低光照下的改进的CAMShift人脸跟踪算法  

Face tracking algorithm based on improved CAMShift in low illumination

作　　者：张林 侯劲 ZHANG Lin;HOU Jin(Sichuan University of Science&Engineering,Yibin 644000,China)

机构地区：[1]四川轻化工大学,四川宜宾644000

出　　处：《现代电子技术》2021年第21期51-55,共5页Modern Electronics Technique

基　　金：国家自然科学基金(61902268);人工智能四川省重点实验室项目(2017RZI02);四川省科技厅重点项目(20ZDYF0919)。

摘　　要：在低光照的情况下,传统的CAMShift在对人脸跟踪时存在亮度较低,跟踪不到目标、目标运动较快,以至于目标跟丢、目标容易受遮挡等问题。针对传统算法存在的一系列问题,提出一种改进的CAMShift算法,将马尔科夫方向预测与LBP纹理特征融入CAMShift算法中。采用LBP纹理特征对跟踪目标进行检测,可以在低光照情况下获取与目标更接近、更好的目标特征与边界框,提高算法的跟踪准确率;针对低光照情况下或在受遮挡情况下目标丢失的问题,采用马尔科夫算法进行目标运动的方向预测,可以缩小检测的位置区域,提高算法的效率。对比实验结果表明,改进算法与传统算法相比具有较高的准确率和实时性。In the case of low illumination,the traditional CAMShift is confronted with problems,such as low brightness,failing of object tracking,object being lost due to fast movement and objects being easily blocked in face tracking.In view of the above,an improved CAMShift algorithm is proposed here,which integrates the Markov direction prediction and the LBP(local binary pattern)texture feature.The LBP texture features are used to detect tracking objects,which can obtain better object features and bounding boxes that are closer to the object under low illumination conditions,and improve the tracking accuracy of the algorithm.In view of the object loss under low illumination or in case of occlusion,the Markov algorithm is used to predict the object movement direction,which can reduce the detection position area and improve the efficiency of the algorithm.The results of contrast experiment show that the improved algorithm has higher accuracy and real-time performance compared with the traditional algorithm.

关 键 词：低光照 人脸跟踪 目标丢失 马尔科夫预测 LBP纹理特征 CAMSHIFT跟踪算法 实时跟踪 目标检测 

分 类 号：TN911.73-34[电子电信—通信与信息系统]