基于扩展的Haar-Like特征和LBP特征的人脸压缩跟踪算法

Face compressive tracking algorithm based on extended Haar-Like feature and LBP feature

针对传统的压缩跟踪算法采用简单的Haar-Like特征而在遮挡、光照变化、物体形变及背景干扰等情况下易产生目标漂移而导致跟踪失败的问题,提出了一种基于扩展的Haar-Like特征和局部二值模式(LBP)特征相结合的改进压缩跟踪算法,并运用于特定目标即人脸的跟踪。利用扩展的Haar-Like特征搜索目标的粗略位置,应用LBP特征充分表征人脸并进行精确跟踪来定位人脸目标的最佳位置。与简单的Haar-Like特征相比,LBP可以构建更稳定的目标表观模型,并扩展原有的Haar-Like特征,使算法在不同环境干扰下更鲁棒,同时也提高了跟踪算法的精度。实验证明:改进后的人脸压缩跟踪算法比传统的算法性能更优越。

Aiming at the problem of traditional compressive tracking algorithm based on simple Haar-Like features may suffer from the tracking drift and lost when situations such as occlusion,illumination variation,deformation and background clutter occur,an improved compressive tracking algorithm based on the combination of extended Haar-Like feature and local binary pattern (LBP) feature is proposed,and applied to a particular human face tracking.Firstly,utilizing extended Haar-Like features to search for the coarse location of target,then applying LBP features to fully characterize the human face and localize the optimum position for fine tracking.Compared with the simple Haar-Like features,the LBP can construct a more stable target appearance model,and extending original Haar-Like features make the algorithm more robust in different circumstances interference and can improve the accuracy of the tracking algorithm as well.Experimental results show that the improved face compressive tracking algorithm is superior to the traditional algorithm.