摘要
视觉目标跟踪是无人机监控领域一项重要的关键技术.无人机在高空领域监控的对象具有距离大而且目标小,因此视差和光照变化等方面跟踪性能不高.在背景感知相关滤波器框架提出一种新的响应融合策略,通过对响应图中的峰值强度进行加权实现目标跟踪的精确定位.通过背景感知相关滤波器对跟踪的目标提取的每个特征;利用峰值信噪比来衡量响应值的峰值强度,对每个响应值进行加权融合.利用融合响应图对目标进行精确定位.该算法在UAV123公开的数据集进行实验,与六个视频跟踪算法进行对比分析.实验表明,该算法表现出较好的性能,特别是提高目标跟踪的运动中视差和光照变化的性能.
Visual target tracking is an important key technology in UAV surveillance.The objects monitored by UAV in high altitude have large distance and small target,so the tracking performance of parallax and light changes is not high.In this paper,a new response fusion strategy was proposed in the framework of background aware correlation filter(BACF).The target tracking can be accurately located by weighting the peak intensity in the response graph.Each feature of the tracked target was extracted by the background sensing correlation filter.The peak signal to noise ratio was used to measure the peak intensity of the response value,and each response value was weighted and fused.Finally,the fusion response graph was used to locate the target accurately.The algorithm was tested in the public data set of UAV123 and compared with 6 video tracking algorithms.Experimental results showed that the algorithm performed well,especially in improving the performance of parallax and illumination variation during target tracking.
作者
欧丰林
杨文元
OU Feng-lin;YANG Wen-yuan(0x09School of Information Engineering,Zhangzhou Institute of Technology,Zhangzhou 363000,China;Fujian Key Laboratory of Granular Computing and Application,Minnan Normal University,Zhangzhou 363000,China)
出处
《哈尔滨商业大学学报（自然科学版）》
CAS
2020年第2期183-189,共7页
Journal of Harbin University of Commerce：Natural Sciences Edition
基金
福建省自然科学基金项目(2018J01549)。
关键词
计算机视觉
目标跟踪
无人机
峰值信噪比
相关滤波
多特征
响应融合
computer vision
target tracking
unmanned aerial vehicle
peak signal to noise ratio
correlation filtering
multi feature
response fusion
