce the influence of position deviation in small target detection, and improve the detection performance for small ships. Seaships dataset was used to verify the above proposed method. Results show that compared with the benchmark model, the precision, recall, and mAP@0.5 of SACAM-YOLOv5 algorithm are improved by 2.2%, 1.0%, and 2.7%, reaching 98.6%, 98.1% and 99.5% respectively, which meet the requirements of ship target detection performance with certain practical engineering significance.

Key words

ship target / detection algorithm / YOLOv5s / context augmentation / dilated convolution / improved loss function

Cite this article

EndNote

Ris (Procite)

Bibtex

Download Citations

QIU Ruicong, ZHOU Haifeng, CHEN Ying, ZHANG Xingjie, Huang Jinman, WENG Weizheng. Ship target detection algorithm based on context augmentation information with switchable dilated convolution[J]. Journal of Dalian Maritime University. 2023, 49(4): 116-125 https://doi.org/10.16411/j.cnki.issn1006-7736.2023.04.013

References

[1]ZHU C, ZHOU H, WANG R, et al. A noval hierarchical method of ship detection from spaceborne optical image based on shape and texture features[J]. IEEE Transactions on Geoscience and Remote Sensing, 2010, 48(9): 3446-3456.

[2]SCHWEGMANN C P, KLEYNHANS W, SALMON B P. Synthetic aperture radar ship detection using Haar-like features[J]. IEEE Geoscience and Remote Sensing Letters, 2016, 14(2): 154-158.

[3]周慧, 严凤龙, 褚娜, 等. 基于特征金字塔模型的高分辨率遥感图像船舶目标检测[J].大连海事大学学报, 2019, 45(4): 131-138.

ZHOU H, YAN F L, CHU N, et al. Ship target detection in high-resolution remote sensing images based on feature pyramid model[J]. Journal of Dalian Maritime University, 2019, 45(4): 131-138. (in Chinese)

[4]赵江洪, 张晓光, 杨璐, 等. 深度学习的遥感影像舰船目标检测[J].测绘科学, 2020, 45(3): 110-116.

ZHAO J H, ZHANG X G, YANG L, et al. Deep Learning for Ship Target Detection in Remote Sensing Images[J]. Surveying and Mapping Science, 2020, 45(3): 110-116. (in Chinese)

[5]段敬雅, 李彬, 董超, 等. 基于YOLOv2的船舶目标检测分类算法[J].计算机工程与设计, 2020, 41(6): 1701-1707.

DUAN J Y, LI B, DONG C, et al. Ship Target Detection and Classification Algorithm Based on YOLOv2[J]. Computer engineering and Design, 2020, 41(6): 1701-1707. (in Chinese)

[6]CHEN D, SUN S, LEI Z, et al. Ship target detection algorithm based on improved YOLOv3 for maritime image[J]. Journal of Advanced Transportation, 2021, 2021: 1-11.

[7]HAN X, ZHAO L, NING Y, et al. ShipYOLO: an enhanced model for ship detection[J]. Journal of Advanced Transportation, 2021, 2021: 1-11.

[8]周旗开, 张伟, 李东锦, 等. 基于改进YOLOv5s的光学遥感图像舰船分类检测方法[J].激光与光电子学进展, 2022, 59(16): 476-483.

ZHOU Q K, ZHANG W, LI D J, et al. Ship Classification Detection Method in optical Remote Sensing Images Based on Improved YOLOv5s[J]. Laser & Optoelectronics Progress, 2022, 59(16): 476-483. (in Chinese)

[9]XIAO J, ZHAO T, YAO Y, et al. Context augmentation and feature refinement network for tiny object detection[J]. 2021.

[10]QIAO S, CHEN L C, YUILLE A. Detectors: Detecting objects with recursive feature pyramid and switchable atrous convolution[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2021: 10213-10224.

[11]BOCHKOVSKIY A, WANG C Y, LIAO H Y M. Yolov4: Optimal Speed and Accuracy of Object Detection[C]//IEEE Conference on Computer Vision and Pattern Recognition. 2020. ArXiv: 2004.10934.

[12]LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC; IEEE Computer Society, 2017: 2117-2125.

[13]LIU S, QI L, QIN H, et al. Path aggregation network for instance segmentation[C]// Proceedings of the IEEE 2018 Conference on Computer Vision and Pattern Recognition. Washington, DC; IEEE Computer Society, 2018: 8759-8768.

[14]YU F, KOLTUN V. Multi-scale context aggregation by dilated convolutions[J]. arXiv preprint arXiv: 1511. 07122, 2015.

[15]XU C, WANG J, YANG W, et al. Detecting tiny objects in aerial images: A normalized Wasserstein distance and a new benchmark[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2022, 190: 79-93.

[16]SHAO Z, WU W, WANG Z, et al. Seaships: A large-scale precisely annotated dataset for ship detection[J]. IEEE Transactions on Multimedia, 2018, 20(10): 2593-2604.

[17]REDMON J, FARHADI A. Yolov3: An incremental improvement[C]//IEEE Conference on Computer Vision and Pattern Recognition, 2018: 1804.02767.

[18]WANG C Y, BOCHKOVSKLY A, LIAO H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023: 7464-7475.