摘要
视频遥感图像中的目标通常呈现小尺度特性,且易受到遮挡、背景干扰和尺度变化等复杂环境影响。针对现有目标跟踪方法对其运动状态估计不够准确的问题,在核相关滤波(kernel correlation filter,KCF)算法框架的基础上,提出一种融合纹理与颜色特征的视频遥感卫星抗遮挡单目标跟踪算法。该算法使用卡尔曼滤波方法对遮挡的目标进行位置预测;提出将FHOG特征和颜色特征融合的最高输出响应作为预测目标位置信息,提高KCF算法在复杂背景下的跟踪精度,同时设计一个尺度滤波器来实现目标尺度的自适应变化。实验结果表明,该算法在目标受遮挡、背景干扰、尺度变化等复杂环境下对视频遥感影像中的移动目标具有良好的跟踪能力,跟踪平均精度和成功率分别可达87.4%和73.4%,优于其他对比方法。
Targets in video remote sensing images often exhibit small-scale characteristics and are susceptible to occlusion,background interference,and scale changes.To address the problem that existing target tracking methods do not accurately estimate their motion states,this paper proposes a single-target tracking algorithm for video remote sensing satellites that integrates texture and color features to resist occlusion based on the framework of KCF algorithm.The algorithm utilizes the Kalman filtering method to predict the position of occluded targets.It combines the FHOG feature and color feature to enhance the tracking accuracy of the KCF algorithm in complex backgrounds,using the highest output response as the predicted target position information.Additionally,a scale filter is designed to achieve adaptive target scale changes.Experimental results demonstrate that the proposed algorithm exhibits good tracking capabilities for moving targets in satellite video remote sensing images under complex environments such as occlusion,background interference,and scale changes,and the average tracking accuracy and success rate can reach 87.4%and 73.4%,which is superior to that of other comparison methods.
作者
许慧
莫楠
XU Hui;MO Nan(College of Geomatics Science and Technology,Nanjing Tech University,Nanjing 211816,China)
出处
《遥感信息》
北大核心
2025年第2期169-176,共8页
Remote Sensing Information
基金
江苏省高等学校(自然科学研究)面上项目(24KJB420004)。
关键词
视频遥感
目标跟踪
卡尔曼滤波
特征融合
尺度自适应
video remote sensing
target tracking
Kalman filtering
feature fusion
scale adaptation
