针对不具备检修资质的人员误入检修现场和现场检修人员难以近距离查看设备铭牌的问题，研究了基于人工智能算法的作业现场人员设备资质识别技术。针对现场作业中的设备和人员分别采用不同的文本检测技术和人脸识别技术，展开设备、人员的资质识别技术研究。文本识别将融合连接文本提议网络（connectionist text proposal network，CTPN）、密集连接卷积网络（dense convolutional network，DenseNet）和连接时序分类（connectionist temporal classification，CTC）算法组合，并设计文本训练模型，提升文本识别率。人脸识别基于OpenCV平台，利用图像预处理和人脸目标算法实现人脸识别，从人脸数据采集、人脸数据训练、人脸检测与识别实现三个方面训练人脸识别的准确性。最后，通过实际应用证明了设备铭牌检测和人脸识别的准确性。
英文摘要:
Aiming at the problem that unqualified personnel enter the maintenance site by mistake and the on-site maintenance personnel are difficult to view the equipment nameplate at close range, the equipment qualification recognition technology of job site personnel based on artificial intelligence algorithm is studied in this paper. According to the equipment and personnel in the field operation, different text detection technology and face recognition technology are adopted respectively, and the qualification recognition technology of equipment and personnel is researched. Text recognition combines the combination of connectionist text proposal network (CTPN), dense convolutional network (DenseNet) and connectionist temporal classification (CTC) algorithms, and designs a text training model to improve text recognition rate. Face recognition is based on the OpenCV platform, and adopts image preprocessing and face target algorithms to achieve face recognition. It trains the accuracy of face recognition from three aspects of face data collection, face data training, and face detection and recognition. Finally, the accuracy of equipment nameplate detection and face recognition is proved through practical applications.