刚性区域特征点的3维人脸识别

发布时间：2018-07-13 20:21

【摘要】：目的针对3维人脸识别中存在表情变化的问题,提出了一种基于刚性区域特征点的3维人脸识别方法。方法该方法首先在人脸纹理图像上提取人脸图像的特征点,并删除非刚性区域内的特征点,然后根据采样点的序号,在人脸空间几何信息上得到人脸图像特征点的3维几何信息,并建立以特征点为中心的刚性区域内的子区域,最后以子区域为局部特征进行人脸识别测试,得到不同子区域对人脸识别的贡献,并以此作为依据对人脸识别的结果进行加权统计。结果在FRGC v2.0的3维人脸数据库上进行实验测试,该方法的识别准确率为98.5%,当错误接受率(FAR)为0.001时的验证率为99.2%,结果表明,该方法对非中性表情下的3维人脸识别具有很好的准确性。结论该方法可以有效克服表情变化对3维人脸识别的影响,同时对3维数据中存在的空洞和尖锐噪声等因素具有较好的鲁棒性,对提高3维人脸识别性能具有重要意义。
[Abstract]:Aim to solve the problem of facial expression change in 3D face recognition, a 3D facial recognition method based on rigid region feature points is proposed. Methods the feature points of the face image are extracted from the texture image, and the feature points in the non-rigid region are deleted, and then, according to the ordinal number of the sampling points, the feature points of the face image are removed. The 3D geometric information of facial image feature points is obtained from the geometric information of human face space, and the subregions of rigid regions centered on feature points are established. Finally, the sub-regions are used as local features for face recognition testing. The contribution of different sub-regions to face recognition is obtained, and based on the results of face recognition, weighted statistics are carried out. Results the experiment was carried out on the 3D face database of FRGC v2.0. The recognition accuracy of the method was 98.5 and the verification rate was 99.2 when the error acceptance rate (far) was 0.001. This method has good accuracy for 3 D face recognition under non neutral expression. Conclusion this method can effectively overcome the effect of facial expression change on 3D face recognition, and it has good robustness to voids and sharp noises in 3D data, and it is of great significance to improve the performance of 3D face recognition.
【作者单位】：辽宁工程技术大学工商管理学院;辽宁工程技术大学软件学院;
【基金】：国家自然科学基金项目(61172144) 辽宁省科技攻关计划项目(2012216026) 辽宁省教育厅科学研究项目(LJYL-049)~~
【分类号】：TP391.41

【相似文献】