Enhancing Attendance Management Systems Using Facial Recognition

Enhancing Attendance Management Systems Using Facial Recognition

Joel Biju, Shreya Sairam, Kishore Kumar, Surendran M.

Department of Computer Science and Engineering, SRM Institute of Science and Technology, Chennai, Tamil Nadu, India 600089

Abstract A facial recognition system is a computer application for automatically identifying or verifying a person from a digital image or a video frame from a video source. It is a biometric technology, based on the identification of the facial features of the person. Traditional method of managing attendance can be a tedious task as it is time-consuming and inaccurate. This paper proposes a facial recognition system for attendance management. This application is based on face detection, feature extraction and recognition algorithms. We use the KLT Algorithm, Viola Jones Algorithm face detection which detect human face using Haar Cascade classifier.

Keywords : Face detection, Attendance systems, Haar Cascade Classifier, Local Binary Pattern Histogram(LBPH).

I . INTRODUCTION

The face is a unique identity of any person. Hence, it can be used to distinguish people from one another. Hence, it also plays a crucial role in applications such as security system, credit and debit card verification or surveillance on identify criminal public places. The people in photos and in real time videos can be identified using facial recognition systems. It is a tier of biometric security. Other types of biometric security are voice, fingerprint and eye recognition. In real time we can use it to unlock phones, find missing people, aid forensic investigations, help the blind, etc. The main aim of this project is to build an attendance system based on face recognition. Different systems can be used for face recognition like attendance systems with fingerprint scanners, RFID tags and readers, facial recognition and location-based attendance systems.

Attendance serves as a fundamental requirement in various institutions, including educational establishments such

1. RELATED WORK

   The primary objective of this review study was to explore solutions proposed by other authors and assess the constraints of their methodologies.

   One of the systems[1] ,uses Convolutional Neural Networks(CNNs) to implement the facial recognition system. A dependable biometric attendance monitoring system has been developed [2]. This system offers a more efficient method for tracking student presence, significantly diminishing the potential for proxy attendance and addressing

   asschools and universities, as well as in employee-based organizations. It plays a pivotal role in determining a student's eligibility to take exams and an employee's remuneration based on the number of days they are present. Unfortunately, a common observation in educational institutions is the practice of proxy attendance, where attendance is falsely recorded even when the individual is absent. This practice provides an unfair advantage to those who engage in it. Accidental errors in attendance record-keeping, whether by teachers in educational institutions or employees in companies, can occur. To mitigate such limitations, an effective solution is to implement an artificial intelligence system that leverages facial recognition technology. This system can automatically identify individuals as they enter, record their attendance, and securely store the data in a database for future reference. The advantages of employing face recognition in this context are numerous:

   • Minimizes the risk of manual errors.

   • Saves time and resources, as administrators are relieved of the task of manually marking attendance.

   • Easily integrates with Human Resource Management Systems (HRMS) or payroll systems, allowing for flexible time and date formats.

   • Provides real-time automated attendance tracking, maintaining up-to-date records.

   • Offers scalability, making it suitable for both daily attendance and generating accurate timesheets for a large workforce.

   This revised version maintains the key points while improving readability and clarity.

   issues like missing attendance records often associated with manual attendance marking.This system proposes the use of Haar Cascade classifiers and Principal Component Analysis(PCA).

   The authors of [3] proposed the use of a facial recognition algorithm based on Eigen values and Eigen vectors.

   Attendance managements systems such as [4] can be more reliable and accurate by using Local Binary Pattern Histogram (LBPH).

   The authors of [5] propose the usage of cloud storage and computing to store the facial information. This methodology has a drawback as the data cannot be fetched in offline mode as it needs active internet connection.

   Another one of the methodologies[6], proposed the usage of KLT algorithm and Voila-Jones algorithm which detects human faces using Haar Cascade classifier and the PCA Algorithm is used for feature selection.

   One of the proposed models[7],implements the AdaBoost algorithm which increases the efficiency of the facial recognition attendance system.

2. METHODOLOGY

   The methodology for a facial recognition system involves several key steps. First, a dataset of facial images is collected, including individuals whose faces the system should recognize, with variations in lighting, poses, expressions, and backgrounds for robustness. These images undergo preprocessing to enhance their quality and consistency, including resizing, normalization, and noise reduction. A face detection algorithm is then employed to locate and extract facial regions within the images. Subsequently, features are extracted from the detected faces using methods like Principal Component Analysis (PCA), Local Binary Pattern (LBP), or Convolutional Neural Networks (CNNs). These features serve as a numerical representation for recognition. A database is created to store the extracted features along with corresponding identity labels, which is the reference for the recognition system. The recognition algorithm, such as Eigenfaces, Fisherfaces or deep learning CNNs, is used to match the input image's features with those in the database for verification or identification. Thresholds are set for decision-making. Testing and validation are performed to evaluate system performance using metrics like accuracy and F1-score. In real-time applications, the system processes live images for various purposes.

   Fig 1: Proposed Methodology for Attendance monitoring.

   1. 1. Image capturing: The process of taking photographs or pictures of objects, scenes, or individuals using a camera or other imaging devices, capturing visual information for various applications and analysis, such as in photography, surveillance, or computer vision.

   2. Image pre-processing: refers to the initial steps taken to clean, enhance, and format raw data, making it suitable for analysis or further processing. In image processing, it involves tasks like noise reduction, resizing, and contrast adjustment to prepare images for analysis or recognition.

   3. Feature Extraction: Feature extraction is the step where relevant characteristics from facial images, such as key points or distinct patterns, are identified and extracted. These features are crucial for creating unique representations of faces and are used for matching and recognition.

   4. Database Creation: The database creation step involves organizing the feature representations of individuals along with corresponding labels, enabling efficient matching during recognitio.

   5. 5.Recognition Algorithm: Recognition algorithms are applied to compare the extracted features from an input image to those stored in the database. These algorithms determine the likelihood of a match and assist in identifying individuals.

   6. Verification/Identification: This phase verifies or identifies individuals based on the algorithm's results, depending on the application's requirements.

   7. 7.Threshold Setting: Setting a threshold for the acceptance or rejection of recognition results is essential to achieve the desired level of accuracy and security in the system.

   8. 8.Testing and Validation: Testing and validation assess the performance of the facial recognition algorithm using various metrics, ensuring it meets the intended objectives and standards.

3. FEATURE SELECTION AND EXTRACTION

   The purpose of face detection is to determine whether an image contains a face. This process makes use of the Haar Cascade classifier, a robust object detection technique pioneered by Paul Viola and Michael Jones. It's a machine learning-based method that involves training a cascade function with images and is commonly utilized to identify diverse objects within different images.

   1. Haar Cascade Classifier

      The Haar Cascade is a machine learning-based object detection technique developed by Paul Viola and Michael Jones. It's widely used for detecting objects or patterns within images and is particularly well-known for its application in facial recognition.The Haar Cascade works by training a classifier with positive and negative images of the object to be detected. In the case of face detection, positive images are those containing faces, while negative images are those

      without faces. The classifier learns to distinguish the features that are characteristic of the object.

      The term "Haar Cascade" refers to the use of Haar-like features, which are simple rectangular patterns with variations in light and dark areas. These features serve as the basis for detecting the object of interest. The Haar Cascade employs a cascading approach, which involves multiple stages of classifiers. In each stage, the classifier eliminates non-matching regions, and only potential candidate regions move forward to the next stage. This cascading approach significantly speeds up the detection process.

      Fig 2: Haar Cascade Feature Selection .

   2. Integral Image

      An integral image, also known as a summed-area table, is a

      computational technique used in computer vision to accelerate the calculation of the sum of pixel values within rectangular regions in an image. It is derived from the original image and stores cumulative sums of pixel values. This allows for efficient and rapid computation of Haar-like features, which are commonly used in object detection and facial recognition algorithms like the Haar Cascade classifier. The integral image simplifies the computation of sum or average values within rectangular regions in images, reducing the need for repetitive calculations and substantially improving the efficiency of object detection and feature extraction algorithms. Its use is especially prominent in tasks like facial recognition, where rapid and accurate analysis of image data is crucial.

      Fig 4: Calculation of sum of pixels using Integral Image

   3. Principal Component Analysis (PCA)

      Principal Component Analysis (PCA) is a dimensionality reduction technique widely used in data analysis and machine learning. It works by transforming data into a new coordinate system, where the first axis (principal component) captures the most significant variation in the data, the second axis captures the second most significant variation, and so on. This reduction in dimensionality simplifies data while retaining its essential characteristics, making it useful for tasks like data visualization, noise reduction, and feature selection.

      PCA is achieved through a mathematical process that involves calculating the covariance matrix of the data, finding its eigenvectors and eigenvalues, and then selecting the top eigenvectors as the principal components. By projecting the data onto these principal components, PCA enables a more compact representation of the original data, which is particularly valuable when dealing with high-dimensional datasets or when you want to identify patterns and reduce redundancy in the data.

      Fig 3: Integral Image

      Fig 5: Face Recognition methodology using PCA

   4. ADABOOST ALGORITHM

      AdaBoost, short for Adaptive Boosting, is a machine learning algorithm that enhances the performance of weak learners to create a robust and accurate ensemble classifier. It works by assigning weights to data points in the training set and repeatedly training weak classifiers while giving more weight to previously misclassified data points. AdaBoost combines the outputs of these weak learners to make a final prediction, and it's widely used in various classification tasks, including face detection, text categorization, and more, due to its ability to handle complex and high-dimensional data effectively.

      AdaBoost is often employed as part of the feature selection or classifier combination process. AdaBoost helps improve the performance of face recognition algorithms by boosting the accuracy of weak classifiers or features.

      AdaBoost allows for the integration of a multitude of features, making it possible to handle complex facial feature combinations, including Haar-like features, Local Binary Patterns (LBP), or Histogram of Oriented Gradients (HOG), among others.

      This approach involves a series of increasingly complex classifiers, where AdaBoost is used to select the most promising regions for further examination. This cascade structure speeds up the detection process, making it suitable for real-time applications.

      Fig 6: AdaBoost Algorithm implementation

   5. LOCAL BINARY PATTERN HISTOGRAM

   Local Binary Pattern Histogram (LBPH) is a texture analysis technique used in computer vision and image processing. It works by examining the individual pixels in an image and creating patterns based on their brightness in relation to their neighboring pixels.

   Local Binary Pattern: For each pixel, LBPH compares its brightness to its neighbors and assigns a binary value of 1 if it's brighter or 0 if it's darker. This process is applied to all pixels in the image, generating a unique pattern for each pixel.

   Histogram Calculation: These patterns are then counted to create a histogram, which shows how frequently each pattern occurs in the image. The resulting histogram is a representation of the image's texture.

   LBPH is particularly useful for tasks that involve recognizing materials or objects based on their textures, such as identifying different types of fabric or distinguishing between surfaces like wood and metal.

   Fig 7: Facial recognition using LBPH

4. MODULE DESCRIPTION

   To create an AI face recognition system for attendance management, you can streamline the architecture by focusing on five or six essential modules. These modules are designed to efficiently manage attendance using facial recognition technology. Here are the key components:

   1. Data Acquisition Module:

      This module captures facial data from a source, typically a camera or a webcam. It is responsible for collecting images or video frames of individuals as they arrive for attendance.

   2. Face Detection and Preprocessing Module:

      This module detects and localizes faces within the captured images or video frames. It may also include preprocessing tasks, such as noise reduction, resizing, and image enhancement to ensure high-quality face images for recognition.

   3. Face Recognition Module:

      The core of the system, this module performs face recognition by comparing the extracted facial features with a database of known individuals. It identifies or verifies individuals and records their attendance.

   4. Database Management Module:

      This module stores and manages the records of individuals' attendance. It includes adding new users, updating records, and efficient retrieval for recognition tasks.

   5. Liveness Detection Module:

      To prevent spoofing and ensure that the person being recognized is physically present, this module checks for liveness by analyzing movements or requiring user interaction (e.g., blinking or nodding) during the recognition process.

   6. User Interface (UI) Module:

      This module provides a user-friendly interface for system configuration and monitoring. It may include features like registration of individuals, viewing attendance logs, and generating reports.

5. CONCLUSION

   In this paper, we have examined various algorithms employed in attendance management systems, each with its own set of advantages and disadvantages. Attendance management is a crucial aspect of any organization, and many institutions still rely on time-consuming manual methods for tracking attendance. The objective of our application is to address this issue by incorporating face recognition technology into the
   attendance management system.
   The primary goal of this application is to provide an efficient and accurate attendance management solution for organizations such as schools, colleges, and institutes. By implementing this technology, we aim to replace the traditional manual methods with a more secure, reliable, and accessible system. An additional advantage is that it doesn't require specialized hardware; a simple camera and computer setup can be used for its implementation. This innovation promises to streamline attendance management and benefit
   both educators and students.
   While our system is expected to deliver strong performance, it's important to acknowledge that it may not be a flawless or
   ideal solution.

6. FUTURE SCOPE

   The future scope of AI face recognition for attendance management systems holds significant promise. It is likely to witness advancements in accuracy, real-time processing, and scalability, making it even more suitable for large institutions and organizations. Integration with other AI technologies, like machine learning and data analytics, can enhance predictive insights and automation. Additionally, as hardware and software continue to evolve, AI face recognition may become more cost-effective and accessible, leading to broader adoption in various sectors beyond education and employee management.

   The future scope of this project is promising and can be extended by integrating hardware components, such as GSM, to enhance communication. Monthly reports highlighting the attendance of defaulting students can be automatically sent to their respective mentors. Furthermore, the development of a dedicated mobile application for students to monitor their attendance is another avenue for improvement.

7. REFERENCES

1. Nandhini R, Duraimurugan N, S.P Chollalingam Face Recognition Attendance System in IJEAT in 2019. https://www.ijeat.org/wpcontent/uploads/papers/v8i3SC

2. E Varadharajan, R Dharani, S.Jeevitha, B Kavinmathi, S. Hemalatha Automatic Attendance Management system using face detection at IC-GET 2016. 2020 Department of Information Technology. https://ieeexplore.ieee.org/abstract/document/7916753

3. Nithya. C, Ramya Bharathi. M, Santhini. M, Sowmya. R, 2020, Face Recognition based-Automatic Attendance Management System, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) NCICCT 2020 (Volume 8 Issue 08)

4. Suman Kumar Jha, Aditya Tyagi, Kundan Kumar, Madhvi Sharma, 2020, Attendance Management System using Face Recognition, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) ENCADEMS 2020 (Volume 8 Issue 10)

5. Dr. Nkolika O. Nwazor , Mumuni M. Olusolape, 2021, Cloud based Attendance Management and Information System, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) Volume 10, Issue 09 (September 2021)

6. Shivam Singh , Prof. S. Graceline Jasmine, 2019, Face Recognition System, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) Volume 08, Issue 05 (May 2019)

7. Sandhya Potadar , Riya Fale , Prajakta Kothawade , Arati Padale, 2021, Attendance Management system using Face Recognition, INTERNATIONAL JOURNAL OF ENGINEERING RESEARCH & TECHNOLOGY (IJERT) Volume 10, Issue 08 (August 2021)