An expression robust 3D face recognition method

In order to address the issue of the poor robustness of 3D face recognition against expression changes, we propose a method for expression-robust 3D face recognition. Firstly, the ridge points on the entire 3D face surface are extracted based on the curvature information of the surface. The ridge points exhibiting the most significant alterations in concavity and convexity are then employed to characterise the face as a faithful representation of the 3D face surface. Secondly, the face is divided into four regions based on an analysis of surface deformation caused by expression changes, and ridge point sets in each regions are characterised using the 3D shape context descriptors. The matching cost is in each regions calculated using a Chi-square test, calculating its matching cost by regions and fusing the matching cost of each regions at the decision level to produce the final recognition result. The experimental results obtained from the CASIA database demonstrate that the proposed method effectively utilises the ridge points on the 3D face surface, incorporates the spatial topology information of the 3D data, and considers the overall facial structure while attenuating the influence of expression changes through the weighted fusion of subregions. This approach achieves high recognition rate and robustness for faces with various expressions.