HLBP与颜色特征自适应融合的粒子滤波目标跟踪改进算法
|
摘要
针对只采用颜色特征的经典粒子滤波目标跟踪算法无法适用相同颜色干扰情况的缺陷,提出一种结合HLBP特征与颜色特征的自适应粒子滤波跟踪算法.该算法采用Haar型局部二值模式算子(Haar local binary pattern,HLBP)提取的HLBP纹理特征与颜色特征结合,通过自适应权值动态调整颜色特征和纹理特征在追踪过程中的比重,实现颜色纹理特征的自适应融合.实验表明,该算法改进了相同颜色干扰情况下的追踪效果,并在目标被遮挡的情况下仍能持续稳定地追踪,提高了追踪的准确度和适用性.
In this paper,an adaptive particle filter tracking algorithm combining HLBP feature and color feature is proposed. This algorithm is based on imperfection of the classical particle filter target tracking algorithm: it only uses the color feature to track and does not perform effectively in the same color interference condition. Our algorithm uses the Haar local binary model operator to extract the HLBP texture feature and combine it with the color feature. Through the opposed model,we dynamically adjust adaptive weights of the color characteristics and texture features in the tracking process,which achieve the adaptive fusion between the color feature and texture feature. Experiments show that the tracking result is greatly strengthened under the same color interference with our algorithm. What's more,in case of occlusions,our algorithm can still track stably and continuously,thus improving the accuracy and applicability of tracking.
In this paper,an adaptive particle filter tracking algorithm combining HLBP feature and color feature is proposed. This algorithm is based on imperfection of the classical particle filter target tracking algorithm: it only uses the color feature to track and does not perform effectively in the same color interference condition. Our algorithm uses the Haar local binary model operator to extract the HLBP texture feature and combine it with the color feature. Through the opposed model,we dynamically adjust adaptive weights of the color characteristics and texture features in the tracking process,which achieve the adaptive fusion between the color feature and texture feature. Experiments show that the tracking result is greatly strengthened under the same color interference with our algorithm. What's more,in case of occlusions,our algorithm can still track stably and continuously,thus improving the accuracy and applicability of tracking.
引文
[1]YIN H P,CHEN B,CAI Y,et al.Vision-based object detection and tracking:a review[J].Acta automatica sinica,2016,42(10):1 466-1 489.
[2]KRISTAN M,KOVACIC S,LEONARDIS A,et al.A two-stage dynamic model for visual tracking[J].IEEE transactions on systems,man,and cybernetics,part B(cybernetics),2010,40(6):1 505-1 520.
[3]TALHA M,STOLKIN R.Particle filter tracking of camouflaged targets by adaptive fusion of thermal and visible spectra camera data[J].IEEE sensors journal,2014,14(1):159-166.
[4]CEHOVIN L,KRISTAN M,LEONARDIS A.Robust visual tracking using an adaptive coupled-layer visual model[J].IEEE transactions on pattern analysis and machine intelligence,2013,35(4):941-953.
[5]XIAO J,STOLKIN R,LEONARDIS A.Single target tracking using adaptive clustered decision trees and dynamic multi-level appearance models[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Boston,2015:4 978-4 987.
[6]闫河,刘婕,杨德红,等.基于特征融合的粒子滤波目标跟踪新方法[J].光电子·激光,2014,25(10):1 990-1 999.YAN H,LIU J,YANG D H,et al.A new particle filter object tracking algorithm based on feature fusion[J].Journal of optoelectronics·laser,2014,25(10):1 990-1 999.(in Chinese)
[7]刘振涛,王朝英,刘卫群.基于双特征的多模块目标跟踪算法[J].计算机工程与设计,2012,33(1):254-258.LIU Z T,WANG C Y,LIU W Q.Multi-block target tracking algorithm based on dual feature[J].Computer engineering and design,2012,33(1):254-258.(in Chinese)
[8]李远征,卢朝阳,高全学,等.基于多特征融合的均值迁移粒子滤波跟踪算法[J].电子与信息学报,2010(2):411-415.LI Y Z,LU Z Y,GAO Q X,et al.Particle filter and mean shift tracking method based on multi-feature fusion[J].Journal of electronics and information technology,2010(2):411-415.(in Chinese)
[9]文凌艳,尹东.基于HLBP特征的运动阴影检测方法[J].光电工程,2016,43(10):84-89.WEN L Y,YIN D.Moving shadow detection approach based on HLBP feature[J].Opto-electronic engineering,2016,43(10):84-89.(in Chinese)
[10]杨德红,闫河,刘婕,等.HLBP纹理特征Mean Shift目标跟踪算法[J].计算机工程与设计,2016,4:26.YANG D H,YAN H,LIU J,et al.Mean shift target tracking algorithm based on HLBP texture feature[J].Computer engineering and design,2016,4:26.(in Chinese)
[11]刘天时,肖敏敏,李湘眷.自适应的Haar型LBP纹理特征提取算法研究[J].计算机工程与科学,2015,37(7):1 381-1 386.LIU T S,XIAO M M,LI X J.An adaptive Haar LBP texture feature extraction algorithm[J].Computer engineering and science,2015,37(7):1 381-1 386.(in Chinese)
[12]何晓川,许录平,冯冬竹,等.结合HLBP模型与色彩位置信息的动目标检测方法[J].西安电子科技大学学报(自然科学版),2015,42(4):27-33.HE X C,XU L P,FENG D Z,et al.HLBP model method with color and location information about moving objects detection[J].Journal of Xidian university,2015,42(4):27-33.(in Chinese)
[13]DU Y,TIAN J,SAN L,et al.Research on particle filter tracking algorithm based on multi-feature covariance matrix[C]//Control,Automation and Robotics(ICCAR),2017 3rd International Conference on.Nagoya:IEEE,2017:753-757.
[14]WANG L,LI L,KONG D.An improved particle filter tracking algorithm based on motion and appearance features[C]//Computational and Information Sciences(ICCIS),2013 Fifth International Conference on.Shiyan,China:IEEE,2013:110-113.
[15]周书仁,殷建平.基于Haar特性的LBP纹理特征[J].软件学报,2013,24(8):1 909-1 926.ZHOU S R,YIN J P.LBP texture feature based on Haar characteristics[J].Journal of software,2013,24(8):1 909-1 926.(in Chinese)
[16]周书仁,王刚,徐岳峰,等.Haar型LBP纹理特征的行人检测研究[J].计算机工程与应用,2016,52(21):175-179.(in Chinese)ZHOU S R,WANG G,XU Y F,et al.Haar characteristics LBP text feature for pedestrian detection[J].Computer engineering and applications,2016,52(21):175-179.(in Chinese)
[17]TALHA M,STOLKIN R.Particle filter tracking of camouflaged targets by adaptive fusion of thermal and visible spectra camera data[J].IEEE sensors journal,2014,14(1):159-166.
[2]KRISTAN M,KOVACIC S,LEONARDIS A,et al.A two-stage dynamic model for visual tracking[J].IEEE transactions on systems,man,and cybernetics,part B(cybernetics),2010,40(6):1 505-1 520.
[3]TALHA M,STOLKIN R.Particle filter tracking of camouflaged targets by adaptive fusion of thermal and visible spectra camera data[J].IEEE sensors journal,2014,14(1):159-166.
[4]CEHOVIN L,KRISTAN M,LEONARDIS A.Robust visual tracking using an adaptive coupled-layer visual model[J].IEEE transactions on pattern analysis and machine intelligence,2013,35(4):941-953.
[5]XIAO J,STOLKIN R,LEONARDIS A.Single target tracking using adaptive clustered decision trees and dynamic multi-level appearance models[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Boston,2015:4 978-4 987.
[6]闫河,刘婕,杨德红,等.基于特征融合的粒子滤波目标跟踪新方法[J].光电子·激光,2014,25(10):1 990-1 999.YAN H,LIU J,YANG D H,et al.A new particle filter object tracking algorithm based on feature fusion[J].Journal of optoelectronics·laser,2014,25(10):1 990-1 999.(in Chinese)
[7]刘振涛,王朝英,刘卫群.基于双特征的多模块目标跟踪算法[J].计算机工程与设计,2012,33(1):254-258.LIU Z T,WANG C Y,LIU W Q.Multi-block target tracking algorithm based on dual feature[J].Computer engineering and design,2012,33(1):254-258.(in Chinese)
[8]李远征,卢朝阳,高全学,等.基于多特征融合的均值迁移粒子滤波跟踪算法[J].电子与信息学报,2010(2):411-415.LI Y Z,LU Z Y,GAO Q X,et al.Particle filter and mean shift tracking method based on multi-feature fusion[J].Journal of electronics and information technology,2010(2):411-415.(in Chinese)
[9]文凌艳,尹东.基于HLBP特征的运动阴影检测方法[J].光电工程,2016,43(10):84-89.WEN L Y,YIN D.Moving shadow detection approach based on HLBP feature[J].Opto-electronic engineering,2016,43(10):84-89.(in Chinese)
[10]杨德红,闫河,刘婕,等.HLBP纹理特征Mean Shift目标跟踪算法[J].计算机工程与设计,2016,4:26.YANG D H,YAN H,LIU J,et al.Mean shift target tracking algorithm based on HLBP texture feature[J].Computer engineering and design,2016,4:26.(in Chinese)
[11]刘天时,肖敏敏,李湘眷.自适应的Haar型LBP纹理特征提取算法研究[J].计算机工程与科学,2015,37(7):1 381-1 386.LIU T S,XIAO M M,LI X J.An adaptive Haar LBP texture feature extraction algorithm[J].Computer engineering and science,2015,37(7):1 381-1 386.(in Chinese)
[12]何晓川,许录平,冯冬竹,等.结合HLBP模型与色彩位置信息的动目标检测方法[J].西安电子科技大学学报(自然科学版),2015,42(4):27-33.HE X C,XU L P,FENG D Z,et al.HLBP model method with color and location information about moving objects detection[J].Journal of Xidian university,2015,42(4):27-33.(in Chinese)
[13]DU Y,TIAN J,SAN L,et al.Research on particle filter tracking algorithm based on multi-feature covariance matrix[C]//Control,Automation and Robotics(ICCAR),2017 3rd International Conference on.Nagoya:IEEE,2017:753-757.
[14]WANG L,LI L,KONG D.An improved particle filter tracking algorithm based on motion and appearance features[C]//Computational and Information Sciences(ICCIS),2013 Fifth International Conference on.Shiyan,China:IEEE,2013:110-113.
[15]周书仁,殷建平.基于Haar特性的LBP纹理特征[J].软件学报,2013,24(8):1 909-1 926.ZHOU S R,YIN J P.LBP texture feature based on Haar characteristics[J].Journal of software,2013,24(8):1 909-1 926.(in Chinese)
[16]周书仁,王刚,徐岳峰,等.Haar型LBP纹理特征的行人检测研究[J].计算机工程与应用,2016,52(21):175-179.(in Chinese)ZHOU S R,WANG G,XU Y F,et al.Haar characteristics LBP text feature for pedestrian detection[J].Computer engineering and applications,2016,52(21):175-179.(in Chinese)
[17]TALHA M,STOLKIN R.Particle filter tracking of camouflaged targets by adaptive fusion of thermal and visible spectra camera data[J].IEEE sensors journal,2014,14(1):159-166.