保角特征结合改进差分进化算法的三维人脸识别
针对三维人脸识别的高复杂度和二维人脸识别无法提供粒状线索的问题,提出一种全自动3D人脸表情识别算法,该算法主要是提供比2D人脸识别更多的线索,同时降低计算复杂度。通过保角映射将3D人脸转换到2D平面,保留了面部变化的线索,提出了基于优化算法的差分进化（DE）算法用于提高识别效率,同时提取最优人脸特征集和分类器参数,加速鲁棒特征池描述了所有预期的人脸特征点。在博斯普鲁斯、FRGC v2及笔者搜集的人脸数据集上的实验结果表明,算法解决了三维人脸识别的高计算复杂度和二维人脸识别的线索低的问题,并在不降低识别性能的前提下大大地节约了成本,相比几种较为先进的三维人脸识别算法,算法取得了更好的识别效果,有...
Saved in:
| Published in | 计算机应用研究 Vol. 33; no. 6; pp. 1898 - 1902 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
四川工程职业技术学院,四川 德阳,618000%四川大学,成都,610065
2016
|
| Subjects | |
| Online Access | Get full text |
| ISSN | 1001-3695 |
| DOI | 10.3969/j.issn.1001-3695.2016.06.066 |
Cover
| Summary: | 针对三维人脸识别的高复杂度和二维人脸识别无法提供粒状线索的问题,提出一种全自动3D人脸表情识别算法,该算法主要是提供比2D人脸识别更多的线索,同时降低计算复杂度。通过保角映射将3D人脸转换到2D平面,保留了面部变化的线索,提出了基于优化算法的差分进化（DE）算法用于提高识别效率,同时提取最优人脸特征集和分类器参数,加速鲁棒特征池描述了所有预期的人脸特征点。在博斯普鲁斯、FRGC v2及笔者搜集的人脸数据集上的实验结果表明,算法解决了三维人脸识别的高计算复杂度和二维人脸识别的线索低的问题,并在不降低识别性能的前提下大大地节约了成本,相比几种较为先进的三维人脸识别算法,算法取得了更好的识别效果,有望应用于一些商业人脸识别系统。 |
|---|---|
| Bibliography: | 51-1196/TP As the problem of the high complexity of 3D face recognition and 2D face recognition not providing granular clues,this paper proposed a fully automatic 3D facial expression recognition algorithm. It provided more clues than that of 2D face recognition and reduced the computational complexity at the same time. Firstly,it transformed 3D face into a 2D plane by conformal mapping,retaining the changing of facial clues. Secondly,it proposed an optimization algorithm based on differential evolution（ DE） algorithm to improve the recognition efficiency,while extracting the best facial feature set and classification parameters,and speed up robust features（ SURF） described all the expected facial feature points. Experimental results on the data sets of Bosphorus,FRGC v2 and gathered face data sets show that the proposed algorithm solves high computational complexity of 3D face recognition and low clues of 2D face recognition. This algorithm greatly reduces the cost without lowering the recognition performance |
| ISSN: | 1001-3695 |
| DOI: | 10.3969/j.issn.1001-3695.2016.06.066 |