基于平滑帧差法的混合高斯模型的研究与应用
|
摘要
鉴于传统混合高斯模型在光照突变、噪声干扰时鲁棒性不高,易造成检测错误等问题,提出一种改进的监控视频前景运动目标提取算法.该算法将帧差法平滑化,再与混合高斯模型相结合,分别采用正序和倒序对样本灰度矩阵处理,增强识别准确度.通过基于灰度图的协方差矩阵导出仿射不变量,根据已有的仿射不变量对灰度图进行识别、分析,得到在晃动视频下的前景运动目标的较准确提取.不同场景的视频检测结果表明,改进算法有效克服了监控摄像头晃动或偏移、光照突变、噪声干扰、"空洞"及"双影"现象,与同类算法相比,具有更高的准确度和鲁棒性.
In view of the problem that the robustness is not high during light mutation and noise interference of the traditional Gaussian mixture model,which easily causes detection errors and other problems,this paper proposes an improved algorithm for foreground moving object detection of video surveillance.The algorithm smoothes the frame difference method.And then combined with the Gaussian mixture model,the algorithm respectively uses the positive sequence and the reverse sequence to the sample gray matrix to enhance the recognition accuracy.The affine invariant is derived from the covariance matrix based on grayscale and the grayscale image is identified and analyzed based on the existing affine invariant,so as to obtain a more accurate extraction of the foreground moving object under shaking video.The video detection results of different scenes show that the improved algorithm effectively overcomes the phenomenon of camera shake or shift,light mutation,noise interference,cavity and double shadow.Compared with other similar algorithms,it has higher accuracy and robustness.
In view of the problem that the robustness is not high during light mutation and noise interference of the traditional Gaussian mixture model,which easily causes detection errors and other problems,this paper proposes an improved algorithm for foreground moving object detection of video surveillance.The algorithm smoothes the frame difference method.And then combined with the Gaussian mixture model,the algorithm respectively uses the positive sequence and the reverse sequence to the sample gray matrix to enhance the recognition accuracy.The affine invariant is derived from the covariance matrix based on grayscale and the grayscale image is identified and analyzed based on the existing affine invariant,so as to obtain a more accurate extraction of the foreground moving object under shaking video.The video detection results of different scenes show that the improved algorithm effectively overcomes the phenomenon of camera shake or shift,light mutation,noise interference,cavity and double shadow.Compared with other similar algorithms,it has higher accuracy and robustness.
引文
[1]Barron J L,Fleet D J,Beauchemin S.Performance of optical flow techniques[J].International Journal of Computer Vision,1994,12(1):42-77.
[2]Wang B,Chen Z,Wang J,et al.A fast passenger detection algorithm based on head area characteristics[J].Procedia Engineering,2012,29(4):184-188.
[3]Dai Kexue.Prospects and current studies on back-ground subtraction techniques for moving objects detection from surveillance video[J].Journal of Image and Graphics,2006,11(7):917-919.
[4]Lee Darshyang.Effective Gaussian mixture learning for video background subtraction[J].IEEE Transactions on Pattern Analysis and Machine Intelligencev,2005,27(5):827-832.
[5]Fan Wenchao,Li Xiaoyu,Wei Kai,et al.Moving targetdetection based on improved Gaussian miture model[J].Computer Science,2015,42(5):286-289.
[6]王小峰,王吉华,王远鹏,等.混合高斯模型和差分法相融合的运动目标检测[J].小型微型计算机系统,2013,34(3):685-688.
[7]陈宝远,霍智超,陈光毅,等.一种改进的三帧差分运动目标检测算法[J].应用科技,2016,43(2):10-13.
[8]袁国武,陈志强,龚健,等.一种结合光流法与三帧差分法的运动目标检测算法[J].小型微型计算机系统,2013,34(3):668-671.
[9]黄晓生,黄萍,曹义亲,等.一种基于PCP的块稀疏RPCA运动目标检测算法[J].华东交通大学学报,2013,30(5):30-36.
[10]郭伟,高媛媛,刘鑫焱.一种改进的基于混合高斯模型的运动目标检测算法[J].计算机工程与应用,2016,52(13):195-200.
[11]亢洁,李晓静.基于RPCA与三帧差分融合的运动目标检测[J].计算机应用研究,2018,35(5):1-7.
[12]赵晓刚,冯全,王书志.一种基于帧间差分与模板匹配的河水表面流速测量方法[J].计算机应用与软件,2017,34(9):68-71,107
[13]田枫.基于高斯混合模型的视频图像人数统计算法研究[J].计算机应用与软件,2017,34(6):187-192.
[2]Wang B,Chen Z,Wang J,et al.A fast passenger detection algorithm based on head area characteristics[J].Procedia Engineering,2012,29(4):184-188.
[3]Dai Kexue.Prospects and current studies on back-ground subtraction techniques for moving objects detection from surveillance video[J].Journal of Image and Graphics,2006,11(7):917-919.
[4]Lee Darshyang.Effective Gaussian mixture learning for video background subtraction[J].IEEE Transactions on Pattern Analysis and Machine Intelligencev,2005,27(5):827-832.
[5]Fan Wenchao,Li Xiaoyu,Wei Kai,et al.Moving targetdetection based on improved Gaussian miture model[J].Computer Science,2015,42(5):286-289.
[6]王小峰,王吉华,王远鹏,等.混合高斯模型和差分法相融合的运动目标检测[J].小型微型计算机系统,2013,34(3):685-688.
[7]陈宝远,霍智超,陈光毅,等.一种改进的三帧差分运动目标检测算法[J].应用科技,2016,43(2):10-13.
[8]袁国武,陈志强,龚健,等.一种结合光流法与三帧差分法的运动目标检测算法[J].小型微型计算机系统,2013,34(3):668-671.
[9]黄晓生,黄萍,曹义亲,等.一种基于PCP的块稀疏RPCA运动目标检测算法[J].华东交通大学学报,2013,30(5):30-36.
[10]郭伟,高媛媛,刘鑫焱.一种改进的基于混合高斯模型的运动目标检测算法[J].计算机工程与应用,2016,52(13):195-200.
[11]亢洁,李晓静.基于RPCA与三帧差分融合的运动目标检测[J].计算机应用研究,2018,35(5):1-7.
[12]赵晓刚,冯全,王书志.一种基于帧间差分与模板匹配的河水表面流速测量方法[J].计算机应用与软件,2017,34(9):68-71,107
[13]田枫.基于高斯混合模型的视频图像人数统计算法研究[J].计算机应用与软件,2017,34(6):187-192.