一种用于监测奶牛的高精度跟踪定位方法  

A High-precision Tracking and Localization Method for Monitoring Cows

作　　者：赵永祥 张国庆 李灯华 罗巍[1,3,4] 陈宏策 于忠德 ZHAO Yongxiang;ZHANG Guoqing;LI Denghua;LUO Wei;CHEN Hongce;YU Zhongde(North China Institute of Aerospace Engineering,Langfang 065000,China;Key Laboratory of Agricultural Monitoring and Early Warning Technology,Ministry of Agriculture and Rural Affairs,Agricultural Information Institute of Chinese Academy of Agricultural Sciences,Beijing 100081,China;Key Laboratory of Advanced Motion Control,Fujian Provincial Education Department,Minjiang University,Fuzhou 350108,China;Aerospace Remote Sensing Information Processing and Application Collaborative Innovation Center of Hebei Province,Langfang 065000,China)

机构地区：[1]北华航天工业学院,河北廊坊065000 [2]中国农业科学院农业信息研究所农业农村部农业监测预警技术重点实验室,北京100081 [3]闽江学院先进运动控制福建省教育厅重点实验室,福建福州350108 [4]河北省航天遥感信息处理与应用协同创新中心,河北廊坊065000

出　　处：《信息与控制》2025年第1期137-149,160,共14页Information and Control

基　　金：农业农村部农业监测预警技术重点实验室开放基金项目(JCYJKFKT2204);河北省中央引导地方科技发展资金项目(236Z7201G,226Z0302G);中国高校产学研创新基金项目(2021ZYA08001)。

摘　　要：针对目标随机运动和遮挡对无人机在复杂场景中跟踪性能的影响问题,提出了一种基于改进CenterTrack的自主无人机(UAV)跟踪定位方法,用于监测目标。首先,设计了一个特征增强模块,提高了对遮挡目标的跟踪性能。其次,结合基于距离的贪婪匹配和搁浅区域,提出一种两阶段匹配算法,缓解了短时间遮挡造成的跟踪中断问题。最后,采用一种定位算法辅助无人机对目标进行精准定位,提高了跟踪性能。在真实的农场环境中,采用所提方法对目标进行了实际监测。实验结果表明,相较于原始CenterTrack算法,所提的跟踪方法在多目标跟踪上的准确度(MA)提高了5.5%,多目标定位精度(Mp)提高了4.3%,识别的F1分数(IF)增加了5.5%。然而,误跟踪的数量(FP)增加了779,漏检和未检测目标的数量(FN)增加了3387。此外,在真实的场景中,所提方法能够准确地跟踪被遮挡和频繁出入无人机相机视野的目标。实验结果验证了该方法在农场动物监测和跟踪方面具有可行性和有效性。To address the issue of the impact of random target motion and occlusion on UAV tracking performance in complex scenarios,an autonomous UAV tracking and positioning method based on an improved CenterTrack is proposed for target monitoring.Firstly,we design a feature enhancement module to improve the tracking performance of occluded targets.Secondly,we propose a two-stage matching algorithm combined with distance-based greedy matching and stranded regions to alleviate the tracking interruption problem caused by short-time occlusion.Finally,we use a localization algorithm to assist the UAV to accurately locate the target and improve the tracking performance.We apply the proposed method to monitoring the targets in a real farm environment practically.The experimental results show that,compared to the original CenterTrack,the tracking algorithm proposed in this paper increases the multi-object tracking accuracy(MA)by 5.5%,the multi-object positioning precision(Mp)by 4.3%,the identification F1 score(IF)by 5.5%,the number of false positives(FP)by 779,and the number of false negatives(FN),including missed and undetected targets,by 3387.In addition,in real scenarios,the proposed method is able to accurately track targets that are occluded and frequently come in and out of the UAV camera's field of view.The experimental results verify that the method is feasible and effective for farm animal monitoring and tracking.

关 键 词：家畜监测 自主无人机 多目标跟踪 CenterTrack 目标定位 

分 类 号：TP24[自动化与计算机技术—检测技术与自动化装置]