CAJ | 학술논문

针对三维人脸识别受表情变化和遮挡情况影响的问题，提出一种基于PDE 形变模型的三维人脸识别算法。首先采用自适应的PDE方法对三维人脸重建，并将PDE表示的三维人脸离散地表示成傅里叶级数的形式，用主成分分析对获得的几何残差进行训练学习，最后建立了结合人脸表情形变和遮挡形变的基于傅里叶级数表示的人脸形变模型，在该模型上获取映射信息进行人脸识别。实验结果表明，该模型对三维人脸形变模型具有良好的描述能力，在人脸表情变化和被遮挡时有更好的鲁棒性和识别率。
침대삼유인검식별수표정변화화차당정황영향적문제，제출일충기우PDE 형변모형적삼유인검식별산법。수선채용자괄응적PDE방법대삼유인검중건，병장PDE표시적삼유인검리산지표시성부리협급수적형식，용주성분분석대획득적궤하잔차진행훈련학습，최후건립료결합인검표정형변화차당형변적기우부리협급수표시적인검형변모형，재해모형상획취영사신식진행인검식별。실험결과표명，해모형대삼유인검형변모형구유량호적묘술능력，재인검표정변화화피차당시유경호적로봉성화식별솔。
According to the problem that expression changes and occlusion affected the three-dimensional face recognition, this paper proposed a three-dimensional face recognition algorithm based on deformable model PDE(partial differential equa-tions).First,it used adaptive PDE method to reconstruct the three-dimensional facial,and represented three-dimensional face with a PDE’s in the form of a Fourier series.Then,it trained geometric residuals by principal component analysis.Finally,it established facial deformation model based on Fourier series,which included facial expression deformation and occlusion de-formation.Results show that the algorithm model has good description ability in many applications.