基于机器视觉的局部遮挡人脸图像识别仿真

Simulation of Local Occlusion Face Image Recognition Based on Machine Vision

与无遮挡条件下的人脸识别不同,在局部遮挡下目标识别中需要考虑有效部位与遮挡部位的区别,准确提取目标区域,为此提出基于机器视觉的局部遮挡人脸图像识别方法。通过成对约束的半监督降维算法对人脸图像降维处理,获取图像中的显著特征,减少图像中的噪声。经过降维处理后,通过机器视觉技术提取局部遮挡下人脸图像的目标区域,确定目标区域的坐标位置。采用Criminisi修复算法获取未修复的人脸图像块数量,搜索最佳匹配块,同时对目标区域填充处理,完成人脸图像修复即可获取完整的人脸图像,最终实现人脸识别。实验结果表明,所提方法可以准确识别不同遮挡率下的人脸,同时可以减少识别时间,降低了遮挡问题对人脸识别效果的影响。

When identifying the target under partial occlusion,it is necessary to consider the difference between the effective part and the occluded part in order to accurately extract the target area.Therefore,this paper presented a method of identifying the face image with partial occlusion based on machine vision.Firstly,the semi-supervised algo-rithm based on pairwise constraint was adopted to reduce the dimensionality of the face image and obtain the salient features of the image,thus reducing the noise.After the dimensionality reduction,the target area of the face image un-der partial occlusion was extracted by machine vision technology,and then the coordinate position of the target area was determined.Moreover,the Criminisi repair algorithm was used to calculate the number of unrepaired face image blocks and search for the best matching block.Meanwhile,the target regions were flled.Finally,the face image repair was completed.Thus,a complete face image could be obtained.In other words,the face recognition was achieved.Ex-perimental results show that the proposed method can accurately identify faces under different shielding rates while re-ducing the recognition time and the impact of occlusion on face recognition effect.