智能權限辨識軟體的開發

Development of Intelligent Authority Identification software

李大維(DA-WEI Li)

指導教授 ： 鄭文昌

朝陽科技大學/資訊學院/資訊工程系/碩士(2019年)

摘要

近幾年來，工業和資訊科技領域發生了重大的變化，進入了工業4.0的時代。製造業將通過互聯網，實現內部和外部網絡的整合，朝著智能化方向發展，因此智慧機械會是工具機產業未來發展的重點。本研究起源於產學合作案，我們在台灣三菱電機「PC軟體NC操作」的獨特架構下，開發出一套智能權限辨識軟體。過往NC工具機上的權限管控機制大多透過鑰匙進行管控，為改善權限管控機制，本APP除了一般的帳號密碼登入，也加入了人臉辨識登入功能，並能根據使用者的權限自動變更NC工具機的操作權限。人臉辨識的部分我們使用FaceNet深度卷積類神經網路作為人臉影像特徵擷取的方法，並直接計算特徵向量的距離當作人臉相似性的度量，因此當新增一個新的使用者時，則不需要重新訓練整個卷積類神經網路。由於距離計算有高誤報率，因此我們提出特徵正規化距離比對法，經實驗驗證此方法確實能有效降低誤報率，我們提出的人臉登入軟體的準確度最好可達98.17%以上，具有實用目標。此軟體在完成後於第27屆台北國際工具機展中配合廠商參展。

關鍵字

智慧機械 ； 權限辨識 ； 人臉辨識 ； 特徵比對

並列摘要

In recent years, major changes have taken place in the industrial and information technology fields, entering the era of Industry 4.0. The manufacturing industry will realize the integration of internal and external networks through the Internet and develop towards the direction of intelligence. Therefore, the smart machinery will be the focus of the future development of the machine tool industry. This research originated from the industry-university cooperation case. We developed an intelligent Authority Identification Application under the unique architecture of Mitsubishi Electric "PC Software NC Operation". In the past, we mostly control the NC machine tool through the key. In order to improve this situation, the software can login not only through account password but also through face recognition, and can automatically change the operation authority of the NC machine tool according to the user's authority. In face recognition, we use FaceNet as the method of facial image feature extraction, and directly calculate the distance of feature vector as the measure of face similarity, so when we add a new user, there is no need to retrain the entire neural network. Above method has a high false positive rate, we use feature normalization distance comparison method. The experiment results show that our method can effectively reduce the false positive rate, the accuracy of our proposed face login software can reach over 98.17%, which means that it has practical goals. After the software is finished, it was exhibited at the 27th Taipei Int'l Machine Tool Show in 2019.

並列關鍵字

smart machinery ； authority identification ； face recognition ； feature comparison

參考文獻

[1] Shaoqing Ren, Kaiming He, Ross Girshick and Jian Sun, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume: 39, Issue: 6, June 1 2017, pp.1137- 1149.

[2] Wen-Chang Cheng, Ting-Yu Wu and Dai-Wei Li, “Ensemble Convolutional Neural Networks for Face Recognition,” in Proceedings of 2018 International Conference on Algorithms, Computing and Artificial Intelligence (ACAI’18), Dec. 21-23, 2018, Sanya, China.

[3] Florian Schroff, Dmitry Kalenichenko and James Philbin, “FaceNet: A Unified Embedding for Face Recognition and Clustering,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’15), June 8-10, 2015, Boston, USA.

[4] Viola and Jones, “Rapid object detection using a boosted cascade of simple features,” in Proceedings of the 2001 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’01), 8-14 Dec. 2001, Kauai, HI, USA.

[5] OpenCV, https://opencv.org/.