侧扫声纳检测沉船目标的轻量化DETR-YOLO法

doi:10.12305/j.issn.1001-506X.2022.08.06

系统工程与电子技术 ›› 2022, Vol. 44 ›› Issue (8): 2427-2436.doi: 10.12305/j.issn.1001-506X.2022.08.06

侧扫声纳检测沉船目标的轻量化DETR-YOLO法

汤寓麟¹, 李厚朴^1,*, 张卫东², 边少锋¹, 翟国君³, 刘敏⁴, 张晓平⁵

1. 海军工程大学电气工程学院, 湖北武汉 430033
2. 军委联合参谋部战场环境体系论证中心, 北京 100088
3. 海军海洋测绘研究所, 天津 300061
4. 中国人民解放军91001部队, 北京 100841
5. 中国地质大学(北京)信息网络中心, 北京 100083

收稿日期:2021-09-22 出版日期:2022-08-01 发布日期:2022-08-24
通讯作者: 李厚朴
作者简介:汤寓麟(1996—), 男, 博士研究生, 主要研究方向为水下目标检测、计算机视觉|李厚朴(1985—), 男, 教授, 博士, 主要研究方向为大地测量数学分析研究|张卫东(1981—), 男, 工程师, 本科, 主要研究方向为地理信息系统|边少锋(1961—), 男, 教授, 博士, 主要研究方向为大地测量|翟国君(1961—), 男, 教授, 博士, 主要研究方向为海洋大地测量、海底地形测量|刘敏(1980—), 男, 高级工程师, 博士, 主要研究方向为大地测量海空重力测量与数据处理技术、海洋测绘及海洋地理信息应用|张晓平(1980—), 女, 工程师, 博士, 主要研究方向为机器学习
基金资助:
国家优秀青年科学基金(42122025);国家自然科学基金(41974005);国家自然科学基金(41971416);国家自然科学基金(42074074);湖北省杰出青年科学基金(2019CFA086)

Lightweight DETR-YOLO method for detecting shipwreck target in side-scan sonar

Yulin TANG¹, Houpu LI^1,*, Weidong ZHANG², Shaofeng BIAN¹, Guojun ZHAI³, Min LIU⁴, Xiaoping ZHANG⁵

1. College of Electrical Engineering, Naval University of Engineering, Wuhan 430033, China
2. System Demonstration Center of Battle Environment Security Bureau of the Central Military Commission, Beijing 100088, China
3. Naval Institute of Oceangraphic Surveying and Mapping, Tianjin 300061, China
4. Unit 91001 of the pLA, Beijing 100841, China
5. Information Network Center, China University of Geosciences (Beijing), Beijing 100083, China

Received:2021-09-22 Online:2022-08-01 Published:2022-08-24
Contact: Houpu LI

摘要/Abstract

摘要：

基于YOLOv5算法的侧扫声纳海底沉船目标检测方法虽然在检测精度和速度上取得了不错的成绩, 但是如何在复杂海洋噪声背景下进一步提高小目标检测的准确性、降低重叠目标漏警和虚警率的同时实现模型的轻量化是一个亟需解决的课题。为此, 本文创新融合DETR(end-to-end object detection with transformers)与YOLOv5结构, 提出了基于DETR-YOLO模型的轻量化侧扫声纳沉船目标检测模型。首先, 加入多尺度特征复融合模块, 提高小目标检测能力。然后, 融入注意力机制SENet(squeeze-and-excitation networks), 强化对重要通道特征的敏感性。最后, 采用加权融合框(weighted boxes fusion, WBF)策略, 提升检测框的定位精度和置信度。实验结果表明, 本文模型在测试集AP_0.5和AP_0.5∶0.95值分别达到84.5%和57.7%, 较Transformer和YOLOv5a模型大幅度提高, 以较小的效率损失和权重增加为代价取得了更高的检测精度, 在提升全场景理解能力和小尺度重叠目标处理能力的同时满足轻量化工程部署需求。

关键词: DETR-YOLO模型, 多尺度特征复融合, 加权融合框

Abstract:

Although the side-scan sonar shipwreck target detection method based on the YOLOv5 algorithm has achieved good results in detection accuracy and speed, however, how to further improve the accuracy of small target detection under the background of complex ocean noise, reduce the missed alarm rate and false alarm rate of overlapping targets, and realize the lightweight of the model is an urgent issue to be solved. To this end, this paper innovatively integrates the structure of DETR (end-to-end object detection with transformers) and YOLOv5, and proposes a lightweight side-scan sonar shipwreck detection model based on the DETR-YOLO model. Firstly, a multi-scale feature complex fusion module is added to improve the detection ability of small targets. Then, the attention mechanism SENet (squeeze-and-excitation networks) is integrated strengthen the sensitivity to important channel features. Finally WBF (weighted boxes fusion) weighted fusion frame strategy is adopted to improve the positioning accuracy and confidence of the detection frame. The experimental results show that the AP_0.5 and AP_0.5∶0.95 values of this model in the test set reach 84.5% and 57.7%, respectively, which are greatly improved compared with the Transfermer and YOLOv5a models.At the expense of smaller detection efficiency loss and weight increase, higher detection accuracy has been achieved. At the same time, it can improve the full-scene understanding ability and the small-scale overlapping target processing ability while meeting the needs of lightweight engineering deployment.

Key words: DETR-YOLO model, multi-scale feature complex fusion, weighted boxes fusion (WBF)

中图分类号:

P227

汤寓麟, 李厚朴, 张卫东, 边少锋, 翟国君, 刘敏, 张晓平. 侧扫声纳检测沉船目标的轻量化DETR-YOLO法[J]. 系统工程与电子技术, 2022, 44(8): 2427-2436.

Yulin TANG, Houpu LI, Weidong ZHANG, Shaofeng BIAN, Guojun ZHAI, Min LIU, Xiaoping ZHANG. Lightweight DETR-YOLO method for detecting shipwreck target in side-scan sonar[J]. Systems Engineering and Electronics, 2022, 44(8): 2427-2436.

图/表 13

图1

图2

图3

图4

图5

图6

图7

图8

图9

表1

表2

图10

图11

参考文献 32

1	LEHARDY P K, MOORE C. Deep ocean search for Malaysia airlines flight 370[C]//Proc. of the Ocean-ST. JOHN's, 2014.
2	王久, 曹敬涛, 卢秉武, 等. 水下目标探测项目设备应用和发展趋势[J]. 珠江水运, 2016, (11): 43- 44. doi: 10.3969/j.issn.1672-8912.2016.11.021
	WANG J , CAO J T , LU B W , et al. Underwater target detection project equipment application and development trend[J]. China Water Transport, 2016, (11): 43- 44. doi: 10.3969/j.issn.1672-8912.2016.11.021
3	赵建虎, 李娟娟, 李萌. 海洋测量的进展及发展趋势[J]. 测绘信息与工程, 2009, 34 (4): 25- 27.
	ZHAO J H , LI J J , LI M . Progress and future trend of hydrographic surveying and charting[J]. Journal of Geomatics, 2009, 34 (4): 25- 27.
4	陈正荣, 王正虎. 多波束和侧扫声纳系统在海底目标探测中的应用[J]. 海洋测绘, 2013, 33 (4): 51- 54. doi: 10.3969/j.issn.1671-3044.2013.04.014
	CHEN Z R , WANG Z H . Research on underwater target detection using side-scan sonar and multibeam sounding system[J]. Hydrographic Survering and Charting, 2013, 33 (4): 51- 54. doi: 10.3969/j.issn.1671-3044.2013.04.014
5	王久, 周健. 侧扫声纳和多波束系统在失事沉船扫测中的综合应用[J]. 中国水运, 2010, 10 (8): 35- 37. doi: 10.3969/j.issn.1006-7973-C.2010.08.020
	WANG J , ZHOU J . Comprehensive application of side-scan sonar and multi-beam system in shipwreck survey[J]. China Water Transport, 2010, 10 (8): 35- 37. doi: 10.3969/j.issn.1006-7973-C.2010.08.020
6	刘陈. 多波束系统、侧扫声纳与磁力仪在海底沉船探测中的比较分析[D]. 北京: 中国地质大学, 2015.
	LIU C. The comparative analysis of multi-beam sounding system, side-scan sonar and magnetometer in the wreck detection[D]. Beijing: China University of Geosciences, 2015.
7	朱建军, 宋迎春, 胡俊, 等. 测绘大数据时代数据处理理论面临的挑战与发展[J]. 武汉大学学报(信息科学版), 2021, 46 (7): 1025- 1031.
	ZHU J J , SONG Y C , HU J , et al. Challenges and development of data processing theory in the era of surveying and mapping big data[J]. Geomatics and Information Science of Wuhan University, 2021, 46 (7): 1025- 1031.
8	赵建虎, 王晓, 张红梅, 等. 侧扫声呐图像分割的中性集合与量子粒子群算法[J]. 测绘学报, 2016, 45 (8): 935- 942.935-942, 951
	ZHAO J H , WANG X , ZHANG H M , et al. The neutrosophic set and quantum-behaved particle swarm optimization algorithm of side-scan sonar image segmentation[J]. Acta Geodaetica et Cartographica Sinica, 2016, 45 (8): 935- 942.935-942, 951
9	王晓. 侧扫声呐图像精处理及目标识别方法研究[D]. 武汉: 武汉大学, 2017.
	WANG X. Research on precise processing of side scan sonar image and object recognition methods[D]. Wuhan: Wuhan University, 2017.
10	朱殿尧, 卞红雨. 侧扫声纳目标自动探测研究[J]. 吉林大学学报(信息科学版), 2008, 26 (6): 627- 631. doi: 10.3969/j.issn.1671-5896.2008.06.014
	ZHU D Y , BIAN H Y . Research of side-scan sonar target auto detection[J]. Journal of Jilin University (Information Science Edition), 2008, 26 (6): 627- 631. doi: 10.3969/j.issn.1671-5896.2008.06.014
11	李海滨, 滕惠忠, 宋海英, 等. 基于侧扫声纳图像海底目标物提取方法[J]. 海洋测绘, 2010, 30 (6): 71- 73.
	LI H B , TENG H Z , SONG H Y , et al. Technology on the extraction of seabed target based on high resolution side-scan sonar[J]. Hydrographic Survering and Charting, 2010, 30 (6): 71- 73.
12	阳凡林, 独知行, 吴自银, 等. 基于灰度直方图和几何特征的声纳图像目标识别[J]. 海洋通报, 2006, (5): 64- 69. doi: 10.3969/j.issn.1001-6392.2006.05.010
	YAN F L , DU Z X , WU Z Y , et al. Object recognizing on sonar image based on histogram and geometric feature[J]. Marine Science Bulletin, 2006, (5): 64- 69. doi: 10.3969/j.issn.1001-6392.2006.05.010
13	吕良, 周超烨, 陈春, 等. 基于虚警函数的侧扫声纳水下目标实时检测方法[J]. 海洋测绘, 2013, 33 (4): 35- 38. doi: 10.3969/j.issn.1671-3044.2013.04.009
	LYU L , ZHOU C Y , CHEN C , et al. Real-time detection of underwater target using side-scan sonar based on false alarm function[J]. Hydrographic Surveying and Charting, 2013, 33 (4): 35- 38. doi: 10.3969/j.issn.1671-3044.2013.04.009
14	朱殿尧, 卞红雨. 侧扫声纳目标自动探测研究[J]. 吉林大学学报(信息科学版), 2008, 26 (6): 627- 631. doi: 10.3969/j.issn.1671-5896.2008.06.014
	ZHU D Y , BIAN H Y . Research of side-scan sonar target auto detection[J]. Journal of Jilin University (Information Science Edition), 2008, 26 (6): 627- 631. doi: 10.3969/j.issn.1671-5896.2008.06.014
15	陈强. 基于水声图像水下目标识别的技术研究[D]. 哈尔滨: 哈尔滨工程大学, 2012.
	CHEN Q. Research-based underwater acoustic images underwater target recognition[D]. Harbin: Harbin Engineering University, 2012.
16	续元君. 水下目标探测关键技术研究[D]. 大连: 大连海事大学, 2011.
	XU Y. Research on the key technologies of detection underwater[D]. Dalian: Dalian Maritime University, 2011.
17	汤寓麟, 金绍华, 边刚, 等. 侧扫声呐识别沉船影像的迁移学习卷积神经网络法[J]. 测绘学报, 2021, 50 (2): 260- 269.
	TANG Y L , JIN S H , BIAN G , et al. The transfer learning with convolutional neural network method of side-scan sonar to identify wreck images[J]. Acta Geodaetica et Cartographica Sinica, 2021, 50 (2): 260- 269.
18	TANG Y L, JIN S H, BIAN G, et al. Wreckage target recognition in side-scan sonar images based on an improved faster R-CNN model[C]//Proc. of the International Conference on Big Data & Artificial Intelligence & Software Engineering, 2020: 348-354.
19	TANG Y L , JIN S H , BIAN G , et al. Shipwreck target recognition in side-scan sonar images by improved YOLOv3 model based on transfer learning[J]. IEEE Access, 2020, 8, 173450- 173460. doi: 10.1109/ACCESS.2020.3024813
20	汤寓麟, 边少锋, 翟国君, 等. 侧扫声纳检测沉船目标的改进YOLOv5法[EB/OL]. [2021-09-01]. https://doi.org/10.13203/j.whugis20210353.
	TANG Y L, BIAO S F, ZHAI G J, et al. Improved YOLOv5 method for detecting shipwreck target with side-scan sonar[EB/OL]. [2021-09-01]. https://doi.org/10.13203/j.whugis20210353.
21	VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Proc. of the Advances in neural Information Processing Systems, 2017: 5998-6008.
22	FEDUS W, ZOPH B, SHAZEER N. Switch transformers: scaling to trillion parameter models with simple and efficient sparsity[EB/OL]. [2021-09-01]. https://arXiv.org/abs/2101.03961.
23	PRANGEMEIER T, REICH C, KOEPPL H. Attention-based transformers for instance segmentation of cells in microstructures[C]//Proc. of the IEEE International Conference on Bioinformatics and Biomedicine, 2020: 700-707.
24	YANG F, YANG H, FU J, et al. Learning texture transformer network for image super-resolution[C]//Proc. of the Computer Vision and Pattern Ecognition, 2020.
25	CHEN H, WANG Y, GUO T, et al. Pre-trained image processing transformer[EB/OL]. [2021-09-01]. https//arXiv.org/abs/2012.00364.
26	ZHANG H, GOODFELLOW I, METAXAS D, et al. Self attention generative adversarial networks[C]//Proc. of the International Conference on Machine Learning, 2019: 7354-7363.
27	CHEN C F, FAN Q, PANDA R. Crossvit: cross-attention multiscale vision transformer for image classification[EB/OL]. [2021-09-01]. https//arXiv.org/abs/2103.14899.
28	CHEN C F, PANDA R, FAN Q. Regionvit: regional-to-local attention for vision transformers[EB/OL]. [2021-09-01]. https//arXiv.org/abs/2106.02689.
29	DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[C]//Proc. of the Computer Vision and Pattern Recongnition, 2021.
30	CARION N, MASSA F, SYNNAEVE G, et al. End-to-end object detection with transformers[C]//Proc. of the European Conference on Computer Vision, 2020.
31	范习健, 李庆武, 黄河, 等. 侧扫声呐图像的3维块匹配降斑方法[J]. 中国图象图形学报, 2012, 17 (1): 68- 74.
	FAN X J , LI Q W , HUANG H , et al. Side-scan sonar image despecking based on block-matching and 3D filtering[J]. Journal of Image and Graphics, 2012, 17 (1): 68- 74.
32	HELLEQUIN L , BOUCHER J M , LURTON X . Processing of high-frequency multibeam echo sounder data for seafloor cha-racterization[J]. IEEE Journal of oceanic Engineering, 2003, 28 (1): 78- 89.

模型	AP_0.5/%	AP_0.5~0.95/%	FPS	权重/MB
Transformer	77.3	43.9	442	15.3
YOLOv5a	81.8	51.6	430	18.6
DETR-YOLO	84.5	57.7	427	20.1

组别	DETR	多尺度特征复融合	SENet	AP_0.5/%	FPS
1	√	－	－	83.16	441
2	√	√	－	84.01	429
3	√	－	√	83.57	438
4	√	√	√	84.52	427

侧扫声纳检测沉船目标的轻量化DETR-YOLO法

Lightweight DETR-YOLO method for detecting shipwreck target in side-scan sonar

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 32

相关文章 0

编辑推荐

Metrics

本文评价