在本文介紹了一個基於人臉識別模塊的跟踪機器人。其中包括三個模塊分別說明：近年來，隨著科學技術的飛速發展，圖像技術正變得越來越成熟，用於圖像處理，圖像控制;計算機視覺是近年來許多人的研究課題。本文主要使用微軟Kinect V2相機開發的產品進行人機交互。在不使用任何其他傳感器的情況下，使用手勢來控制移動機器人執行命令，以使移動機器人跟隨人的動作。結果表明，移動機器人可以有效地接收姿勢動作來執行命令，並且可以準確地追踪人。
除了近年來深度學習的興起之外，深度學習的應用也越來越多，本文利用深度學習來加強對人物的判斷力。所使用的方法分為四個部分：（a）使用哈爾特徵算法找出已經判斷的人臉。（b）從屏幕上收集臉部樣本。（c）收集的樣本是分類器並使用自適應增強級聯分類器進行訓練。（d）驗證培訓模式。實驗結果證明該系統能夠實時正確判斷人員，並取得良好的效果。
在一般的人臉識別中，經常遇到一些外界因素，比如那種情況人臉不是面向相機，角度過大，識別精度低，系統無法識別。本文分為三部分：1，採用多特徵點算法採集樣本，隨機改變模塊的對比度和亮度，增加樣本的多樣性。 2.卷積神經網絡訓練模型可以讓系統記住我的臉部特徵數據。 3.驗證模型。實驗結果表明，使用多特徵點算法可以提高人臉在不同角度的識別率，而在使用較小數據集的情況下，該方法可以成功識別人臉。

This article describes a following robot based on face recognition modules. This includes three modules that are illustrated in recent years, with the rapid development of science and technology. Image technology is becoming more and more mature, for image processing, image control; computer vision is the research subject of many people in recent years. This article is main using the product developed by Microsoft Kinect V2 camera for human-machine interaction. Without using any other sensors, the user use gestures to control the mobile robot to execute the command, so that the mobile robot is follow the people action. The result indicates that the mobile robot can effectively receive the gesture action to execution the command, and can accurately follow people.
In addition to the rise of deep learning in recent years, the application of deep learning has been increasing. This paper uses deep learning to strengthen the judgment of the characters. The method used is divide into four parts: (a) Use the Haar-like feature algorithm to find out the face that has been judge. (b) Collect face samples from the screen. (c) The collected samples were classifier and trained using the Adaptive boosting cascading classifier. (d) Verification of training model. The experimental results prove that the system can correctly determine the personnel in real time and have a good effect.
In general face recognition, some external factors are often encountered, such as the case that the face is not facing the camera, the angle is too large, the recognition accuracy is low, and the system cannot identify. This paper is divided into three parts: 1. Use multi-feature point algorithm to collect samples and randomly change the contrast and brightness of the module to increase sample diversity. 2. Convolutional Neural Network training model can let the system remember my face features data. 3. Validation model. The experimental results show that using multi-feature point algorithm can improve the recognition of face in different angles, and in the case of using smaller data sets, this method can identify human faces.

[1] A. M. Ahmad and H. A. Youssef, "3D sensor-based Moving Human Tracking Robot with Obstacle Avoidance," 2016.
[2] Y. L. J. L. Z. Ju, "Data Fusion-based Real-Time Hand Gesture Recognition with Kinect V2 " Portsmouth, UK 2016.
[3] Y. Wang, J. Sun, J. Li, and D. Zhao, "Gait recognition based on 3D skeleton joints captured by kinect," Phoenix, AZ, USA 2016.
[4] H. Alabbasi, A. Gradinaru, F. Moldoveanu, and A. Moldoveanu, "Human motion tracking & evaluation using Kinect V2 sensor," 2015.
[5] S. Amir, A. Waqar, M. A. Siddiqui, M. Majeed, H. Haseeb, and S. H. Mehmood, "Kinect Controlled UGV," Wireless Personal Communications, vol. 95, no. 2, pp. 631-640, Jul 2017.
[6] M. G. D. F. T. P. R.-G. D. Gonzalez-Aguilera, "Analysis and Evaluation Between the First and the Second Generation of RGB-D Sensors," IEEE Sensors Journal, vol. 15, pp. 6507 - 6516, 2015.
[7] Q. Du, R. Liu, and Y. Pan, "Depth extraction for a structured light system based on mismatched image pair rectification using a virtual camera," pp. 1086 - 1093, 2017.
[8] J. Gan, X. Wang, and Y. Zhai, "A Real-Time Face Recognition System Based on IP Camera and Image Sets Algorithm," in 2015 IEEE 12th Intl Conf on Ubiquitous Intelligence and Computing and 2015 IEEE 12th Intl Conf on Autonomic and Trusted Computing and 2015 IEEE 15th Intl Conf on Scalable Computing and Communications and Its Associated Workshops (UIC-ATC-ScalCom), 2015, pp. 1524-1528.
[9] L. Cuimei, Q. Zhiliang, J. Nan, and W. Jianhua, "Human face detection algorithm via Haar cascade classifier combined with three additional classifiers," in 2017 13th IEEE International Conference on Electronic Measurement & Instruments (ICEMI), 2017, pp. 483-487.
[10] M. Heshmat, M. Girgis, W. M. Abd-Elhafiez, and S. Elaw, "An efficient scheme for face detection based on contours and feature skin recognition," in 2015 Tenth International Conference on Computer Engineering & Systems (ICCES), 2015, pp. 255-260.
[11] M. A. S. Mollah, M. A. A. Akash, M. Ahmed, and M. A. H. Akhand, "Improvement of haar feature based face detection incorporating human skin color analysis," in 2016 International Conference on Medical Engineering, Health Informatics and Technology (MediTec), 2016, pp. 1-6.
[12] M. Yu, L. Yun, Z. Chen, and F. Cheng, "Research on video face detection based on AdaBoost algorithm training classifier," in 2017 First International Conference on Electronics Instrumentation & Information Systems (EIIS), 2017, pp. 1-6.
[13] K. Goyal, K. Agarwal, and R. Kumar, "Face detection and tracking: Using OpenCV," in 2017 International conference of Electronics, Communication and Aerospace Technology (ICECA), 2017, vol. 1, pp. 474-478.
[14] A. R. Bhatia, N. M. Patel, and N. C. Chauhan, "Parallel implementation of face detection algorithm on GPU," in 2016 2nd International Conference on Next Generation Computing Technologies (NGCT), 2016, pp. 674-677.
[15] L. A. Elrefaei, A. Alharthi, H. Alamoudi, S. Almutairi, and F. Al-rammah, "Real-time face detection and tracking on mobile phones for criminal detection," in 2017 2nd International Conference on Anti-Cyber Crimes (ICACC), 2017, pp. 75-80.
[16] V. Kazemi and J. Sullivan, "One millisecond face alignment with an ensemble of regression trees," in 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014, pp. 1867-1874.
[17] S. Huang, F. Shi, and X. Yao, "Sparse representation with dense matching for face recognition from single sample per person," in 2017 Chinese Automation Congress (CAC), 2017, pp. 6557-6562.
[18] K. Awedat, A. Essa, V. Asari, and D. Stoppenbrink, "Sparse representation based classification performance under different optimization forms for face recognition," in 2017 IEEE National Aerospace and Electronics Conference (NAECON), 2017, pp. 34-37.
[19] Y. Sun, X. Wang, and X. Tang, "Deep Learning Face Representation from Predicting 10,000 Classes," in 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014, pp. 1891-1898.
[20] Y. C. Yi Sun, Xiaogang Wang, Xiaoou Tang, "Deep Learning Face Representation by Joint Identification-Verification," Montreal, Canada, 2014.
[21] H. Fan, Z. Cao, Y. Jiang, Q. Yin, and C. Doudou, " Learning Deep Face Representation," CoRR, vol. abs/1403.2802, 2014.
[22] Y. Sun, X. Wang, and X. Tang, "Deeply learned face representations are sparse, selective, and robust," in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 2892-2900.
[23] H. Li, Z. Lin, X. Shen, J. Brandt, and G. Hua, "A convolutional neural network cascade for face detection," in 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 5325-5334.
[24] D. Pagliari and L. Pinto, "Calibration of Kinect for Xbox One and Comparison between the Two Generations of Microsoft Sensors," 30 October 2015.
[25] P. Sturm, "Pinhole camera model," in Computer Vision, pp. 610{613, Springer, 2014.
[26] L. Valgma, "3D reconstruction using Kinect v2 camera," Computer Engineering, University of Tartu, Tartu Ülikool, 2016.
[27] F. Schlagenhauf, P. P. Sahoo, and W. Singhose, "Comparison of single-kinect and dual-kinect motion capture of upper-body joint tracking," Gold Coast, QLD, Australia 17-20 Dec. 2017.
[28] LIU Yang, SHANG Zhaowei. Traffic gesture recognition based on Kinect skeleton data. Computer Engineering and Applications, 2015, 51（3）：157-161.
[29] Izhar Haq, Shahzad Anwar, Kamran Shah, Muhammad Tahir Khan, Shaukat Ali Shah and Jonathan A Coles, Fuzzy Logic Based Edge Detection in Smooth and Noisy Clinical Images, PLOS ONE, 10, 9, (e0138712), (2015).