基于高阶门控卷积的面部识别算法在电站身份识别中的应用

Application of Facial Recognition Algorithm Based on High-order Gated Convolution for Power Station Identification

电站安全与国家安全和人民生产生活息息相关,因此对电站的出入人员进行准确的身份识别尤为重要。针对电站出入人员携带安全帽导致的面部特征采集不完全等问题,文章提出了融合高阶门控卷积模块对MTCNN和Face Net网络模型进行改进增强,通过显式建模人脸面部特征向量之间的高阶关联,使得模型可以学习到更加准确的特征映射函数,从而提高面部识别的准确率。在LFW数据集上,提出的改进方法相比于原Facenet模型的99.63%的准确率提高到了99.68%的准确率。最后,在电站的实际应用场景中,出入识别的准确率分别为99.12%和98.93%。

The security of the power station is closely related to national security and people's production and life,so it is particularly important to accurately identify the access personnel of the power station.Aiming at the problems such as incomplete collection of facial features caused by helmets carried by power station access personnel,this paper proposes to improve and enhance the MTCNN and FaceNet network models by integrating high-order gated convolution modules.By explicitly modeling the high-order correlation between facial feature vectors,the model can learn more accurate feature mapping functions to improve the accuracy of facial identification.On the LFW dataset,the improved method proposed in this paper improves the accuracy of the original FaceNet model by 99.63% to 99.68%.Finally,in the actual application scenario of the power station,the accuracy of the access identification is 99.12%and 98.93% respectively.