基于改进重采样的粒子滤波红外车辆跟踪算法
|
摘要
针对传统粒子滤波算法状态方程无法利用多帧信息及重采样阶段的粒子种类缺失问题,提出基于改进重采样的粒子滤波红外车辆跟踪算法,对图像进行预处理,增强图像;引入图像准则并结合多帧信息对粒子滤波的状态方程予以改良,在保证以目标帧间变化为基础的前提下将图像信息更多的结合在状态方程中,提高算法的抗干扰能力;在重采样阶段利用粒子权值设定阈值,在保留原始大权重粒子的基础上引入受小权重粒子影响的新粒子,抑制粒子权重过于集中,保证粒子的多样性。经过实验验证,提出的算法在精确性与抗干扰性方面与传统粒子滤波方法相比有较大提升。
The state equation can't use the multi-frame information and the particle species in resampling stage are lack in the traditional particle filter algorithm.In view of these problems,a particle filter algorithm based on improved resampling is proposed in the detection and tracking stage.Firstly,the image is preprocessed and enhanced.Then the image criterion is introduced and the state equation of particle filter is improved by combining multi-frame information.As a prerequisite for ensuring the change of target frames,the image information is more combined in the state equation to improve the anti-interference capability of the algorithm.In the resampling stage,the threshold value is set by the particle weight,and the new particle affected by the small weight particle is introduced based on reserving the large weight particle,which can not only restrain the excessive concentration of particle weight,but also ensure the diversity of particles.The experimental results show that the proposed algorithm is more accurate,and has improved anti-interference capability compared with the traditional particle filter.
The state equation can't use the multi-frame information and the particle species in resampling stage are lack in the traditional particle filter algorithm.In view of these problems,a particle filter algorithm based on improved resampling is proposed in the detection and tracking stage.Firstly,the image is preprocessed and enhanced.Then the image criterion is introduced and the state equation of particle filter is improved by combining multi-frame information.As a prerequisite for ensuring the change of target frames,the image information is more combined in the state equation to improve the anti-interference capability of the algorithm.In the resampling stage,the threshold value is set by the particle weight,and the new particle affected by the small weight particle is introduced based on reserving the large weight particle,which can not only restrain the excessive concentration of particle weight,but also ensure the diversity of particles.The experimental results show that the proposed algorithm is more accurate,and has improved anti-interference capability compared with the traditional particle filter.
引文
[1] SHERLOCK B.Wavelet-based Feature Extraction for Target Recognition and Minefield Detection[R].Final Progress Report on ONR Grant,N000149910091,2002:172-178.
[2] URAS S,GIROSI F,VERRI A,et al.A Computational Approach to Motion Perception[J].Biol Cybern,1988,60(2):79-87.
[3] DEL B A,NESI P,SANZ J C.Optical Flow Computation Using Extended Constraints[J].IEEE Trans Image Process,1996,5(5):720-739.
[4] 李智勇,沈振康,谌海新.一种对斑点目标高精度质心跟踪方法[J].宇航学报,1995(2):33-40.
[5] 林宗坚,刘政荣.从遥感影像提取道路信息的方法评述[J].武汉大学学报(信息科学版),2003(1):90-93.
[6] ?MER O.A Non-uniform Haar Wavelet Method for Numerically Solving Two-dimensional Convection-dominated Equations and Two-dimensional Near Singular Elliptic Equations[J].Computers and Mathematics with Applications,2019,77(7):1799-1820.
[7] 郑永斌,黄新生,丰松江.SIFT和旋转不变LBP相结合的图像匹配算法[J].计算机辅助设计与图形学学报,2010,22(2):286-292.
[8] 范九伦,赵凤.灰度图像的二维Otsu曲线阈值分割法[J].电子学报,2007(4):751-755.
[9] 米向荣,曹建芳.基于局部敏感直方图的鲁棒目标跟踪[J].计算机工程与设计,2018,39(11):3526-3530.
[10] FUKUNAGA K,HOSTETLER L D.The Estimation of the Gradient of a Density Function,with Applications in Pattern Recognition[J].IEEE Transactions on Information Theory,1975,21(1):32-40.
[11] LU L,FEI M,WANG H,et al.A New Meanshift Target Tracking Algorithm by Combining Feature Points from Gray and Depth Images[J].Chinese Journal of Liquid Crystals & Displays,2017(2):545-555.
[12] TANG J,ZHONG Y,ZHANG X,et al.Block Mean Shift Tracking Combing Gradient with Hue Information[J].Telecommunication Engineering,2015,55(9):1019-1024.
[13] 高璐,张大志,田金文.红外序列图像目标跟踪的自适应Kalman滤波方法[J].红外与激光工程,2007(5):729-732.
[14] 李蔚,李辉.多特征融合的优化粒子滤波红外目标跟踪[J].激光与红外,2014,44(1):35-40.
[15] 刘宏,李锦涛,刘群,等.融合颜色和梯度特征的运动阴影消除方法[J].计算机辅助设计与图形学学报,2007(10):1279-1285.
[16] GAUTAM A K,SINHA L K,BHUTIYANI M R.An Innovative Approach for Detection of Armoured Vehicle in Airborne Thermal Imagery Using Morphological Processing and Texture Feature Extraction[J].Journal of Intelligent Systems,2017,26(2):359-370.
[2] URAS S,GIROSI F,VERRI A,et al.A Computational Approach to Motion Perception[J].Biol Cybern,1988,60(2):79-87.
[3] DEL B A,NESI P,SANZ J C.Optical Flow Computation Using Extended Constraints[J].IEEE Trans Image Process,1996,5(5):720-739.
[4] 李智勇,沈振康,谌海新.一种对斑点目标高精度质心跟踪方法[J].宇航学报,1995(2):33-40.
[5] 林宗坚,刘政荣.从遥感影像提取道路信息的方法评述[J].武汉大学学报(信息科学版),2003(1):90-93.
[6] ?MER O.A Non-uniform Haar Wavelet Method for Numerically Solving Two-dimensional Convection-dominated Equations and Two-dimensional Near Singular Elliptic Equations[J].Computers and Mathematics with Applications,2019,77(7):1799-1820.
[7] 郑永斌,黄新生,丰松江.SIFT和旋转不变LBP相结合的图像匹配算法[J].计算机辅助设计与图形学学报,2010,22(2):286-292.
[8] 范九伦,赵凤.灰度图像的二维Otsu曲线阈值分割法[J].电子学报,2007(4):751-755.
[9] 米向荣,曹建芳.基于局部敏感直方图的鲁棒目标跟踪[J].计算机工程与设计,2018,39(11):3526-3530.
[10] FUKUNAGA K,HOSTETLER L D.The Estimation of the Gradient of a Density Function,with Applications in Pattern Recognition[J].IEEE Transactions on Information Theory,1975,21(1):32-40.
[11] LU L,FEI M,WANG H,et al.A New Meanshift Target Tracking Algorithm by Combining Feature Points from Gray and Depth Images[J].Chinese Journal of Liquid Crystals & Displays,2017(2):545-555.
[12] TANG J,ZHONG Y,ZHANG X,et al.Block Mean Shift Tracking Combing Gradient with Hue Information[J].Telecommunication Engineering,2015,55(9):1019-1024.
[13] 高璐,张大志,田金文.红外序列图像目标跟踪的自适应Kalman滤波方法[J].红外与激光工程,2007(5):729-732.
[14] 李蔚,李辉.多特征融合的优化粒子滤波红外目标跟踪[J].激光与红外,2014,44(1):35-40.
[15] 刘宏,李锦涛,刘群,等.融合颜色和梯度特征的运动阴影消除方法[J].计算机辅助设计与图形学学报,2007(10):1279-1285.
[16] GAUTAM A K,SINHA L K,BHUTIYANI M R.An Innovative Approach for Detection of Armoured Vehicle in Airborne Thermal Imagery Using Morphological Processing and Texture Feature Extraction[J].Journal of Intelligent Systems,2017,26(2):359-370.