为解决侧扫声呐图像自动检测样本不足的问题，在不增加试验成本和模型复杂度的基础上，依托现有样本，通过增加标签、调整尺寸、区分频率、叠加数量扩增声图样本，结合迁移学习，有效提升了小样本侧扫声呐图像自动检测性能。实验结果显示，YOLOv7模型直接对原始样本进行检测，精确率和召回率的调和平均数为0.617，经过新扩增的4个声图样本进行迁移学习后，调和平均数分别增加了0.267、0.158、0.166和0.306，检测性能均得到了提升。实验数据进一步显示，样本数量和相似度是影响检测效果的重要因素，因此，在现有样本基础上，挖掘图像特征、扩增声图数量、开展迁移学习，是提升小样本侧扫声呐图像自动检测性能的有效手段。
In order to solve the problem of poor detection performance caused by insufficient samples in the automatic detection of side-scan sonar images, based on the existing samples, four methods including adding labels, adjusting size, discriminating frequency and superimposing number were adopted, and transfer learning was carried out by using new sonogram. The automatic detection performance of small sample side scan sonar image is effectively improved. The experimental results show that the YOLOv7 model directly trains and detects the existing samples, and the harmonic average (F1) of the accuracy rate and recall rate is 0.617. After transfer learning of the four acoustic atlas amplified by the new method, the F1 value is increased by 0.267, 0.158, 0.166 and 0.306, respectively, and the detection performance has been improved. Further analysis of the experimental data shows that the similarity between the number of samples and the transferred image is still the core factor affecting the detection effect. Therefore, based on existing images, expanding the number of sonograms, mining image features, and carrying out transfer learning are effective means to improve the automatic detection performance of small sample side-scan sonar images.
2025,47(14): 87-94 收稿日期：2024-8-19
DOI：10.3404/j.issn.1672-7649.2025.14.014
分类号：TP391.41
基金项目：山东省重点研发计划项目（2021JMRH0104）
作者简介：赵桁(1986-),男,博士,工程师,研究方向为海上搜救打捞与声呐图像的自动检测与识别
参考文献：
[1] ZHAO H, HAN S P, XU J F, et al. A review of intelligent detection methods for underwater targets in sonar images[C]// 2023 IEEE 7th Information Technology and Mechatronics Engineering Conference, 2023. China: IEEE, 2023.
[2] PENG S, JIANG W, PI H, et al. Deep snake for real time instance segmentation[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), 2020.
[3] DING M, HUO Y, YI H, et al. Learning depth-guided convolutions for monocular 3D object detection[C]//Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR), 2020.
[4] SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[J]. IEEE Trans Pattern Anal Mach Intell, 2017(4): 640-651.
[5] TAIGMAN Y, YANG M, RANZATO M, et al. Deep face: closing the gap to human-level performance in face verification[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014.
[6] CHEN Q, TIAN J, HUANG H N, et al. Study on SAS image segmentation using SVM based on statistical and texture features[J]. Chinese Journal of Scientific Instrument, 2013, 34(6): 214-221.
[7] WANG T, PAN G F, ZHANG J B. Research on submarine sediment classification based on texture features of side-scan sonar images [C]// Proceedings of the 2020 Western China Acoustics Academic Exchange Conference. Jiuquan : 2020 Western China Acoustics Academic Exchange Conference, 2020.
[8] DONG L G, SHAN R, LIU H M, et al. Shipwreck identification with side scan sonar image based on fractal texture[J]. Marine Geology & Quaternary Geology, 2021, 41(4): 232-239.
[9] LUO J H, JIANG J P, ZHU P M. Automatic extraction of the side-scan sonar imagery outlines based on mathematical morphology[J]. Haiyang Xuebao, 2016, 38(5): 150-157.
[10] 谭志. 基于深度学习的目标检测与识别技术[M]. 北京: 化学工业出版社, 2022.
[11] GIRSHICK R, DONAHUE J, DARRELl T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014.
[12] GIRSHICK R. Fast R-CNN[C]// IEEE International Conference on Computer Vision (ICCV), 2015.
[13] HE K, ZHANG X, REN X, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 37(9): 1904-1916.
[14] REN S, HE K, GIRSHICK R, et al. Faster R-CNN: Towards real-time object detection with region proposal networks[C]// Annual Conference on Neural Information Processing Systems (NIPS), 2015.
[15] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-timed object detection[C]// IEEE Conference on Computer Vision and Pattern Recognition(CVPR), 2016.
[16] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot multibox detector[C]// European Conference on Computer Vision (ECCV), 2016.
[17] LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]// IEEE Conference on Computer Vision and Pattern Recognition, 2017.
[18] LI H, JIANG Q. A river ship target detection algorithm based on YOLOv3 and transfer learning [J]. Journal of Guangxi Academy of Sciences, 2023, 39(3) : 331-339.
[19] LI Q Z, LI Y B, NIU J. Real-time detection of underwater fish based on improved YOLO and transfer learning[J]. Pattern Recognition and Artificial Intelligence, 2019, 32(3) : 193-203.
[20] FAN B, WU J, SUN L, et al. Lightweight mesh target detection based on improved YOLOv3 and transfer learning[J]. Journal of Yunnan University: Natural Sciences Edition. 2022, 44(3): 471-479.
[21] 陈佳辉, 陈岚萍, 夏小云, 等, 基于迁移学习的海底底质声呐图像分类[J]. 计算机仿真, 2022, 39(1): 229-233.
[22] 武铄, 王晓, 张丹阳, 等, 联合迁移学习和深度学习的侧扫声呐沉船识别方法[J]. 河南科技, 2021, 770(36): 36-40.
[23] WANG C Y, BOCHKOVSKIY A, LIAO H Y, YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023.
