基于MATLAB的动态前景目标特征提取与运动跟踪
|
摘要
动态前景目标识别和提取是计算机视觉领域的重要内容。对动态图像进行前景目标提取与运动跟踪,通过改进高斯混合背景模型,提出一种基于自适应特征加权的前景目标提取算法,目的是对动态画面中的图像特征进行识别并精确提取所需要的画面。根据高斯模型组合多个图像特征,针对组合特征空间的各个子空间构建似然图像,通过似然图像特征分析与加权,提取最具有差异性的特征。根据图像前景特征在图像帧与帧之间的不同,提高前景目标跟踪的鲁棒性。试验结果表明,改进算法在提取前景目标上比传统算法提高了精度,目标跟踪效果好。
The identification and extraction of dynamic foreground targets is an important topic in the field of computer vision.The research content of this subject is to extract foreground image and motion tracking for moving images.By improving the Gaussian mixture background model,a foreground target extraction algorithm based on adaptive feature weighting is proposed.The purpose is to identify the image features in the dynamic picture and accurately extract the desired foreground image.According to the Gaussian model,a plurality of image features are combined,and a likelihood image is constructed for each subspace of the combined feature space,and the most different features are extracted by likelihood image feature analysis and weighting.The image foreground feature is used to improve the robustness of foreground target tracking.The experimental results show that the improved algorithm improves the accuracy of the foreground target compared with the traditional algorithm,and the target tracking effect is obvious,which verifies the feasibility of the method.
The identification and extraction of dynamic foreground targets is an important topic in the field of computer vision.The research content of this subject is to extract foreground image and motion tracking for moving images.By improving the Gaussian mixture background model,a foreground target extraction algorithm based on adaptive feature weighting is proposed.The purpose is to identify the image features in the dynamic picture and accurately extract the desired foreground image.According to the Gaussian model,a plurality of image features are combined,and a likelihood image is constructed for each subspace of the combined feature space,and the most different features are extracted by likelihood image feature analysis and weighting.The image foreground feature is used to improve the robustness of foreground target tracking.The experimental results show that the improved algorithm improves the accuracy of the foreground target compared with the traditional algorithm,and the target tracking effect is obvious,which verifies the feasibility of the method.
引文
[1]刘洁.基于光流法的运动目标检测和跟踪算法研充[D].徐州:中国矿业大学,2015.
[2]夏瑜.图像处理技术研究[D].无锡:江南大学,2015.
[3]严晓明.一种基于改进帧差法的运动目标检测[J].莆田学院学报,2011(5):69-72.
[4]李键,丁学明.视频前景提取算法研究[J].软件导刊,2018(2):135-139.
[5]刘国宏,郭文明.改进的中值滤波算法[J].计算机光盘软件与应用,2012(6):122-123.
[6]刘嘉敏,彭玲,袁佳成.基于二维变分模态分解和自适应中值滤波的图像去噪方法[J].计算机应用研究,2017,34(5):1700-1705.
[7]江登表,程小辉.基于高斯运动模型的车辆检测[J].电视技术,2009,33(8):103-106.
[8]张学习,杨宜民.彩色图像工程中常见的颜色空间及其转换[J].计算机工程与设计,2008,29(5):1210-1212.
[9]黄玉.基于改进混合高斯模型的前景检测算法研究[D].长沙:长沙理工大学,2014.
[10]于明,郭团团.基于高斯混合模型的人群异常检测[J].软件导刊,2017(11):59-62.
[11]李鸿.基于混合高斯模型的运动检测及阴影消除算法研究[D].天津:中国民航大学,2013.
[12]刘令芝.基于视频的车辆检测与跟踪研究[D].湘潭:湘潭大学,2011.
[13]HEIKKILA M,PIETIKAINEN M,HEIKKILA J.A texture-based method for detecting moving objects[EB/OL].http://www.docin.com/p-1395391692.html.
[14]JIAN M,JUNG C.Interactive image segmentation using adaptive constraint propagation[J].IEEE Transactions on Image Process,2016,25(3):1301-1311.
[15]李培华.序列图像中运动目标跟踪方法[M].北京:科学出版社,2010:32-35.
[16]李霉,刘昌化,年福忠.基于自适应背景的多特征融合目标跟踪[J].计算机应用,2013(3):891-902.
[17]胡小锋,赵辉.Visual C++/MATLAB图像处理与识别实用案例精选[M].北京:人民邮电出版社,2004.
[18]GHASSABEH Y A.Asymptotic stability of equilibrium points of mean shift algorithm[J].Machine Learning,2015,98(3):359-368.
[19]王强光,伍铁军.智能监控系统中的运动目标分类方法研究[J].工业控制计算机,2011,24(1):16-20.
[20]黄超群.基于混合高斯模型和Kalman滤波器的运动目标检测与跟踪[D].昆明:云南大学,2010.
[2]夏瑜.图像处理技术研究[D].无锡:江南大学,2015.
[3]严晓明.一种基于改进帧差法的运动目标检测[J].莆田学院学报,2011(5):69-72.
[4]李键,丁学明.视频前景提取算法研究[J].软件导刊,2018(2):135-139.
[5]刘国宏,郭文明.改进的中值滤波算法[J].计算机光盘软件与应用,2012(6):122-123.
[6]刘嘉敏,彭玲,袁佳成.基于二维变分模态分解和自适应中值滤波的图像去噪方法[J].计算机应用研究,2017,34(5):1700-1705.
[7]江登表,程小辉.基于高斯运动模型的车辆检测[J].电视技术,2009,33(8):103-106.
[8]张学习,杨宜民.彩色图像工程中常见的颜色空间及其转换[J].计算机工程与设计,2008,29(5):1210-1212.
[9]黄玉.基于改进混合高斯模型的前景检测算法研究[D].长沙:长沙理工大学,2014.
[10]于明,郭团团.基于高斯混合模型的人群异常检测[J].软件导刊,2017(11):59-62.
[11]李鸿.基于混合高斯模型的运动检测及阴影消除算法研究[D].天津:中国民航大学,2013.
[12]刘令芝.基于视频的车辆检测与跟踪研究[D].湘潭:湘潭大学,2011.
[13]HEIKKILA M,PIETIKAINEN M,HEIKKILA J.A texture-based method for detecting moving objects[EB/OL].http://www.docin.com/p-1395391692.html.
[14]JIAN M,JUNG C.Interactive image segmentation using adaptive constraint propagation[J].IEEE Transactions on Image Process,2016,25(3):1301-1311.
[15]李培华.序列图像中运动目标跟踪方法[M].北京:科学出版社,2010:32-35.
[16]李霉,刘昌化,年福忠.基于自适应背景的多特征融合目标跟踪[J].计算机应用,2013(3):891-902.
[17]胡小锋,赵辉.Visual C++/MATLAB图像处理与识别实用案例精选[M].北京:人民邮电出版社,2004.
[18]GHASSABEH Y A.Asymptotic stability of equilibrium points of mean shift algorithm[J].Machine Learning,2015,98(3):359-368.
[19]王强光,伍铁军.智能监控系统中的运动目标分类方法研究[J].工业控制计算机,2011,24(1):16-20.
[20]黄超群.基于混合高斯模型和Kalman滤波器的运动目标检测与跟踪[D].昆明:云南大学,2010.