基于HSV-HOG的改进TLD目标跟踪方法  

Improved TLD Target Tracking Method Based on HSV-HOG

作　　者：孙春梅 谢明 王婷 Sun Chunmei;Xie Ming;Wang Ting(College of Electrical Engineering and Control Science,Nanjing Tech University,Nanjing 211816,China)

机构地区：[1]南京工业大学电气工程与控制科学学院,南京211816

出　　处：《科技通报》2017年第10期87-91,108,共6页Bulletin of Science and Technology

基　　金：国家重点基础研究发展计划(2015CB857100)

摘　　要：针对目前现有的TLD(跟踪-学习-检测)算法易受阴影、遮蔽、摄像机晃动或是目标快速运动的影响,提出基于HSV-HOG的改进TLD目标跟踪方法。首先,在跟踪初始化前通过加入HSV颜色空间提高TLD算法初始化速度以及抗噪性,使得TLD算法在阴影、抖动的干扰下依然能够实现较好的目标跟踪。若TLD算法选取的跟踪目标受到遮蔽、运动过快,则在算法中加入自适应kalman滤波预测目标物体可能存在的区域,缩小跟踪器的跟踪范围,提高跟踪速度,并在检测器加入后验HOG特性,对已缩小的预测区域进行检测,增强了检测器的判别和检测能力。实验证明,改进的追踪方法具有较好的鲁棒性和跟踪精度。The current existing TLD(tracking learning detection)algorithm is easy to be affected by the light shadow,shadowing,camera shaking or fast motion of the target,this paper propose an improvedTLD target tracking method based on HSV-HOG.Firstly,by adding the HSV color space to improve the speed of initialization and the ability of the noise immunity of the TLD algorithm,the TLD algorithm can achieve better target tracking under the shadow and Jitter interference.If the tracking target of TLD tracking algorithm is seriously covered,the motion speed is too fast,then the TLD algorithm is added to the Kalman filter to predict the target object of current frame which may be exist in a number of small regions,which can narrow the range of tracking,improve the speed of tracking,and TLD detectors with posterior HOG features detect the predicting regional and enhance the discriminant ability and detection efficiency of the detector.The experimental results show that The improved tracking method has better robustness and tracking accuracy.

关 键 词：TLD算子 HSV颜色空间 KALMAN滤波器 后验HOG特性 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]