基于边缘差分的运动目标检测方法
|
摘要
针对视频序列中的运动目标检测问题,提出了一种新的基于边缘差分的运动目标检测方法。通过改进的边缘检测算法提取视频序列中相邻帧的边缘图像并作差分,采用改进的Otsu法对差分图像进行阈值分割,得到运动目标检测的结果。结合Prewitt算子和Sobel算子改进的边缘检测算法能够获取纹理丰富、细节显著的边缘图像,边缘差分结果更加理想;改进的Otsu法联合类内方差能够很好地抑制噪声,保留更多的纹理细节。实验结果表明,提出的方法能够提取更加完整的目标区域,对背景噪声更加鲁棒。与最近一些同类算法相比,在背景运动和光照变化条件下,该方法具有更加优越的运动目标检测性能。
A new moving object detection algorithm based on edge subtraction is proposed to solve the problem of moving object detection in video sequences.Firstly,two edge images are detected from two adjacent frames through an improved object detection algorithm and subtracts the former from the latter,and then a meliorative Otsu method is used to determine the threshold of edge subtraction image,finally the result of moving object detection is acquired after threshold segmentation.Edge images obtained by the edge detection algorithm integrating Prewitt operator with Sobel operator have abundant texture and notable details,and the results of edge subtraction are more ideal.The meliorative Otsu method combined with within-class variance restrains noise well and reserves more details.The computer simulation demonstrates that the target regions detected by this method are more complete and distinguishable with less noise.Compared with some methods proposed recently,it performs better particularly when the images have illumination changing and background moving.
A new moving object detection algorithm based on edge subtraction is proposed to solve the problem of moving object detection in video sequences.Firstly,two edge images are detected from two adjacent frames through an improved object detection algorithm and subtracts the former from the latter,and then a meliorative Otsu method is used to determine the threshold of edge subtraction image,finally the result of moving object detection is acquired after threshold segmentation.Edge images obtained by the edge detection algorithm integrating Prewitt operator with Sobel operator have abundant texture and notable details,and the results of edge subtraction are more ideal.The meliorative Otsu method combined with within-class variance restrains noise well and reserves more details.The computer simulation demonstrates that the target regions detected by this method are more complete and distinguishable with less noise.Compared with some methods proposed recently,it performs better particularly when the images have illumination changing and background moving.
引文
[1]LUAN Qinglei,ZHAO Weisong.Moving object detection al-gorithm based on three-frame-difference of moving backgroundand edge information[J].Opto-Electronic Engineering,2011,38(10):77-83(in Chinese).[栾庆磊,赵为松.动背景下帧差分法与边缘信息融合的目标检测算[J].光电工程,2011,38(10):77-83.]
[2]WU Junqin,LIU Hao,LUO Yong.Moving object detectionalgorithm based on image sequences fusion[J].Computer En-gineering and Design,2012,33(12):4614-4618(in Chi-nese).[吴君钦,刘昊,罗勇.图像融合的运动目标检测算法研究[J].计算机工程与设计,2012,33(12):4614-4618.]
[3]HOU Zhiqiang,HAN Chongzhao.A background reconstruc-tion algorithm based on pixel intensity classification[J].Jour-nal of Software,2005,16(9):1568-1576(in Chinese).[侯志强,韩崇昭.基于像素灰度归类的背景重构算法[J].软件学报,2005,16(9):1568-1576.]
[4]GAN Minggang,CHEN Jie,LIU Jin,et al.Moving objectdetection algorithm based on three-frame-differencing and edgeinformation[J].Journal of Electronics and Information Tech-nology,2010,32(4):894-897(in Chinese).[甘明刚,陈杰,刘劲,等.一种基于三帧差分和边缘信息的运动目标检测方法[J].电子与信息学报,2010,32(4):894-897.]
[5]Zhao H,Sun X J,Meng C H.Moving object de-tection basedon T-test combined with kirsch operator[C]//Wuhan,Chi-na:3rd International Conference on Image Analysis and SignalProcessing,2011:199-203.
[6]Tong N N,Duan X H.Research on detection algorithm for ve-hicle monitoring system[C]//Beijing:The First ChineseConference on Intelligent Surveillance,2002:612-616.
[7]Jagadish H P,Shambhavi D S.A novel digital algorithm forSobel edge detection[J].Communications in Computer and In-formation Science,2010,70:91-95
[8]Deng C X,Ma W F,Y Y.An edge detection app-roach ofimage fusion based on improved sobel operator[C]//Shang-hai:International Congress on Image and Signal Processing,2011:1189-1193.
[9]LIU Shaohua,ZHANG Miaojun,XIONG Zhihui,et al.Arobust and efficient video moving object detection and trackingalgorithm[J].Acta Automatica Sinica,2009,35(8):1055-1062(in Chinese).[刘少华,张茂军,熊志辉,等.一种鲁棒高效的视频运动目标检测与跟踪算法[J].自动化学报,2009,35(8):1055-1062.]
[10]HAO Haogang,CHEN Jiaqi.Moving object detection algo-rithm based on five frame difference and background difference[J].Computer Engineering,2012,38(4):146-148(inChinese).[郝毫刚,陈家琪.基于五帧差分和背景差分的运动目标检算法[J].计算机工程,2012,38(4):146-148.]
[11]Gao H,Xu W B,Sun J,et al.Multilevel thresholding forimage segmentation through an improved quantum-behavedparticle swarm algorithm[J].IEEE Transactions on Instru-mentation and Measurement,2010,59(4):934-946.
[12]Wang H Z,Dong Y.An improved image segmentation algo-rithm based on Otsu method[C]//International Symposiumon Photo-electronic Detection and Imaging:Related Technolo-gies and Applications,2008.
[13]HU Bin,GONG Ningsheng.An improved Otsu’thresholdingsegmentation method[J].Micro electronics&Computer,2009,26(12):153-155(in Chinese).[胡斌,宫宁生.一种改进的Otsu阈值分割[J].微电子学与计算机,2009,26(12):153-155.]
[14]XU Xiangyang,SONG Enmin,JIN Lianghai.Characteristicanalysis of threshold based on Otsu criterion[J].Acta Elec-tronica Sinica,2009,37(12):2716-2719(in Chinese).[徐向阳,宋恩民,金良海.Otsu准则的阈值性质分析[J].电子学报,2009,37(12):2716-2719.]
[2]WU Junqin,LIU Hao,LUO Yong.Moving object detectionalgorithm based on image sequences fusion[J].Computer En-gineering and Design,2012,33(12):4614-4618(in Chi-nese).[吴君钦,刘昊,罗勇.图像融合的运动目标检测算法研究[J].计算机工程与设计,2012,33(12):4614-4618.]
[3]HOU Zhiqiang,HAN Chongzhao.A background reconstruc-tion algorithm based on pixel intensity classification[J].Jour-nal of Software,2005,16(9):1568-1576(in Chinese).[侯志强,韩崇昭.基于像素灰度归类的背景重构算法[J].软件学报,2005,16(9):1568-1576.]
[4]GAN Minggang,CHEN Jie,LIU Jin,et al.Moving objectdetection algorithm based on three-frame-differencing and edgeinformation[J].Journal of Electronics and Information Tech-nology,2010,32(4):894-897(in Chinese).[甘明刚,陈杰,刘劲,等.一种基于三帧差分和边缘信息的运动目标检测方法[J].电子与信息学报,2010,32(4):894-897.]
[5]Zhao H,Sun X J,Meng C H.Moving object de-tection basedon T-test combined with kirsch operator[C]//Wuhan,Chi-na:3rd International Conference on Image Analysis and SignalProcessing,2011:199-203.
[6]Tong N N,Duan X H.Research on detection algorithm for ve-hicle monitoring system[C]//Beijing:The First ChineseConference on Intelligent Surveillance,2002:612-616.
[7]Jagadish H P,Shambhavi D S.A novel digital algorithm forSobel edge detection[J].Communications in Computer and In-formation Science,2010,70:91-95
[8]Deng C X,Ma W F,Y Y.An edge detection app-roach ofimage fusion based on improved sobel operator[C]//Shang-hai:International Congress on Image and Signal Processing,2011:1189-1193.
[9]LIU Shaohua,ZHANG Miaojun,XIONG Zhihui,et al.Arobust and efficient video moving object detection and trackingalgorithm[J].Acta Automatica Sinica,2009,35(8):1055-1062(in Chinese).[刘少华,张茂军,熊志辉,等.一种鲁棒高效的视频运动目标检测与跟踪算法[J].自动化学报,2009,35(8):1055-1062.]
[10]HAO Haogang,CHEN Jiaqi.Moving object detection algo-rithm based on five frame difference and background difference[J].Computer Engineering,2012,38(4):146-148(inChinese).[郝毫刚,陈家琪.基于五帧差分和背景差分的运动目标检算法[J].计算机工程,2012,38(4):146-148.]
[11]Gao H,Xu W B,Sun J,et al.Multilevel thresholding forimage segmentation through an improved quantum-behavedparticle swarm algorithm[J].IEEE Transactions on Instru-mentation and Measurement,2010,59(4):934-946.
[12]Wang H Z,Dong Y.An improved image segmentation algo-rithm based on Otsu method[C]//International Symposiumon Photo-electronic Detection and Imaging:Related Technolo-gies and Applications,2008.
[13]HU Bin,GONG Ningsheng.An improved Otsu’thresholdingsegmentation method[J].Micro electronics&Computer,2009,26(12):153-155(in Chinese).[胡斌,宫宁生.一种改进的Otsu阈值分割[J].微电子学与计算机,2009,26(12):153-155.]
[14]XU Xiangyang,SONG Enmin,JIN Lianghai.Characteristicanalysis of threshold based on Otsu criterion[J].Acta Elec-tronica Sinica,2009,37(12):2716-2719(in Chinese).[徐向阳,宋恩民,金良海.Otsu准则的阈值性质分析[J].电子学报,2009,37(12):2716-2719.]