摘要
针对普通相机无法获取人脸图片的深度信息而导致识别率低的问题,提出了一种基于点云和RGB图像的三维人脸增强识别方法。首先,利用深度相机采集点云数据和RGB图像,并利用Transformer的Attention机制建立图像的融合机制。其次,在FaceNet的MobileNet中增加encoder网络层,即PointTransformer,可以将数据分成多块计算。在encoder网络中,利用Attention机制将点云与RGB图像融合,以增强3D人脸识别的准确率。结果表明:基于点云和RGB图像的三维人脸增强识别方法的准确率达到99.67%,验证了此方法的可行性。
To solve the problem of low recognition rate due to the fact that the depth information of face images cannot be obtained by ordinary cameras,a 3D face enhancement recognition method based on point cloud and RGB images is proposed.Firstly,the depth camera is used to collect point cloud data and RGB images,and the attention mechanism of Transformer is utilized to establish the image fusion mechanism.Secondly,the encoder network layer,i.e.,PointTransformer,is added to the Mobilenet of Facenet,which can be realized to divide the data into multiple pieces for calculation.In the encoder network,the attention mechanism is utilized to fuse point cloud with RGB image to enhance the accuracy of 3D face recognition.The simulation results show that the accuracy of the 3D face enhancement recognition method based on point cloud and RGB image is 99.67%,which verifies the feasibility of this method.
作者
毛广志
丁彬
唐飞洋
刘聪聪
郁钱
MAO Guangzhi;DING Bin;TANG Feiyang;LIU Congcong;YU Qian(School of Computer Engineering,Jiangsu Institute of Technology,Changzhou 213001,China)
出处
《江苏理工学院学报》
2024年第4期85-89,共5页
Journal of Jiangsu University of Technology
基金
江苏省大学生创新创业大赛重点项目“基于点云和RGB图像的3D增强人脸识别”(202311463101Z)。
