Modern Technique Of Lecture Attendance Using Face Recognition. Shreya Nallawar, Neha Giri, Neeraj Deshbhratar, Shamal Sane, Trupti Gautre, Avinash Bansod Bapurao Deshmukh College Of Engineering, Sewagram, RTM Nagpur University nsshreya93@gmail.com, nehagiri0406@gmail.com, neeraj6488@gmail.com,shamalsane93@yahoo.com, truptigautre2@gmail.com,ironeyeme@gmail.com A B S T R A C T Attendance management of students in any institution is a very lengthy process and even time consuming. Furthermore, biometrics attendance system is also available. These methods too are time consuming, since each time students have to form a queue for scanning their thumb impression. In this paper, we have proposed a system which deals in terms of face recognition using real time camera. The work presented in this paper proposes a method to automatically take the attendance of student using face recognition. Continuous observation improves the precision of attendance. The attendance will be recorded by using camera(s), attached in front of the class which is continuously capturing images of the students. It will then compare the faces with the student s database and marks the attendance. The important key of this paper is to design a better student attendance system with ease interface and accurate results. Regarding to the student and lecturer sides, the system is working without any preparation and no extra effort and the most important thing is to increase the quality of our educational system. Index Terms: Face Recognition, Face Detection, Monitoring, Attendance System, and System Access Control. I. INTRODUCTION Nowadays, taking attendance in any field is very important task so as to maintained the record of student, employee etc. Traditionally, student attendance is taken manually by using attendance sheet given by faculty member in class which is very time consuming. We observed that the technique used was very time consuming and even many demerits has been observed by us such as wastage of paper, interruption in classroom etc. previously a system has been proposed by continues observation which improves the performance for the estimation of the attendance. Implementation of the system has to be carried out on accordance of some techniques named as face detection and face recognition. The Face detection and face recognition are very advanced in terms of computer authentication technology. The technology of student attendance system is used to support the teacher for checking student attendance in modern way. It is gradually evolving to a universal biometric solution since it requires virtually zero effort from the user end while compared with other biometric options. The system is going to work by some techniques such as the picture is taken by camera then processed towards the detection as the detected face image is obtained face recognition has to be done which is divided into further parts namely face alignment, preprocessing, feature extraction, face matching where the image is converted into gray scale image and the result has to be seen. This has been done by using PCA algorithm. This technique is considered to be one of the most successful for image processing or analysis. Face recognition is a biometric method for identifying any individual by the features of their face. Applications of face detection are widely used in areas such as security systems, criminal identification etc. Using a pre stored database, we can identify one or more identities in the scene. The general block 59 2014, IJAFRSE All Rights Reserved www.ijafrse.org
diagram for face detection system consists of three blocks. The first is the Face Detection, the second is feature extraction, and the third is Face recognition. The general block diagram is as shown in below figure 1. Figure 1. General Block Diagram of Face Recognition The paper is organized in this way: Section I gives the introduction of the paper. Literature survey is given in the section II. The method is explained in this section III. Implementation techniques are explained in section IV. The algorithms used are explained in section V. Concludes the paper. II. LITERATURE REVIEW Here the details about the referred paper, author are given below Identification of human faces [11] Author: A.J. GoldStein, L. D. Harmon, and A.B. Lesk Lesk used 21 different facial marks such as hair color and lip thickness for processing the image recognition. The problem was, it had to be manually compute the measurements and location. Integrated system for monitoring and recognition student during class session Author: Mohammad A. Alia, Abdelfatah Aref Tamimi and Omaima N. A. Al-Allaf In this paper, the proposed attendance system is based on face detection and recognition authentication methods. The system is minimizing the lecturer exhaustion since the lecturer can check their student s attendance by their own systems. The camera changes its direction to monitor the students during the lecture. Classroom Attendance System Using Facial Recognition System [2] Author: Abhishek Jha The PCA algorithm for face detection and PCA and LDA for feature extraction has been used. The whole system is implemented in MATLAB. The main advantage of their system is that, the student can also keep track of their attendance by their own login id. A Biometric Authentication Approach using Face Recognition System [12] Authors: Pallabi Saikia, and Margaret Kathing Here the feature extraction was carried out using PCA algorithm. The eigenface method was implemented for feature extraction and classification. 60 2014, IJAFRSE All Rights Reserved www.ijafrse.org
A MatLab based face recognition system using image processing Author: Jawad Nagi, Syed Khaleel Ahmed, Farrukh Nagi They have presented a novel face recognition technique which uses feature from Discrete Cosine Transform (DCT) coefficient along with a SOM based classifier. This makes there system well suited for low cost, real time hardware implementation. Face Recognition Based on Principal Component Analysis [3] Author: Ali Javed The paper was a research work of face recognition system by using PCA which is eigenvector based multivariate analyses. Real Time Face Recognition Using AdaBoost Improved Fast PCA Algorithm [1] Author: K. Susheel Kumar, Shitala Prasad, Vijay Bhaskar Semwal, R C Tripathi. For the detection of face a real time human face AdaBoost and haar cascade has been used and PCA along with LDA algorithm for face recognition which imparts the high accuracy rate. III. PROPOSED SYSTEM In the earlier systems the attendance of the students has been marked for the whole day at once. Thus if a student is present for only a lecture and then left the class, even then he will be marked as present for whole day. We propose a system for automated attendance of the student which will mark the attendance for each lecture in the classroom via face detection and face recognition. We will create the database of face of each of the student of each classroom. A digital camera in the center of front wall on the classroom is placed. The camera will get automatically ON for some time period in the mid the lecture and it captures the image of the student in the classroom. The image with the best localization of faces of student will be considered for further processing. The image is then processed for face detection via Viola and Jones face detection methods. It will detect all the faces present in the image with maximum efficiency. Then the detected face images are been compared with the images of the student in the database for recognition process. The database contains the record of all of the students in each class. If any of the face in the picture taken in the classroom is matched with the image in the database then the student with that face is marked as present in the classroom. 61 2014, IJAFRSE All Rights Reserved www.ijafrse.org
Figure 2. Block Diagram of Attendance system using Face Recognition IV. METHODS We have proposed a system in which we will mark the attendance of the student for each lecture. The camera attached in the class room will click the image of the students in the classroom in the mid of each lecture. The clicked image is then processed for face detection. This process separates the facial area from the rest of the image. The facial area of all of the students is extracted then segmentation is done to align the image properly. This process is done to find the best localization and normalization of the image. Then face recognition is done which has following steps- Face alignment, feature extraction, feature matching. If the face image is matched with an image of student in the database, then the student has attended the lecture and is marked as present in the classroom. Figure 3: System Architecture V. IMPLEMENTATION TECHNIQUES Principal Component analysis (PCA) is based on information theory approach. It identifies the subspace of image space spanned by the training face image data and de-correlates the pixel values. As compared to Linear Discriminant Analysis (LDA), PCA algorithm does more of feature classification and LDA does data classification A. Face Detection Face detection is a technology that determines the location and sizes of human faces in an image. It detects faces and ignores anything else, such as building, chairs, and trees. It is a starting point for face recognition. Most of the face detection methods focus on detecting frontal faces. These methods are categorized [6] into four types: Knowledgebase, Feature invariant, Template matching and Appearance- Based. Each method involves color segmentations, pattern matching, statistical analysis and complex transform. Face detection is an important part of face recognition as to implement the automatic face recognition. And therefore we will use the algorithm based on Ada Boost and Haar cascade in which we will detect the face by the face model. Face model contains the shapes and motions of faces. In this technique we found different shapes on face models such as rectangle, triangle, circle, heart, and square. And again if some other techniques are used to implement this the high level of accuracy rate will impart. B. Face Segmentation Segmentation is one of the very first steps in automatic face recognition system. The aim of the segmentation is to make the image more represent able. As the camera will click the picture of the 62 2014, IJAFRSE All Rights Reserved www.ijafrse.org
students in the classroom which will goes through the process for face recognition where we found face image of the whole classroom, thus it will segments face of the each student and preprocess the features of each student and removes the noise from the picture. C. Face Recognition Face Recognition is automatic identification or verification of a person from an image/video. It is one of the most active and widely used techniques because of its reliability, accuracy in the process of recognizing and verifying the person s identity. Problem that may occur with face recognition aredifferent people may look similar, characteristic of the face may change with time. Face can be recognized by two approaches that are based on geometry of face and based on appearance of face. The recognition process is done by comparing the extracted features from the image with the one previously stored in the database. Face recognition can be implemented bay using Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Discrete Cosine Transform (DCT) etc. Figure 4: Block diagram of Face Recognition Face Recognition has been divided in following parts: 1. Preprocessing The detected image is processed for removing the noise and sets the unique intensity of light for the face image. Preprocessing is also known Normalization. The input colored image is converted into gray scale image 2. Face Alignment After face segmentation the normalization and localization has to found out of the face. Alignment has to perform to estimate the outline of the facial components such as nose, eyes, mouth and ears as after detection the position of the face is roughly estimated. Thus after normalization the face image is converted into gray scale image. 3. Feature Extraction After the previous two steps feature extraction has been performed which enhances the various segments of the face namely eyebrows, nose, mouth, ears. Feature extraction includes two algorithms: 63 2014, IJAFRSE All Rights Reserved www.ijafrse.org
Facial component are to be selected with high level of accuracy. Normal center of gravity (NCG) is to be determined. Normal center of gravity is determined of each features of the face denoted by asterisk (*). Figure 5: Determination of NCG VI. ALGORITHMS USED A. Viola and Jones Face Detection For Viola and Michael Jones in their 2003 article titled Robust Real Time Face Detection, proposed the face detection methods. They describe how one can use machine-learning technique to construct sets of meaningful feature that encode image properties and will detect faces. It is quite fast method but optimization can further speed up the detection. Viola and Jones had made three key innovations: the first was the new representation of image call the integral image, for faster feature computation, the second was the use of the AdaBoost (Adaptive Boosting) machine learning algorithm for selecting simple and efficient classifiers and the third and last was a method of combining classifiers into a Cascade that quickly eliminates background regions and focus computational attention on more promising areas of the image. 1. Haar- Like Feature Haar-like features are rectangular digital image that provide a method for encoding the properties of the image in a form that can be computed much more quickly as compared to individual pixels. Simple Haarlike features are composed of two adjacent rectangles, located at any scale and position within an image, and is referred to as 2-rectangle feature. The feature is defined as the difference between the some of image intensities within each rectangle. Viola and Jones also extended this set by defining similar features composed of 3 and 4 rectangles. This types of features are quite course when compared to alternatives such as steerable filters, however, there computational efficiency more than makes for their limitations. 2. Integral Image 64 2014, IJAFRSE All Rights Reserved www.ijafrse.org
Haar-like features can be calculated quickly by using an image representation known as integral image. It is an application of summed area tables. The integral image can be calculated in single pass. Each feature can be calculated in a constant time by using it. 3. Feature Selection Viola and Jones hypothesized and discovered through experimentation that a very small number of feature can be form into effective classifier. For a classifier to be effective the set it is trained on must contain a good range of facial variation. No single feature can be used as an effective classification function. The AdaBoost algorithm creates the striner classifiers by searching the set of all weighted combinations of weak classifier and selective the most successful combination. The newly obtained strong classifier is combined with the optimal threshold which enables it to best separate faces from non faces. 4. Attentional Cascade The advent of the attentional cascade is the most important innovation of Viola and Jones methods. Its focuses first on removal of negative regions of the image while including all positive ones. The method was to use two neural networks: the algorithm first uses the faster neural network to select regions of interest before running the slower neural network which is from complex then first network and is used to pick out the faces from the image. The initial stages are created by adjusting the AdaBoost by latest staged use more complex classifier to reduce overall falls positive weights. 5. Algorithm and Implementation By the above steps the training process is completed and a classifier cascade has been created with desire properties. The detection algorithm simply scans all possible sub windows of an image at a range of scales, running the cascade on each window. If a sub window passes the final level of the cascade then the sub window will contain a face. In some steps normalization may occur. Firstly the image intensities of both the images (training and test images)must be normalized to the same scale. Then, while running the cascade on the sub windows, the rectangle sums within each feature must be scaled accordingly. Lastly training is done on variance normalized images and therefore, the test windows must be variance normalized as well. B. Principal Component Analysis (PCA) Principal Component Analysis (PCA) algorithm [2] is used to recognize the faces in the image. It is mathematically defined as an orthogonal linear transformation that transforms the data to a new coordinate system. It involves the procedure that transforms the number of possibly correlated variables called Principal components. It involves the calculation of the Eigen value decomposition of a data covariance matrix or singular value decomposition of a data matrix, after mean centering the data for each attribute. The results of PCA are in terms of component scores and loading. PCA is theoretically the optimal linear scheme for compressing a set of high dimensional vectors into a set of lower dimensional vectors and then reconstructing the original sets. PCA algorithm is as follows: A data matrix (m x n) for each image is created which is then converted into an m * n x 1 matrix having rows equal to the product of number of rows and columns of the original matrix. 65 2014, IJAFRSE All Rights Reserved www.ijafrse.org
A mean matrix is created for all the different image matrices. The mean matrix is calculated by adding all the columns of data matrix divided by the total number of columns. The mean subtracted data matrix is obtained by subtracting the mean image from all the image matrices. The covariance matrix is obtained by multiplying the mean subtracted matrix by its transpose to make it a square matrix in next phase. The system then finds the Eigen vectors and Eigen values. For N dimensional vectors there will be N Eigen values and Eigen vectors Then the Eigen image is created by multiplying mean subtracted data matrix with the Eigen vectors. Eigen vectors with highest Eigen value is the principal component of the data set having maximum information. The weight matrix is then calculated by multiplying the transposed large Eigen image with the mean subtracted data matrix. After these steps the system can recognize any face image by comparing it with the main weight matrix. IV. ADVANTAGES Database of all the students is located in central database. Only one camera is to be connected to accessing PC. Saving of our resources (such as paper) and time. VIII. CONCLUSION In order to obtain the attendance in classroom lecture, we proposed the attendance management system based on face recognition in the classroom lecture. The system estimates the attendance of each student by continuous clicking of images for some time period and finds the best localized image for processing. The system allows the lecturer to check his/her student attendance automatically by using personal computer (PC) without any extra cost and effort. As well as, the proposed system needs only the basic requirements such as; camera, PC, and local network. The overall system is implemented in MATLAB. IX. REFERENCES [1] K. Susheel Kumar, Shitala Prasad, Vijay Bhaskar Semwal and R.C Tripathi, Real Time Face Recognition Using Adaboost Improved Fast PCA Algorithm, Internatinal journal of Artificial intelligence and Application, Application, Vol. 2, No.3, july.2011, DOI: 10.5121/ijaia.2011.2305 45 [2] Abhishek Jha, Classroom Attendance System Using Face Recognition System, The International journal of Mathematics, science, technology and Management, Vol. 2, ISSD: 2319-8125. 66 2014, IJAFRSE All Rights Reserved www.ijafrse.org
[3] Ali Javed, Face Recognition Based on Principal Component Analysis, I. J. Image, Graphics and Signal Processing. Vol. 2, DOI: 10.5815/ijigsp.2013.02.06, pp [4] W. Zhao, R. Chellappa, P. J. Phillips and A. Rosenfeld, Face recognition: A literature survey, ACM Computing Surveys, 2003, vol. 35, no. 4, pp. 399-458. [5] Yang J, Zhang D, Alejandro F F, Yang J Y, Two dimensional PCA: a new approach to appearancebased face representation and recognition, Pattern Analysis and Machine Intelligence, 2004, 26: 131-137. [6] Balwant Singh, Sunil Kumar, Paurush Bhulania, Lecture Attendance System WITH Face Recognition And Image Processing, International Journal Of Advance Research In Science And Engineering, Vol. No.2, Issue No.3, March, 2013 ISSN-2319-8354(E), http://www.ijarse.com IJARSE [7] Dayanand S. Shilwant,Dr. A.R.Karwankar, Student Monitoring By Face Recognition System, International Journal of Electronics, Communication & Soft Computing Science and Engineering, Vol. 2, Issue 2, ISSN: 2277-9477, 24 [8] D. Cristinacce and T. Cootes, Facial feature detection using adaboost with shape constraints, in Proc. 14th British Machine Vision Conference, Norwich, UK, Sep.2003, pp. 231 240. [9] E. Hjelmås, and B. K. Low, Face detection: A survey,computer Vision and Image Understanding, Vol. 83, No. 3, Sept. 2001, pp. 236-274. [10] S. Rao, and K. J. Satoa. An Attendance MonitorinG System Using Biometrics Authentication. International Journal of Advanced Research in Computer Science and Software Engineering. Volume. 3, Issue 4, pp. 379-383. 2013 [11] A. J. Goldstein, L. D. Harmon, and A. B. Lesk, Identification of human faces, Proc IEEE, May 1971, Vol. 59, No. 5, 748-760 [12] Pallabi Saikia, Margaret Kathing, A Biometric Authentication Approach using Face Recognition System, International Journal of Advanced Research in Computer Science and Software Engineering, March 2014, Vol. 4, Issue 3 67 2014, IJAFRSE All Rights Reserved www.ijafrse.org