摘要
针对核相关滤波器(KCF)跟踪算法在遮挡环境中存在跟踪漂移及跟踪失败的问题,提出融合卡尔曼预测的改进KCF跟踪方法。采用灰度图像和方向梯度直方图(HOG)特征融合用于检测目标位置;利用平均峰值相关能量(APCE)进行目标遮挡判断并在KCF算法上改进:若发生遮挡情况,则用卡尔曼滤波算法预测下一时刻的目标位置,并代替原KCF算法中的目标,防止跟踪漂移;若无遮挡出现,则用KCF算法继续跟踪,从而实现遮挡环境下目标的有效跟踪。采用OTB100数据集对视频长时间和遮挡序列进行跟踪实验。结果表明:论文算法与KCF算法对比,在遮挡情况下目标跟踪精度和成功率分别提升11.9%和13.4%,说明论文提出算法的成功率和跟踪精度均显著提高。
To address the issues of tracking drift and failure in the kernel correlation filter(KCF)tracking algorithm in occlud-ed environments,an improved KCF tracking method integrating Kalman prediction is proposed.The grayscale image and directional gradient histogram(HOG)feature fusion are used to detect target position.The average peak correlation energy(APCE)is used for target occlusion detection and improving on KCF algorithm.If occlusion occurs,the Kalman filter algorithm is used to predict the tar-get position at the next moment and replace the target in the original KCF algorithm to prevent tracking drift.If no occlusion occurs,the KCF algorithm continues to track,thus achieving effective tracking of targets in occluded environments.The the OTB100 dataset is used for tracking long duration and occlusion sequences in videos.The results show that compared with the KCF algorithm,the proposed algorithm improves the target tracking accuracy and success rate by 11.9%and 13.4%respectively under occlusion,indi-cating a significant improvement in the success rate and tracking accuracy of the proposed algorithm.
作者
张银环
薛静云
韦永全
ZHANG Yinhuan;XUE Jingyun;WEI Yongquan(Civil&Architectural Engineering,Weinan Vocational&Technical College,Weinan 714000;School of Mechatronic Engineering,Xi'an Technological University,Xi'an 710021;Shaanxi Railway Institute,Weinan 714000)
出处
《舰船电子工程》
2024年第8期55-60,共6页
Ship Electronic Engineering
基金
智能建造与人工智能(青年)科技创新团队(编号:WZYQNKETD202309)
渭南市科技局项目“电网互动模式中不平衡负载下的电动汽车充电器关键技术研究”(编号:2022ZDYFJH-134)
陕西省“十四五”教育科学规划2023年度课题项目“元宇宙赋能职业教育,构建‘五维一体’实训教学新形态”(编号:SGY23Y3178)资助。
关键词
核相关滤波器
目标跟踪
卡尔曼滤波
遮挡环境
nuclear correlation filter
target tracking
Kalman filtering
occluded environment Class Number TP39
