Many existing face recognition algorithms require manual intervention, and an imbalance exists between the size of the cropped image and the running speed.An automatic face recognition method based on self-organizing map ( SOM) was proposed to address these issues.First, AdaBoost face detection algorithm was used to obtain a face image, which was then converted into a grayscale image with the same size.Subsequently, the training subset of the image was used for training SOM classifier.Finally, the similarity measurement was used to complete the classification.The effectiveness of the proposed algorithm was verified by the results of the tests carried out for Li-bor Spacek dataset, extended Yale B dataset, and FERET dataset.Compared with the LDA, LDP, and WSR algo-rithm, the proposed method achieves fully automated processing.The recognition rates of the three test sets were all 97.00％.%针对当前许多人脸识别算法需要人工干预以及剪裁图像尺寸与运行速度不平衡的问题,提出一种利用自组织网络的自动人脸识别方法.利用AdaBoost人脸检测算法定位来获取人脸图像,并转化为相同尺寸的灰度图像;将图像的训练子集用于训练自组织映射(SOM)分类器,再利用相似性度量完成分类.在Libor Spacek数据集、扩展Yale B数据集和FERET数据集上进行验证,结果显示提出的方法实现了全自动处理,三个测试集的识别率均在97.00％左右,均优于对比算法.
展开▼
