臺灣博碩士論文加值系統

在全球疫情升溫的情況下，人們紛紛戴上了口罩，而當人們使用FACE ID解鎖手機時，卻發現無法解鎖，拿下口罩也有防疫安全的疑慮，因為這個問題，臉部辨識(face_recognition)再次成為計算機領域中一個重要的議題，原因在於，戴上口罩的人臉無法做出有效的身分辨識，由於口罩無法作為臉部辨識的配件，以及人臉大多數特徵被口罩的遮擋，鼻子、嘴巴、臉型輪廓等等，造成特徵點的減少，而被口罩遮擋的這些特徵在人臉辨識中，是重要的臉部辨識關鍵點，致使無法完成臉部辨識。

由於臉部辨識的特徵大量減少，加上辨識困難等原因，因此本論文探究臉部辨識無法實現戴口罩完成人臉辨識的原因。本研究提出的方法為，針對人臉眼睛特徵部分做人臉相似度辨識。

因此本論文利用opencv中的haar cascade將戴口罩的人臉圖片，眼睛周圍的特徵擷取後，經過opencv的特徵向量(eigenvector)轉化，並比對其特徵向量，最後利用eigen_face的比較方式來得到相似度，完成戴口罩也能做到臉部辨識的效果。

本研究最後將得到的人臉相似度設定在80%，與open face data的人臉進行向量相似度比較後，得到準確度約為87%。達到透過向量計算能夠完成戴口罩的人臉辨識。

When the global epidemic heats up, people are wearing masks one after another, but when people use FACE ID to unlock their mobile phones, they find that they cannot be unlocked. Taking off the mask also has concerns about epidemic prevention and safety. Because of this problem, face recognition once again It has become an important topic in the computer field, because the face wearing a mask cannot make effective identification, because the mask cannot be used as an accessory for face recognition, and most of the features of the face are blocked by the mask, nose, mouth , face contours, etc., resulting in the reduction of feature points, and these features blocked by masks are important face recognition key points in face recognition, making it impossible to complete face recognition.

Since the features of face recognition are greatly reduced and the recognition is difficult, this paper discusses the reasons why face recognition cannot achieve face recognition with masks. The method proposed in this study is to identify the face similarity between the feature parts of the face and the eyes.

Therefore, this paper uses the haar cascade in opencv to extract the features around the face and eyes of people wearing masks, transforms them through the feature vector of opencv, compares the feature vectors, and finally uses the method of comparing eigen_face to get the similarity, even if Wearing a mask and wearing a mask can also achieve the effect of face recognition.

In this study, the obtained face similarity is set to 80%. After comparing the vector similarity with open face data, the accuracy is about 87%. It can realize face recognition with masks through vector calculation.

目 錄
摘要…………………………………………………………………………………………………………………… i
ABSTRACT………………………………………………………………………………………………………….. ii
致謝…………………………………………………………………………………………………………………… iii
目錄…………………………………………………………………………………………………………………… iv
圖目錄……………………………………………………………………………………………………………….. vi
表目錄……………………………………………………………………………………………………………….. vii
壹、緒論………………………………………………………………………………………………………….… 1
一、研究背景…………………………………….………………………………………………………. 1
二、研究動機與目的…………………………………………………………………………………. 2
三、研究範圍與限制…………………………………………………………………………………. 4
四、研究架構……………………………………………………………………………………………. 5
五、研究貢獻…………………………………………………………………………………………….. 8
貳、文獻探討……………………………………………………………………………………………………. 9
一、FACE_DETECTION………………………………………………………………………………… 9
二、FACE_RECOGNITION…………………………………………………………………………….10
三、FEATURE_EXTRACTION…………………………………………………………………………14
四、EIGEN_FACE………………………………………………………………………………………….15
五、EIGEN_VECTOR…………………………………………………………………………………….16
六、EUCLIDEAN_DISTANCE………………………………………………………………………….17
七、結語……………………………………………………………………………………………………. 17
參、研究方法與設計………………………………………………………………………………………….18
一、實驗設計流程概述…………………………………………………………………………….. 18
二、臉部特徵定位………………………………………………………………………………………21
三、特徵轉換…………………………………………………………………………………………… 24
 
肆、結果與討論………………………………………………………………………………………………….26
一、初步實驗結果…………………………………………………………………………………….. 26
二、主要實驗……..…………………………………………………………………………………….. 27
三、實驗結果與討論…………………………………………………………………………………. 35
四、比較結果與討論………………………………………………………………………………… 36
伍、結論與未來展望…..……………………………………………………………………………………. 37
一、圖片大小的限制………………………………………………………………………………… 37
二、相似度的設定………………………………………………………………………………………37
三、相似度不佳解決方案………………………………………………………………………… 37
陸、參考文獻………………………………………………………………………………………………….. 38

 
圖目錄
圖1-1:未戴口罩能正常人臉辨識圖………………………………………………………………………. 3
圖1-2: 戴口罩後無法正常進行人臉偵測圖………………………….………………………………. 3
圖1-3:系統流程圖……………………………………………………………………………………………….. 6
圖2-1: 三元組損失示意圖…………………………………………………………………………………… 13
圖3-1: 本研究使用之戴口罩人臉圖 ……………………………………………………………..…… 18
圖3-2: 眼睛特徵定位後結果圖………………………………………………………………..………… 19
圖3-3：經過矩陣化結果圖…….………………………………………………………………….……… 20
圖3-4：特徵向量結果圖………………………………………………………………………….………….. 21
圖3-5: 部分人臉公開資料庫圖………………………………………………………….………………. 22
圖3-6：輸入戴口罩人臉圖…………………………………………………………………………………. 24
圖3-7：進行特徵定位程式結果圖………………………………………………………………………. 25
圖3-8：經特徵定位裁切後的眼睛位置圖…………………………………………………………… 25
圖3-9：特徵定位經過矩陣化結果圖…………………………………………………………………… 26
圖3-10：部分特徵向量圖…………………………………………………………………………………… 27
圖3-11：歐氏距離結果圖…………………………………………………………………………………… 27
圖4-1：主要實驗架構圖………………………………………………………………………………….….. 30
圖4-2：隨機抽樣與準確折線圖………………………………………………………………………….. 31

 
表目錄
表4-1:初步實驗相似度表………………………………………………………………………………………28
表4-2:進行特徵辨識比較表………………………………………………………………………………... 32

[1]Kumar, A., Kaur, A., & Kumar, M. (2019). Face detection techniques: a review. Artificial Intelligence Review, 52(2), 927-948.

[2]Viola, P., & Jones, M. J. (2004). Robust real-time face detection. International journal of computer vision, 57(2), 137-154.

[3] Rowley, H. A., Baluja, S., & Kanade, T. (1998). Neural network-based face detection. IEEE Transactions on pattern analysis and machine intelligence, 20(1), 23-38.

[4] Yang, S., Luo, P., Loy, C. C., & Tang, X. (2016). Wider face: A face detection benchmark. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 5525-5533).

[5] Hsu, R. L., Abdel-Mottaleb, M., & Jain, A. K. (2002). Face detection in color images. IEEE transactions on pattern analysis and machine intelligence, 24(5), 696-706.

[6] 李一二, 人臉識別學習筆記一：入門篇,https://iter01.com/515955.html,2022

[7]ageitgey,face_recognition,https://github.com/ageitgey/face_recognition,2022

[8] Zhao, W., Chellappa, R., Phillips, P. J., & Rosenfeld, A. (2003). Face recognition: A literature survey. ACM computing surveys (CSUR), 35(4), 399-458.

[9] Tolba, A. S., El-Baz, A. H., & El-Harby, A. A. (2006). Face recognition: A literature review. International Journal of Signal Processing, 2(2), 88-103.

[10] Damer, N., Grebe, J. H., Chen, C., Boutros, F., Kirchbuchner, F., & Kuijper, A. (2020, September). The effect of wearing a mask on face recognition performance: an exploratory study. In 2020 International Conference of the Biometrics Special Interest Group (BIOSIG) (pp. 1-6). IEEE.
 
[11] Hermans, A., Beyer, L., & Leibe, B. (2017). In defense of the triplet loss for person re-identification. arXiv preprint arXiv:1703.07737.

[12] 文杰,三元組損失概述,https://skyfaker.cc/2019/08/22/san-yuan-zu-sun-shi-mo-xing/,(2022)

[13] Wang, H., Hu, J., & Deng, W. (2017). Face feature extraction: a complete review. IEEE Access, 6, 6001-6039.

[14] 張靜遠, 張冰, & 蔣興舟. (2000). 基於小波變換的特徵提取方法分析. 信號處理, 16(2), 156-162.

[15] 黃俊欽（1996）。灰階人臉辨識之研究。國立交通大學資訊科學學系碩士論文，新竹市。 取自https://hdl.handle.net/11296/5yjgn7

[16] Wei, P., Zhou, Z., Li, L., & Jiang, J. (2018). Research on face feature extraction based on K-mean algorithm. EURASIP Journal on Image and Video Processing, 2018(1), 1-9.

[17]肖漢光, & 蔡從中. (2009). 特徵向量的歸一化比較性研究. 電腦工程與應用, 45(22), 117-119.

[18] 沈永泰（2010）。改進人臉偵測器在不同光照環境之效能。國立臺灣科技大學機械工程系碩士論文，台北市。 取自https://hdl.handle.net/11296/3x297y

[19] 陳聖元（2007）。設計與實現一個平行人臉偵測器。國立高雄應用科技大學電機工程系碩士班碩士論文，高雄市。 取自https://hdl.handle.net/11296/ck398p

 
[20] Huang, C. L., & Chen, C. W. (1992). Human facial feature extraction for face interpretation and recognition. Pattern recognition, 25(12), 1435-1444.

[21] Kim, S. K., Park, Y. J., Toh, K. A., & Lee, S. (2010). SVM-based feature extraction for face recognition. Pattern Recognition, 43(8), 2871-2881.

[22] Sobottka, K., & Pitas, I. (1998). A novel method for automatic face segmentation, facial feature extraction and tracking. Signal processing: Image communication, 12(3), 263-281.

[23] Zhang, J., Yan, Y., & Lades, M. (1997). Face recognition: eigenface, elastic matching, and neural nets. Proceedings of the IEEE, 85(9), 1423-1435.

[24] 吳天樂（2021）。有無戴口罩之人臉辨識系統。元智大學電機工程學系乙組碩士論文，桃園縣。 取自https://hdl.handle.net/11296/228mz5