基于多相机信息协同的水面漂浮物跟踪方法研究

doi:10.12305/j.issn.1001-506X.2026.08.05

系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (8): 2545-2559.doi: 10.12305/j.issn.1001-506X.2026.08.05

• 电子技术 • 上一篇

基于多相机信息协同的水面漂浮物跟踪方法研究

陈任飞¹(), 陈佳慧¹(), 李忠文²(), 王思霖¹()

1. 新疆大学经济与管理学院，新疆乌鲁木齐 830046
2. 大连理工大学经济管理学院，辽宁大连 116024

收稿日期:2025-06-19 修回日期:2025-09-25 出版日期:2025-12-08 发布日期:2025-12-08
通讯作者: 李忠文 E-mail:18504281750@xju.edu.cn;20230500005@stu.xju.edu.cn;lizhongweng@dlut.edu.cn;wangsilin0618@126.com
作者简介:陈任飞（1992—），男，副教授，博士，主要研究方向为人工智能、模式识别、先进计算
陈佳慧（2006—），女，本科生，主要研究方向为深度学习
王思霖（1988—），女，讲师，博士，主要研究方向为应急管理、调度优化
基金资助:
新疆维吾尔自治区自然科学基金（2025D01C262；2023D01C30）；国家自然科学基金（72474043；72362032）；新疆维吾尔自治区天池英才项目资助课题

Water surface floating object tracking method based on multi-camera information collaboration

Renfei CHEN¹(), Jiahui CHEN¹(), Zhongwen LI²(), Silin WANG¹()

1. Faculty of Economics and Management，Xinjiang University，Urumqi 830046，China
2. Institute of Systems Engineering，Dalian University of Technology，Dalian 116024，China

Received:2025-06-19 Revised:2025-09-25 Online:2025-12-08 Published:2025-12-08
Contact: Zhongwen LI E-mail:18504281750@xju.edu.cn;20230500005@stu.xju.edu.cn;lizhongweng@dlut.edu.cn;wangsilin0618@126.com

摘要/Abstract

摘要：

漂浮物智能跟踪技术对水环境治理至关重要，然而现有方法受限于复杂背景干扰、精度与效率失衡及单相机视场局限，存在检测精度与跟踪性能不足的问题。本文提出基于多相机信息协同的漂浮物跟踪方法，通过整合轻量级骨干网络和动态特征金字塔网络以平衡检测精度和效率，在核相关滤波算法中引入了快速定向梯度直方图和金字塔尺度估计策略，并提出了改进的图像匹配算法，以实现精确跟踪和匹配；构建以漂浮物监测覆盖相关性为核心的多相机空间联合机制，提出主跟踪相机和辅助跟踪相机的选择策略、3种相机跟踪状态选择方案及协同跟踪姿态关系。6种漂浮物的数据集实验结果表明，目标身份F1（identity document F1，IDF1）分数、目标身份精度（identity document precision，IDP）和目标身份召回率（identity document recall，IDR）分别达到89.16%、94.52%和92.01%，速度达到48.86 f/s，有效解决了复杂环境下的多尺度漂浮物检测与跨相机持续跟踪难题。

关键词: 深度学习, 水面漂浮物, 多相机跟踪, 信息协同

Abstract:

Intelligent tracking technology of floating objects is crucial for water environment governance. However, existing methods are limited by complex background interference, imbalance between accuracy and efficiency, and single camera field of view limitations, resulting in insufficient detection accuracy and tracking performance. This study proposes a floating object tracking method based on multi-camera information collaboration, by integrating a lightweight backbone network and a dynamic feature pyramid network to balance detection accuracy and efficiency, introducing a fast directional gradient histogram and pyramid scale estimation strategy in the kernel correlation filtering algorithm, and proposing improved image matching algorithms for accurate tracking and matching; constructing a joint multi-camera spatial mechanism centered on the relevance of floating object monitoring coverage, the selection strategy of main tracking camera and auxiliary tracking camera, three camera tracking state selection schemes and cooperative tracking attitude relationship are proposed. The experimental results of the dataset of six types of floating objects show that the identity document F1(IDF1), identity document precision(IDP) and identity document recall(IDR) reach 89.16%, 94.52% and 92.01%, respectively, and the speed reaches 48.86 f/s, which effectively solves the difficulties of multi-scale floating object detection and cross-camera continuous tracking in complex environments.

Key words: deep learning, surface floater, multi-camera tracking, information collaboration

中图分类号:

陈任飞, 陈佳慧, 李忠文, 王思霖. 基于多相机信息协同的水面漂浮物跟踪方法研究[J]. 系统工程与电子技术, 2026, 48(8): 2545-2559.

Renfei CHEN, Jiahui CHEN, Zhongwen LI, Silin WANG. Water surface floating object tracking method based on multi-camera information collaboration[J]. Systems Engineering and Electronics, 2026, 48(8): 2545-2559.

图/表 18

图1

图2

图3

图4

图6

图5

表1

图7

表2

表3

表4

图8

图9

表5

表6

表7

图10

图11

参考文献 32

1	RAO J J, XU K, CHEN J B, et al. Sea-surface target visual tracking with a multi-camera cooperation approach[J]. Sensors, 2022, 22 (2): 693. doi: 10.3390/s22020693
2	MACLEOD M, ARP HPH, TEKMAN M, et al. The global threat from plastic pollution[J]. Science, 2021, 373 (6550): 61- 65. doi: 10.1126/science.abg5433
3	CHEN R F, PENG Y, LI Z W, et al. Floating object detection using double-labelled domain generalization[J]. Engineering Applications of Artificial Intelligence, 2024, 133, 108500. doi: 10.1016/j.engappai.2024.108500
4	ZHU J L, LI Q L, GAO C B, et al. Camera-aware re-identification feature for multi-target multi-camera tracking[J]. Image and Vision Computing, 2023, 142, 104889.
5	CHEN R F, WU J, PENG Y, et al. Detection and tracking of floating objects based on spatial-temporal information fusion[J]. Expert Systems with Applications, 2023, 225, 120185. doi: 10.1016/j.eswa.2023.120185
6	JIA T, VRIES D R, KAPELAN Z, et al. Detecting floating litter in freshwater bodies with semi-supervised deep learning[J]. Water Research, 2024, 266, 122405. doi: 10.1016/j.watres.2024.122405
7	WANG M, LIU Y D, LIU H P. An enhanced domain generalization method for object detection based on text guided feature disentanglement[J]. Digital Signal Processing, 2025, 156, 104855. doi: 10.1016/j.dsp.2024.104855
8	YIN L, ZHONG Y, GUO N R, et al. A robust and adaptive kernelised correlation filter with fuzzy-enhanced template matching and motion-aware target candidate prediction[J]. Knowledge-Based Systems, 2025, 325, 113947. doi: 10.1016/j.knosys.2025.113947
9	LI S P, ZHANG C H, SHANG Y L, et al. Early prediction of lithium battery cycle life with variable sequence feature extraction based on recursive kernel canonical correlation analysis[J]. IEEE Trans. on Instrumentation and Measurement, 2025, 74, 1- 10.
10	FRANK H Y, CAI J R, LIU C X, et al. Cooperative multi-camera vehicle tracking and traffic surveillance with edge artificial intelligence and representation learning[J]. Transportation Research Part C, 2023, 148, 103982. doi: 10.1016/j.trc.2022.103982
11	XU X S, DENG H X, SHANG Y Y, et al. Multi-camera multi-cow tracking under non-overhead views[J]. Expert Systems with Applications, 2025, 291, 128519. doi: 10.1016/j.eswa.2025.128519
12	MA Y H, SONG Z X, WANG Y S, et al. Improving the spatial resolution of full-field vibration measurements for hollow structures using a speckle compensation calibration method in multi-camera DIC[J]. Optics Communications, 2025, 585, 131865. doi: 10.1016/j.optcom.2025.131865
13	CHEN L, MA L, ZHAN W, et al. Research on the method of anti-occlusion of surgical instrument tracking based on multi-camera module information fusion[J]. Measurement, 2025, 239, 115480. doi: 10.1016/j.measurement.2024.115480
14	YOU S S, YAO H T, XU C S. Multi-target multi-camera tracking with optical-based pose association[J]. IEEE Trans. on Circuits and Systems for Video Technology, 2020, 31 (8): 3105- 3117.
15	PERI N, KHORRAMSHAHI P, RAMBHATLA S S, et al. Towards real-time systems for vehicle re-identification, multi-camera tracking, and anomaly detection[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020: 132−141.
16	SRIGRAROM S, SIE N J L, CHENG H, et al. Multi-camera multi-drone detection, tracking and localization with trajectory-based re-identification[C]//Proc. of the Second International Symposium on Instrumentation, Control, Artificial Intelligence, and Robotics, 2021: 1131−1137.
17	JIANG N, FANG J L, XU J H, et al. SSD based on contour–material level for domain adaptation[J]. Pattern Analysis and Applications, 2021, 24 (3): 1221- 1229. doi: 10.1007/s10044-021-00986-w
18	陈任飞, 彭勇, 李忠文. 基于持续无监督域适应策略的水面漂浮物目标检测方法[J]. 系统工程与电子技术, 2023, 45 (11): 3391- 3401.
	CHEN R F, PENG Y, LI Z W. A novel detector for floating objects based on continual unsupervised domain adaptation strategy[J]. Systems Engineering and Electronics, 2023, 45 (11): 3391- 3401.
19	CHENG X P, LU F, LIU Y H. Lightweight hybrid model based on MobileNet-v2 and Vision Transformer for human–robot interaction[J]. Engineering Applications of Artificial Intelligence, 2024, 127, 107288. doi: 10.1016/j.engappai.2023.107288
20	LIN T Y, DOLLAR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]// Proc. of the IEEE Conference on Computer Vision and Pattern Recognition, 2017: 24−32.
21	ZHOU S L, LI G L, WANG Q J, et al. Rotor attitude estimation for spherical motors using multiobject kalman KCF algorithm in monocular vision[J]. IEEE Trans. on Industrial Electronics, 2022, 70 (1): 265- 275.
22	FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al. Object detection with discriminatively trained part-based models[J]. IEEE Trans. on Pattern Analysis and Machine Intelligence, 2009, 32 (9): 1627- 1645.
23	BOULKENAFET Z, KOMULAINEN J, HADID A. Face antispoofing using speeded-up robust features and fisher vector encoding[J]. IEEE Signal Processing Letters, 2016, 24 (2): 141- 145.
24	SAFY M, SHI G, AMEIN A S. Semi-automatic image registration using harris corner detection and random sample consensus (RANSAC)[C]//Proc. of the IET International Radar Conference, 2013: 131−142.
25	骆志明, 周易非, 李睿, 等. 数字图像相关法辅助的条纹投影轮廓术相位展开方法[J]. 光学学报, 2024, 44(14): 134−142.
	LUO Z M, ZHOU Y F, LI R, et al. Digital image correlation-assisted phase unwrapping for temporal phase shifting fringe projection profilometry[J]. Acta Optica Sinica, 2024, 44(14): 134−142.
26	CIAPARRONE G, SÁNCHEZ F L, TABIK S, et al. Deep learning in video multi-object tracking: a survey[J]. Neurocomputing, 2020, 381, 61- 88. doi: 10.1016/j.neucom.2019.11.023
27	HE L, LIU G L, TIAN G H, et al. Efficient multi-view multi-target tracking using a distributed camera network[J]. IEEE Sensors Journal, 2019, 20 (4): 2056- 2063.
28	PANG B, LI Y Z, ZHANG Y F, et al. Tubetk: adopting tubes to track multi-object in a one-step training model[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 6308−6318.
29	LI P, ZHANG J B, ZHENG Z, et al. State-aware re-identification feature for multi-target multi-camera tracking[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2019: 1506−1516.
30	HOU Y Z, ZHENG L, WANG Z D, et al. Locality aware appearance metric for multi-target multi-camera tracking[EB/OL]. [2025-05-19]. https: //arxiv.org/abs/1911.12037.
31	YANG X P, YE J, LU J C, et al. Box-grained reranking matching for multi-camera multi-target tracking[C]//Proc. of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2022: 251−262.
32	LIU X C, XU T F, WANG Y, et al. BACTrack: building appearance collection for aerial tracking[J]. IEEE Trans. on Circuits and Systems for Video Technology, 2024, 34 (6): 5002- 5017. doi: 10.1109/TCSVT.2023.3340372

视野区域	漂浮物范围	旋转方向
1	$ \boldsymbol{r }\in \left( {\left. {9\text {π} /8,11\text {π} /8} \right]} \right. $	左上
2	$ \boldsymbol{r }\in \left( {\left. {11\text {π} /8,13\text {π} /8} \right]} \right. $	上
3	$ \boldsymbol{r }\in \left( {\left. {13\text {π} /8,15\text {π} /8} \right]} \right. $	右上
4	$ \boldsymbol{r }\in \left( {\left. {7\text {π} /8,9\text {π} /8} \right]} \right. $	左
5	$ \left\| {\boldsymbol{r}} \right\| \lt R $	不变
6	$ \boldsymbol{r }\in \left[ {0,\text {π} /8} \right] \cup \left( {\left. {15\text {π} /8,2\text {π} } \right]} \right. $	右
7	$ \boldsymbol{r }\in \left( {\left. {7\text {π} /8,9\text {π} /8} \right]} \right. $	左下
8	$ \boldsymbol{r }\in \left( {\left. {3\text {π} /8,5\text {π} /8} \right]} \right. $	下
9	$ \boldsymbol{r }\in \left( {\left. {\text {π} /8,3\text {π} /8} \right]} \right. $	右下

场景	数量	分辨率	帧率/（f/s）
漂浮物尺度变化	15	1 920×1 080	30
水波干扰	15	1 920×1 080	30
光影变化	15	1 920×1 080	30
漂浮物平面外旋转	15	1 920×1 080	30

场景	类别/张	类别/张	实例/张	实例/张
漂浮物尺度变化	27 835	3.03	47 093	5.13
水波干扰	18 736	2.21	37 982	4.49
光影变化	13 092	2.06	28 739	4.52
漂浮物平面外旋转	9 837	1.76	20 913	3.74
总计	69 500	2.35	13 4727	4.55

类型	训练集（70%）		验证集（20%）		测试集（10%）
类型	图像/张	标签/个	图像/张	标签/个	图像/张	标签/个
水葫芦	6 836	34 180	1 953	9 766	976	4 883
塑料瓶	4 453	17 812	1 272	5 089	637	2 545
浮萍	3 333	16 665	952	4 761	477	2 381
金属罐	2 497	7 491	713	2 140	357	1 070
纸盒	2 293	9 172	655	2 621	328	1 310
塑料袋	2 087	10 435	596	2 981	298	1 491
总计	21 499	95 755	6 141	27 358	3 073	13 680

方法	评价指标	相机 15	相机16	相机17	相机22	相机23	相机24	相机25	总体
TRACTA^[27]	IDF1	78.28	81.45	81.23	79.34	80.19	83.65	77.91	80.29
	IDP	79.12	83.24	82.91	82.7	84.23	78.95	75.26	80.92
	IDR	75.15	79.32	78.42	75.84	80.27	74.92	73.36	76.75
TubeTK^[28]	IDF1	69.26	70.37	71.62	69.25	72.97	73.12	70.97	71.08
	IDP	70.13	73.94	72.33	72.82	75.67	74.25	72.77	73.13
	IDR	65.24	68.85	69.83	64.71	69.76	70.24	68.13	68.11
SA_ReID ^[29]	IDF1	70.43	69.92	72.31	73.93	66.25	68.16	60.26	68.75
	IDP	69.35	70.54	74.85	76.12	70.24	67.15	63.52	70.25
	IDR	65.61	63.54	68.86	67.27	62.97	63.63	58.24	64.30
LAAM ^[30]	IDF1	73.44	75.15	72.95	69.83	74.21	77.63	75.94	74.16
	IDP	74.93	77.72	73.81	67.24	73.85	80.26	79.14	75.28
	IDR	67.95	72.73	70.34	67.42	68.55	74.84	70.27	70.30
MCMT ^[31]	IDF1	84.38	87.39	89.27	88.25	86.16	88.29	85.13	86.98
	IDP	88.32	86.28	89.25	88.02	89.93	87.75	88.91	88.35
	IDR	81.93	82.37	86.79	83.28	86.82	88.93	84.24	84.91
BACTrack ^[32]	IDF1	81.03	82.38	84.17	83.19	81.11	84.03	80.23	82.31
	IDP	82.81	83.36	85.28	84.19	83.04	84.92	82.02	83.66
	IDR	78.92	80.17	80.28	81.22	81.17	80.16	78.91	80.12
MCMCT ^[11]	IDF1	86.11	86.29	88.19	87.29	88.28	88.37	87.06	87.37
	IDP	90.28	89.39	93.14	92.17	90.28	92.11	93.82	91.60
	IDR	88.21	89.93	90.06	89.36	89.94	88.25	90.91	89.52
本方法	IDF1	88.26	89.87	90.16	88.46	89.99	89.27	88.11	89.16
	IDP	94.38	93.17	95.45	94.33	93.19	95.93	95.19	94.52
	IDR	91.03	92.39	93.11	91.93	92.72	89.91	92.98	92.01

基于多相机信息协同的水面漂浮物跟踪方法研究

Water surface floating object tracking method based on multi-camera information collaboration

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 18

参考文献 32

相关文章 15

编辑推荐

Metrics

本文评价

实验	单相机参数初始化	多相机空间联合	ACLE	TP/%	FPS(f/s)
1	√	×	6.78	90.08	46.37
2	×	√	6.26	90.54	43.28
3	√	√	4.02	93.65	48.86

目标数量	评价指标	相机 15	相机16	相机17	相机22	相机23	相机24	相机25	总体
1	IDF1	88.26	89.87	90.16	88.46	89.99	89.27	88.11	89.16
	IDP	94.38	93.17	95.45	94.33	93.19	95.93	95.19	94.52
	IDR	91.03	92.39	93.11	91.93	92.72	89.91	92.98	92.01
2	IDF1	87.22	88.93	88.32	86.19	88.71	88.22	87.16	87.82
	IDP	93.91	91.93	94.02	92.91	92.88	93.81	94.03	93.36
	IDR	89.38	91.22	92.31	89.84	90.97	87.81	90.22	90.25
3	IDF1	85.02	85.18	86.02	83.39	86.27	85.29	85.61	85.25
	IDP	90.02	89.01	91.38	89.92	90.17	90.07	91.42	90.28
	IDR	87.39	89.22	90.28	86.92	88.39	85.31	88.31	87.97
4	IDF1	81.23	82.34	83.21	82.15	83.34	83.16	82.40	82.55
	IDP	84.24	86.15	87.21	86.86	88.33	87.26	83.19	86.18
	IDR	76.46	78.34	83.52	80.87	85.29	85.42	80.21	81.44
5	IDF1	79.21	80.43	81.65	81.74	80.53	81.22	80.32	80.73
	IDP	82.31	84.94	86.14	85.25	86.26	85.34	81.74	84.57
	IDR	74.82	77.24	81.82	78.43	83.98	83.76	78.47	79.79
6	IDF1	76.16	78.27	80.44	79.53	78.74	80.84	78.23	78.89
	IDP	79.32	81.43	83.34	83.24	84.21	84.11	79.47	82.16
	IDR	72.63	75.34	79.36	76.14	81.76	82.53	75.84	77.66

[1]	韩成杰, 彭聪, 施苏牧, 王宇, 纪江南. 基于多尺度交互融合网络的作动器故障诊断[J]. 系统工程与电子技术, 2026, 48(7): 2319-2332.
[2]	袁胜智, 王少蕾, 李静, 肖任楷, 谭浩天, 赖建伟, 许江涛. 基于小波变换和改进残差神经网络的故障诊断方法[J]. 系统工程与电子技术, 2026, 48(6): 2031-2041.
[3]	陈垣光, 刘晓凯, 肖思瑶, 张顺生, 崔红元, 陈晓莹, 刘莹. 低信噪比下面向时空特征融合的雷达目标智能检测方法[J]. 系统工程与电子技术, 2026, 48(5): 1523-1531.
[4]	杨泽莹, 陈卓, 徐志明, 艾小锋, 吴其华, 王玲. 空间目标ISAR图像部件正向标记方法[J]. 系统工程与电子技术, 2026, 48(4): 1165-1173.
[5]	林炅, 高朋, 洪文兴. 物理信息神经网络在空天领域的应用与展望[J]. 系统工程与电子技术, 2026, 48(4): 1264-1282.
[6]	程鲲, 雷洪涛, 吕志轩. 改进YOLOv8的轻量级无人机跟踪方法[J]. 系统工程与电子技术, 2026, 48(3): 737-750.
[7]	张龙, 朱连宏, 杨波, 雷震, 冯轩铭. MFA-Net：一种面向复杂对抗环境的反舰导弹智能识别多模态自适应融合网络[J]. 系统工程与电子技术, 2026, 48(3): 894-907.
[8]	田弘宇, 葛松虎, 郭宇, 崔中普, 梁潇. 低干噪比下多特征融合的通信干扰识别方法[J]. 系统工程与电子技术, 2026, 48(3): 1061-1071.
[9]	包玉刚, 贾皓翔, 赵旦峰. 基于高阶递归网络的单幅图像去雨滴模型[J]. 系统工程与电子技术, 2026, 48(1): 12-21.
[10]	陈海青, 汪刘应, 刘顾, 王龙, 葛超群, 陈孟州. 合成孔径雷达图像去噪算法研究进展[J]. 系统工程与电子技术, 2026, 48(1): 94-105.
[11]	李开明, 张袁鹏, 罗迎, 代肖楠. 弹道导弹雷达目标识别研究进展[J]. 系统工程与电子技术, 2025, 47(9): 2870-2889.
[12]	季策, 马相宇, 牟晓宇, 赵家毅. TS-GRU-VTA：基于深度学习的车辆信道估计方案[J]. 系统工程与电子技术, 2025, 47(9): 3093-3098.
[13]	闫小伟, 凌冲, 石胜斌. 地表未爆子弹药快速检测系统设计与实现[J]. 系统工程与电子技术, 2025, 47(8): 2639-2645.
[14]	张琬滢, 高由兵, 李泽一, 李鹏飞, 张伟. 基于双路径特征融合网络的背景信号抑制算法[J]. 系统工程与电子技术, 2025, 47(7): 2406-2413.
[15]	张新征, 闫梦可, 朱晓林. 噪声伪标签容忍的半监督SAR目标识别[J]. 系统工程与电子技术, 2025, 47(6): 1796-1805.