改进的统计模型三维人脸特征点标定算法框架
|
摘要
自动三维人脸特征点标定是计算机视觉领域的研究热点,其广泛应用于人脸识别,人脸模型配准,表情识别,脸部动画等领域。通过对三维人脸样本统计建模,采用遗传算法对待匹配模型的生成数目进行参数优化,利用模型相似性匹配方法及其映射关系对三维人脸特征点进行自动标定。首先,对三维人脸数据预处理,然后对其统计建模并通过模型形变得到有映射关系的基准模型和待匹配模型。利用遗传算法对待匹配模型中的待匹配模型生成数目参数进行优化,生成与之对应的待匹配模型数;接着计算待测模型与待匹配模型的相似度。最后,利用模型相似度和模型映射关系,间接得到待测模型的特征点。实验结果表明,提出的算法是可行的,能够在一定程度上提高原有算法的效率。该算法可以自动标定三维人脸模型的特征点,当距离阈值为10像素时,39个三维人脸特征点定位的准确率都可以达到100%,并有效解决了传统方法中三维人脸模型平滑区域特征点精度不高的问题。
Automatic 3D facial feature point extraction is a hot field of computer vision, which is widely used in face recognition, face model alignment, facial expression recognition, facial animation and other areas. By means of the statistical modeling of 3D face samples and the parameter optimization challenge in the number of for-matched models is solved by Genetic Algorithm(GA). The 3D facial feature points can be calibrated automatically by using the model similarity matching method and its mapping relation. Firstly, the statistic models of 3D face model are constructed. Then get the mapping relation of the reference model and the matching model by the model deformation and then using the parameter about the number of for-matched model by Genetic Algorithm(GA)to generate the corresponding matching model. Secondly,calculate the similarity between the test model and matched model. Finally, the feature points of the model to be measured are indirectly obtained with the model similarity and the projection relationship. The experimental results show that the proposed algorithm is feasible and very effective. The method can automatically extract 3D face feature points. When the threshold value of the distance is 10 pixels, 39 of the 3D face feature points localization accuracy in this article can reach100%. At the same time it can also solve the problem of traditional method of 3D face model smooth area feature point with low precision effectively.
Automatic 3D facial feature point extraction is a hot field of computer vision, which is widely used in face recognition, face model alignment, facial expression recognition, facial animation and other areas. By means of the statistical modeling of 3D face samples and the parameter optimization challenge in the number of for-matched models is solved by Genetic Algorithm(GA). The 3D facial feature points can be calibrated automatically by using the model similarity matching method and its mapping relation. Firstly, the statistic models of 3D face model are constructed. Then get the mapping relation of the reference model and the matching model by the model deformation and then using the parameter about the number of for-matched model by Genetic Algorithm(GA)to generate the corresponding matching model. Secondly,calculate the similarity between the test model and matched model. Finally, the feature points of the model to be measured are indirectly obtained with the model similarity and the projection relationship. The experimental results show that the proposed algorithm is feasible and very effective. The method can automatically extract 3D face feature points. When the threshold value of the distance is 10 pixels, 39 of the 3D face feature points localization accuracy in this article can reach100%. At the same time it can also solve the problem of traditional method of 3D face model smooth area feature point with low precision effectively.
引文
[1]Xu C H,Tan T N,Li S,et al.Learning effective intrinsic features to boost 3D-based face recognition[C]//ECCV2006:Proc of the 9th European Conference on Computer Vision,Graz,Austria,2006.
[2]Conde C,Cipolla R,Rodrguez-Aragcn L J.3D facial feature location with spin images[C]//IAPR:Conference on Machine Vision Applications,Tsukuba Science City,Japan,2005.
[3]Conde C,Serrano A.3D facial normalization with spin images and influence of range data calculation over face verification[C]//CVRP 2005:Computer Vision and Pattern Recognition,San Diego,2005.
[4]Lu X G,Colbry D,Jain A.Three-dimensional model based face recognition[C]//ICPR 2004:Proceedings of the 17th International Conference on Pattern Recognition,2004:362-366.
[5]张广鹏,张艳宁,郭哲.基于精确主轴分析及ICP的三维人脸配准[J].计算机工程与应用,2006,42(29):62-64.
[6]Feng Q,Pan J S,Yan L,et al.Adaptively nearest feature point classifier for face recognition[M]//Innovations in Bioinspired Computing and Applications.[S.l.]:Springer,2014:115-122.
[7]李康,尚鹏,耿国华.基于深度图像的人脸模型特征点自动标定[J].计算机科学,2014,41(4):287-291.
[8]Cosar S,Cetin M.A graphical based solution to the facial feature point tracking problem[J].Image and Vision Computing,2011,29(5).
[9]Gizatdinova Y,Surakka V.Automatic edge-based localization of facial features from images with complex facial expression[J].Pattern Recognition Letters,2010,31(15):2436-2446.
[10]Li S,Wang H.Realistic 3D face modeling and application research[J].International Journal of Electronic Commerce Studies,2014,4(2):203-212.
[11]王成章.基于三维模型的人脸识别算法的研究[D].北京:北京工业大学,2008.
[12]王跃明,潘纲,吴朝晖.三维人脸识别研究综述[J].计算机辅助设计与图形学学报,2008,20(7):819-829.
[13]Adam D,Tibbalds.Three dimensional human face acquisitions for recognition[D].Cambridge:University of Cambridge,1998.
[14]张雨禾,耿国华,刘伦椿,等.基于L邻域分割的结构性纹理合成方法[J].哈尔滨工程大学学报,2014,35(10):1265-1271.
[15]Zhao W Y.Robust image based 3D face recognition[D].University of Maryland,1999.
[16]王旭东,任炳俐,孙尚迪.基于二维PCA算法的三维人脸识别[J].电子制作,2013(12).
[17]冯筠,陈雨,仝鑫龙,等.三维颅骨特征点的自动标定[J].光学精密工程,2014,22(5):1388-1394.
[18]王劭寅.基于区域统计模型的正颌手术仿真关键算法研究[D].西安:西北大学,2012.
[19]Cootes T F,Taylor C J,Cooper D H.Active shape models-their training and applications[J].Computer Vision and Image Understanding,1995,61(1):38-59.
[20]Holland J H.Adaptation in natural and artificial systems[M].Ann Arbor:The University of Michigan Press,1975.
[21]麻宏静,张德同,冯筠.基于相对角分布聚类和支持向量机的3维人脸特征点匹配技术的研究[J].中国图象图形学报,2011,16(5):850-856.
[22]麻宏静.基于相对角聚类和支持向量机的人脸特征点定位技术研究[D].西安:西北大学,2010.
[2]Conde C,Cipolla R,Rodrguez-Aragcn L J.3D facial feature location with spin images[C]//IAPR:Conference on Machine Vision Applications,Tsukuba Science City,Japan,2005.
[3]Conde C,Serrano A.3D facial normalization with spin images and influence of range data calculation over face verification[C]//CVRP 2005:Computer Vision and Pattern Recognition,San Diego,2005.
[4]Lu X G,Colbry D,Jain A.Three-dimensional model based face recognition[C]//ICPR 2004:Proceedings of the 17th International Conference on Pattern Recognition,2004:362-366.
[5]张广鹏,张艳宁,郭哲.基于精确主轴分析及ICP的三维人脸配准[J].计算机工程与应用,2006,42(29):62-64.
[6]Feng Q,Pan J S,Yan L,et al.Adaptively nearest feature point classifier for face recognition[M]//Innovations in Bioinspired Computing and Applications.[S.l.]:Springer,2014:115-122.
[7]李康,尚鹏,耿国华.基于深度图像的人脸模型特征点自动标定[J].计算机科学,2014,41(4):287-291.
[8]Cosar S,Cetin M.A graphical based solution to the facial feature point tracking problem[J].Image and Vision Computing,2011,29(5).
[9]Gizatdinova Y,Surakka V.Automatic edge-based localization of facial features from images with complex facial expression[J].Pattern Recognition Letters,2010,31(15):2436-2446.
[10]Li S,Wang H.Realistic 3D face modeling and application research[J].International Journal of Electronic Commerce Studies,2014,4(2):203-212.
[11]王成章.基于三维模型的人脸识别算法的研究[D].北京:北京工业大学,2008.
[12]王跃明,潘纲,吴朝晖.三维人脸识别研究综述[J].计算机辅助设计与图形学学报,2008,20(7):819-829.
[13]Adam D,Tibbalds.Three dimensional human face acquisitions for recognition[D].Cambridge:University of Cambridge,1998.
[14]张雨禾,耿国华,刘伦椿,等.基于L邻域分割的结构性纹理合成方法[J].哈尔滨工程大学学报,2014,35(10):1265-1271.
[15]Zhao W Y.Robust image based 3D face recognition[D].University of Maryland,1999.
[16]王旭东,任炳俐,孙尚迪.基于二维PCA算法的三维人脸识别[J].电子制作,2013(12).
[17]冯筠,陈雨,仝鑫龙,等.三维颅骨特征点的自动标定[J].光学精密工程,2014,22(5):1388-1394.
[18]王劭寅.基于区域统计模型的正颌手术仿真关键算法研究[D].西安:西北大学,2012.
[19]Cootes T F,Taylor C J,Cooper D H.Active shape models-their training and applications[J].Computer Vision and Image Understanding,1995,61(1):38-59.
[20]Holland J H.Adaptation in natural and artificial systems[M].Ann Arbor:The University of Michigan Press,1975.
[21]麻宏静,张德同,冯筠.基于相对角分布聚类和支持向量机的3维人脸特征点匹配技术的研究[J].中国图象图形学报,2011,16(5):850-856.
[22]麻宏静.基于相对角聚类和支持向量机的人脸特征点定位技术研究[D].西安:西北大学,2010.