面向无人机视觉制导的自适应目标跟踪方法

doi:10.12382/bgxb.2024.0284

摘要/Abstract

摘要：

为解决无人机制导中跟踪目标尺度变化大、外形变化大、推理速度慢、数据集缺失的问题,提出了一种面向无人机视觉制导的自适应目标跟踪方法。自适应搜索区域机制通过分析制导过程调整搜索区域解决尺度变化快的问题;自适应模板更新机制通过更新模板特征解决外形变化大的问题。此外,该方法在骨干网络引入FasterNet Block,在跟踪头引入无锚机制,减少推理的时间。最后,构建并公开了一个包含12个制导视频的测试数据集 Guidance UAV以评估算法在视觉制导中的性能。实验结果表明,该方法不仅在通用无人机跟踪数据集 UAV123 上适用,而且在 Guidance UAV 上实现了最先进的性能,同时在机载设备Jetson Xavier NX 上保持15f/s的速度。室内无人机制导打击实验证明了算法的有效性。

关键词: 视觉制导, 目标跟踪, 无人机, 自适应, 模板更新, 搜索区域

Abstract:

Object tracking is an important part of UAV visual guidance,and the tracking accuracy determines the combat effectiveness of UAVs on the battlefield.There are many problems with general object tracking in UAV guidance:fast scale change,large appearance change,slow inference speed and lack of dataset.In order to solve the problems,this paper proposes Guidance-Tracker,an adaptive UAV siamese tracker for visual guidance.The adaptive search region mechanism adjusts the search region to fit the fast scale change by analyzing the guidance process,and the adaptive template updating mechanism updates the template feature to fit large appearance change.Besides,FasterNet Block and anchor-free mechanism are respectively introduced in the backbone network and the tracking head to speed up the inference time.In addition,an object tracking test dataset Guidance UAV for visual guidance,which contains 12 videos, is constructed to evaluate the performance of trackers in visual guidance.The experiment results indicate that Guidance-Tracker is not only suitable for use on the general UAV tracking dataset UAV123,but also achieves state-of-the-art performance on the Guidance UAV,while maintaining a speed of 15f/s on the onboard device Jetson Xavier NX.The indoor guidance experiments shows the effectiveness of the Guidance Tracker.

Key words: visual guidance, object tracking, UAV, adaptive, template updating, search region

杨绪祺, 谭启凡, 苏航, 谭浩. 面向无人机视觉制导的自适应目标跟踪方法[J]. 兵工学报, 2025, 46(2): 240284-.

YANG Xuqi, TAN Qifan, SU Hang, TAN Hao. Guidance-Tracker:An Adaptive UAV Siamese Tracker for Visual Guidance[J]. Acta Armamentarii, 2025, 46(2): 240284-.

图/表 12

图1 两种经典算法在仿真环境中执行制导的效果图

Fig.1 Simulated results of two classic algorithms executing guidance tasks in simulation environment

图2 自适应跟踪方法的网络结构图

Fig.2 Network structure of adaptive tracking method

图3 图像大小与路程的函数关系图

Fig.3 Function relationship between shape of image and distance of UAV.

图4 自适应模板更新的伪代码

Fig.4 Pseudocode of adaptive templates updating

图5 基于锚点的跟踪头

Fig.5 Anchor-based tracker

表1 Guidance UAV测试数据集

Table 1 Guidance UAV dataset

参数	数值
初始高度	20~100m
飞行速度	5~40m/s
图像尺寸	1152*720
数据数量	2015
有遮挡段数	4
无目标段数	2
视频帧率	30f/s

表2 对比试验结果

Table 2 The results of comparative experiment

算法	来源	Guidance UAV			UAV 123
算法	来源	IOU₁₀₀	IOU₂₀	EOA_A	成功率	精度
KCF^[3]	TPAMI2014	24.73	14.59	34.11	52.17	65.52
SiamRPN^[6]	CVPR2018	37.17	27.46	44.91	64.32	75.98
DasiamRPN^[12]	ECCV2018	33.34	17.22	43.66	66.11	77.53
SiamRPN++^[8]	CVPR2019	43.26	35.72	48.87	69.62	83.14
SiamFC++^[7]	AAAI2020	33.09	19.20	41.68	65.61	71.34
SiamCAR^[21]	CVPR2020	38.85	31.53	45.12	68.04	78.08
SiamBAN^[12]	CVPR2020	41.90	33.77	42.67	69.48	83.17
TCTrack^[13]	CVPR2022	47.37	37.41	50.82	71.18	86.77
Guidance	Ours	49.24	40.83	51.24	67.72	83.34

图6 不同算法在Guidance UAV下的跟踪效果展示

Fig.6 Tracking performances of different methods on Guidance UAV

表3 比较两种自适应方法对跟踪性能的影响

Table 3 The influences of two adaptive methods on tracking performance

Backbone	搜索	模板	IOU₁₀₀	IOU₂₀	EAO_A	推理速度
Alexnet			33.09	19.20	41.68	11.2
Fastnet Block			35.11	21.76	44.51	17.3
	√		44.81	35.46	48.63	16.5
		√	41.22	29.09	46.31	15.7
	√	√	49.24	40.83	51.24	15.1

图7 不同参数设置的实验结果图

Fig.7 Experimental results of different parameters

图8 室内场景无人机制导打击实验图

Fig.8 Indoor guidance experimental result

表4 不同跟踪算法在室内制导实验的结果

Table 4 Results of indoor guidance experiments using different tracking algorithms

算法	飞行速度/(m·s^-1)	总次数	成功次数
SiamRPN	2	4	3
SiamRPN	3	4	2
SiamFC++	2	4	2
SiamFC++	3	4	1
TCTrack	2	4	0
Ours	2	4	4
	3	4	4
	4	4	4

参考文献 24

[1]	BOLME D S, BEVERIDGE J R, DRAPER B A, et al. Visual object tracking using adaptive correlation filters[C]// Proceedings of 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition. San Francisco,CA,USA: IEEE, 2010:2544-2550.
[2]	F.HENRIQUES J, CASEIRO R, MARTINS P, et al. Exploiting the circulant structure of tracking-by-detection with kernels[C]// Proceedings of 2012 European Conference on Computer Vision Part IV. Florence,Italy: Springer, 2012:702-715.
[3]	HENRIQUES J F, CASEIRO R, MARTINS P, et al. High-speed tracking with kernelized correlation filters[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(3):583-596. doi: 10.1109/TPAMI.2014.2345390 pmid: 26353263
[4]	DANELLJAN M, KHAN F S, FELSBERG M, et al. Adaptive color attributes for real-time visual tracking[C]// Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus,OH,USA: IEEE, 2014:1090-1097.
[5]	LI B, YAN JJ, WU W, et al. High performance visual tracking with siamese region proposal network[C]// Proceedings of 2018 IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City,UT,US: IEEE, 2018:8971-8980.
[6]	BERTINETTO L, VALMADRE J, HENRIQUES J F, et al. Fully-convolutional siamese networks for object tracking[C]// Proceedings of 2016 European Conference on Computer Vision Part II.Amsterdam, The Netherlands: Springer, 2016:850-865.
[7]	XU Y D, WANG Z Y, LI Z X, et al. SiamFC++:Towards robust and accurate visual tracking with targetestimation guidelines[C]// Proceedings of the 2020 AAAI Conference on Artificial Intelligence.New York,NY, US:AAAI, 2020:12549-12556.
[8]	LI B, WU W, WANG Q, et al. SiamRPN++:Evolution of siamese visual tracking with very deep networks[C]// Proceedings of 2019 IEEE Conference on Computer Vision and Pattern Recognition. Long Beach,CA,US: IEEE, 2019:4282-4291.
[9]	WU Y, LIM J, YANG M. Object tracking benchmark[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1834-1848. doi: 10.1109/TPAMI.2014.2388226 pmid: 26353130
[10]	MUELLER M, SMITH N, GHANEM B. A benchmark and simulator for UAV tracking[C]// Proceedings of 2016 European Conference on Computer Vision Part I. Amsterdam,The Netherlands: Springer, 2016:445-461.
[11]	CHEN J R, KAO S, HE H, et al. Run,don't walk:Chasing higher FLOPS for faster neural networks[C]// Proceedings of 2023 IEEE Conference on Computer Vision and Pattern Recognition. Vancouver,BC,Canada: IEEE, 2023:12021-12031.
[12]	ZHU Z, WANG Q, LI B, et al. Distractor-Aware Siamese Networks for Visual Object Tracking[C]// Proceedings of 2016 European Conference on Computer Vision Part IX. Munich,Germany: Springer, 2018:103-119.
[13]	CAO Z A, HUANG Z Y, PAN L, et al. TCTrack:temporal contexts for aerial tracking[C]// Proceedings of 2022 IEEE Conference on Computer Vision and Pattern Recognition. New Orleans,LA,US: IEEE, 2022:14778-14788.
[14]	HOWARD A G, ZHU M L, CHEN B, et al. MobileNets:efficient convolutional neural networks for mobile vision applications[OL]. CoRR,arXiv, 2017. http://arxiv.org/abs/1704.04861
[15]	ZHANG X Y, ZHOU X Y, LIN M X, et al. ShuffleNet:An extremely efficient convolutional neural network for mobile devices[C]// Proceedings of 2018 IEEE Conference on Computer Vision and Pattern Recognition. Salt Lake City,UT,US: IEEE, 2018:6848-6856.
[16]	TIAN Z, SHEN C H, CHEN H, et al. FCOS:Fully convolutional one-stage object detection[C]// Proceedings of 2019 IEEE International Conference on Computer Vision. Seoul,Korea (South): IEEE, 2019:9626-9635.
[17]	DUAN K W, BAI S, XIE L X, et al. CenterNet:Keypoint triplets for object detection[C]// Proceedings of 2019 IEEE International Conference on Computer Vision. Korea (South): IEEE, 2019:6568-6577.
[18]	吴超, 吴绍斌, 李子睿, 等. 基于人机交互的免锚检测和跟踪系统设计[J]. 兵工学报, 2022, 43(10):2565-2575.
	WU C, WU S B, LI Z R, et al. Anchor-free detection and tracking system design based on human-computer interaction[J]. Acta Armamentarii, 2022, 43(10):2565-2575. (in Chinese) doi: 10.12382/bgxb.2021.0682
[19]	谭建豪, 郑英帅, 王耀南, 等. 基于中心点搜索的无锚框全卷积孪生跟踪器[J]. 自动化学报, 2021, 47(4):801-812
	TAN J H, ZHENG Y S, WANG Y N, et al. AFST:Anchor-free fully convolutional siamese tracker with searching center point[J]. Acta Automatica Sinica, 2021, 47(4):801-812. (in Chinese)
[20]	李绍明, 储珺, 冷璐, 等. 目标跟踪中基于IoU和中心点距离预测的尺度估计[J]. 自动化学报, 2024, 50(4):1001-1014.
	LI S M, CHU J, LENG L, et al. Accurate scale estimation with IoU and distance between centroids for object tracking[J]. Acta Automatica Sinica, 2024, 50(4):1001-1014. (in Chinese)
[21]	GUO D Y, WANG J, CUI Y, et al. SiamCAR:Siamese fully convolutional classification and regression for visual tracking[C]// Proceedings of 2020 IEEE Conference on Computer Vision and Pattern Recognition. Seattle,WA,US: IEEE, 2020:6269-6277.
[22]	YU Y C, XIONG Y L, HUANG W L, et al. Deformable siamese attention networks for visual object tracking[C]// Proceedings of 2020 IEEE Conference on Computer Vision and Pattern Recognition. Seattle,WA,US: IEEE, 2020:6727-6736.
[23]	姜珊, 底晓强, 韩成. 融合时空特性的孪生网络视觉跟踪[J]. 兵工学报, 2021, 42(9):1940-1950. doi: 10.3969/j.issn.1000-1093.2021.09.015
	JIANG S, DI X Q, HAN C. Siamese network for visual tracking with temporal-spatial property[J]. Acta Armamentarii, 2021, 42(9):1940-1950. (in Chinese) doi: 10.3969/j.issn.1000-1093.2021.09.015
[24]	张永梅, 赖裕平, 马健喆, 等. 基于视频的装甲车和飞机检测跟踪及轨迹预测算法[J]. 兵工学报, 2021, 42(3):545-554. doi: 10.3969/j.issn.1000-1093.2021.03.010
	ZHANG Y M, LAI Y P, MA J Z, et al. A video-based prediction algorithm for armored vehicle and aircraft detection/tracking and trajectory[J]. Acta Armamentarii, 2021, 42(3):545-554. (in Chinese)