Design and Case Study of a Face Recognition-Based Automated Attendance System Using LBPH Algorithm

Design and Case Study of a Face Recognition-Based Automated Attendance System Using LBPH Algorithm

Hariom Agnihotri(1), Minakshi Yeole(2), Aman Pardeshi(3), Vina Ghadge (4)

(1)Head of Department, Diploma in Computer Engineering (2)Senior Lecturer, Diploma in Computer Engineering

(3)Lecturer, Diploma in Computer Engineering (4)Lecturer, Diploma in Computer Engineering

Department of Computer Engineering, Shri Gulabrao Deokar Polytechnic, Jalgaon, Maharashtra, India

Abstract – Traditional attendance systems in educational institutions often rely on manual processes that are time- consuming and prone to human errors. To overcome these limitations, this paper presents the design and case study of a face recognition-based automated attendance system using the Local Binary Pattern Histogram (LBPH) algorithm. The proposed system captures facial images through a camera, detects faces using computer vision techniques, and recognizes individuals by comparing them with stored facial data in a database. Once a match is identified, the system automatically records attendance, reducing the need for manual intervention. The implementation utilizes modern image processing and machine learning techniques to ensure accuracy and efficiency in real-time attendance tracking. The results demonstrate that the system provides a reliable and secure solution for automated attendance management in educational institutions.

Keywords Face Recognition, Automated Attendance System, LBPH Algorithm, Computer Vision, Image Processing, Machine Learning

1. INTRODUCTION

   The rapid advancement of Artificial Intelligence (AI) and Machine Learning (ML) technologies has significantly transformed various sectors, including healthcare, security, banking, and education. In recent years, educational institutions have increasingly adopted intelligent systems to automate administrative and academic processes. Among these processes, student attendance management remains a critical yet traditionally manual task. Conventional attendance systems, which rely on roll calls or signature-based registers, are time- consuming, prone to human error, and vulnerable to proxy attendance. These limitations highlight the need for a reliable, automated, and efficient attendance monitoring mechanism.

   Face recognition technology, a subdomain of computer vision, has emerged as a robust biometric identification technique due to its non-intrusive and contactless nature. Unlike fingerprint or RFID-based systems, face recognition does not require physical interaction with devices, making it suitable for classroom environments. With the integration of machine learning algorithms, face recognition systems can accurately identify

   individuals by analysing facial features and comparing them with stored datasets.

   Various approaches have been proposed for facial recognition, including deep learning-based Convolutional Neural Networks (CNNs) and classical machine learning techniques such as Eigenfaces, Fisherfaces, and Local Binary Pattern Histogram (LBPH). While deep learning models provide high accuracy, they demand large datasets, high computational resources, and specialized hardware such as GPUs. Such requirements may not be practical for diploma-level institutions with limited infrastructure. Therefore, lightweight and efficient algorithms like LBPH offer a practical alternative for real-time classroom deployment.

   The Local Binary Pattern Histogram (LBPH) algorithm is widely recognized for its simplicity, computational efficiency, and robustness under controlled lighting conditions. It works by extracting local texture features from facial images and generating histograms for classification. Due to its low resource requirement and ease of implementation using OpenCV, LBPH is well-suited for academic case studies and institutional-level applications.

   This paper presents the design and implementation of a face recognition-based automated attendance system using the LBPH algorithm. A case study was conducted in a Diploma in Computer Engineering classroom environment at Shri Gulabrao Deokar Polytechnic, Jalgaon. The system captures student images through a webcam, detects faces using Haar Cascade classifiers, recognizes registered students using the trained LBPH model, and records attendance automatically in a structured database. The performance of the system is evaluated based on recognition accuracy, processing time, and operational feasibility in a real classroom setting.

   The primary objective of this study is to demonstrate that an efficient, cost-effective, and scalable attendance system can be implemented using classical machine learning techniques without the need for complex deep learning architectures. The proposed approach aims to reduce manual workload, prevent proxy attendance, and enhance institutional efficiency while

   maintaining practical feasibility for diploma-level educational institutions.

2. LITERATURE REVIEW

   Face recognition has been widely studied in the field of computer vision and biometric authentication systems. Over the years, various techniques and algorithms have been proposed to improve the accuracy, efficiency, and reliability of facial recognition systems. Researchers have explored multiple approaches ranging from classical statistical methods to modern deep learning models.

   One of the earliest approaches in face recognition was the Eigenfaces method proposed by Turk and Pentland, which used Principal Component Analysis (PCA) to represent facial images as a combination of eigenvectors. Although this method was effective for basic recognition tasks, its performance was significantly affected by changes in lighting conditions, facial expressions, and pose variations.

   Another widely used method is the Fisherfaces technique, which applies Linear Discriminant Analysis (LDA) to maximize the separability between different face classes. While Fisherfaces improved recognition accuracy compared to Eigenfaces, it still required well-aligned facial images and controlled environmental conditions for optimal performance.

   Later research introduced the Local Binary Pattern Histogram (LBPH) algorithm, which focuses on extracting local texture features from facial images. LBPH converts facial images into binary patterns by comparing each pixel with its neighboring pixels and then creates histograms representing these patterns. This method has proven to be highly effective for face recognition tasks, particularly in environments with moderate variations in illumination.

   Recent advancements in deep learning have led to the development of Convolutional Neural Networks (CNNs) for face recognition. CNN-based systems such as FaceNet and DeepFace have demonstrated very high recognition accuracy. However, these models require large training datasets, high computational resources, and specialized hardware such as GPUs, which may not always be feasible for smaller educational institutions.

   Several studies have applied face recognition technology for automated attendance systems in classrooms and workplaces. These systems typically involve capturing facial images through cameras, detecting faces using computer vision techniques, recognizing individuals using trained models, and automatically recording attendance in a database. Such systems help reduce manual effort, prevent proxy attendance, and improve the overall efficiency of attendance management.

   The proposed system in this paper utilizes the Local Binary Pattern Histogram (LBPH) algorithm for face recognition due to its simplicity, low computational requirements, nd suitability for real-time implementation in educational environments. By integrating face detection, recognition, and

   automated database recording, the system provides an efficient and reliable solution for attendance management.

3. PROPOSED SYSTEM AND METHODOLOGY

   The proposed system aims to automate the process of attendance management in educational institutions using face recognition technology. The system utilizes computer vision techniques and the Local Binary Pattern Histogram (LBPH) algorithm to identify individuals based on their facial features. By integrating face detection, recognition, and automated database recording, the system eliminates the need for manual attendance marking and reduces the possibility of proxy attendance.

   The system captures real-time facial images of students using a webcam installed in the classroom. These images are processed using face detection algorithms to locate the facial region. Once the face is detected, the system extracts facial features and compares them with previously stored facial data in the database. If a match is found, the system automatically records the attendance of the corresponding student along with the date and time.

   The proposed system is designed to be efficient, cost-effective, and suitable for deployment in educational institutions such as schools, colleges, and diploma institutes. By using the LBPH algorithm, the system ensures reliable performance with relatively low computational requirements, making it practical for real-time applications.

   1. System Architecture

      The overall architecture of the proposed system consists of several functional components that work together to perform face detection, recognition, and attendance recording. The system begins with image acquisition using a webcam, which captures real-time video frames of students present in the classroom.

      The captured images are processed using a face detection module based on the Haar Cascade classifier. This module identifies and isolates facial regions from the captured frames. After detecting the faces, the system converts the images into grayscale format to simplify feature extraction.

      The recognized faces are then processed using the LBPH algorithm, which extracts local texture patterns from the facial images. These extracted features are compared with the trained dataset stored in the system. When a match is found, the attendance management module automatically updates the attendance database with the student’s identity and timestamp.

      Webcam Camera Image Capture

      Face Detection Module(Haar Cascade Algorithm) Face Recognition Module (LBPH Algorithm) Student Database Match

      Attendance Report Update (CSV/ MySQL Database)

       

      Fig. 1. System Architecture of Proposed System.

   2. Algorithmic Flow

      The algorithmic flow of the proposed system describes the sequence of steps involved in detecting faces, recognizing individuals, and recording attendance automatically. The process begins with capturing real-time video input from a webcam and continues through a series of image processing and recognition stages until attendance is recorded in the database.

      In the recognition phase, the system performs real-time face recognition and automatically marks attendance.

      • Capture live video stream through a webcam.
      • Detect faces in each video frame.
      • Extract facial features using the trained LBPH model.
      • Compare extracted features with the stored dataset.
      • If a match is found, the system marks attendance automatically.
      • If no match is found, the system labels the face as Unknown

        D. LBPH Algorithm Description

        The Local Binary Pattern Histogram (LBPH) algorithm is used for facial feature extraction and recognition in the proposed system. The algorithm analyzes local texture patterns in grayscale facial images and represents them as numerical feature vectors.

        Step 1: Local Binary Pattern (LBP) Calculation

        For each pixel in the grayscale image:

        • Compare the center pixel with its neighbouring pixels.
        • Assign a binary value to each neighbour: 1 if neighbour centre pixel

          0 if neighbour < centre pixel

        • Convert the resulting binary pattern into a decimal number.

        Step 2: Histogram Generation

        • Divide the facial image into multiple grids.
        • Compute the histogram of LBP values for each grid.
        • Concatenate all histograms to form a single feature vector representing the face.

          Step 3: Classification

        • Compare the generated feature vector with stored training vectors.
        • Compute the similarity distance using Euclidean distance.
        • The minimum distance indicates the best match.

        Mathematical Representation

        The LBP value for a pixel is calculated as:

        LBP(xc, yc) = (p=0P1) s(ip ic) × 2^p

        Fig. 2. Flowchart of Proposed System

   3. Working Methodology

   The methodology of the proposed system is divided into two main phases: the Training Phase and the Recognition Phase.

   1. Training Phase

      During the training phase, the system collects and processes facial images of registered students in order to create a trained recognition model.

      • Collect facial images of enrolled students
      • Convert images into grayscale format for efficient processing
      • Detect facial regions using the Haar Cascade classifier
      • Extract facial features using the LBPH algorithm
      • Train the LBPH model and store the feature vectors in the dataset
   2. Recognition Phase

   Where:

   ic = intensity of centre pixel

   ip = intensity of neighbouring pixel s(x) = 1 if x 0, otherwise 0

   E. Implementation Environment

   The proposed face recognition-based attendance system was developed using commonly available software tools and hardware components to ensure ease of deployment and cost- effectiveness. The system was implemented using the Python programming language due to its extensive support for machine learning and image processing libraries.

   The OpenCV library was used to perform face detection and recognition tasks. Haar Cascade classifiers were utilized for detecting facial regions in captured images, while the Local Binary Pattern Histogram (LBPH) algorithm was applied for facial feature extraction and recognition. Additionally, the NumPy library was used for numerical computations and handling image data efficiently.

   The system was executed on a standard laptop equipped with a built-in webcam for capturing real-time video input. The webcam continuously captures frames from the classroom environment, which are then processed by the system for face detection and recognition.

   For attendance storage, a lightweight database approach was adopted using a CSV file or MySQL database to store student attendance records along with timestamps. This allows easy retrieval, modification, and analysis of attendance data when required.

   The implementation environment used in the proposed system is summarized as follows:

   • Programming Language: Python
   • Libraries Used: OpenCV, NumPy
   • Hardware Platform: Standard Laptop with Webcam
   • Daabase: CSV File / MySQL Database
   • Operating System: Windows / Linux

   The use of lightweight tools and widely available hardware ensures that the system can be easily implemented in educational institutions without requiring specialized computational infrastructure.

4. IMPLEMENTATION AND CASE STUDY
   1. Experimental Setup

      The proposed face recognition-based automated attendance system was implemented using the Python programming language along with the OpenCV library for computer vision operations. The system was deployed and tested in a classroom environment of the Diploma in Computer Engineering department at Shri Gulabrao Deokar Polytechnic, Jalgaon.

      The experimental setup consisted of a standard laptop equipped with an Intel i5 processor and 8GB RAM, along with a built-in HD webcam used for capturing facial images in real time. The software environment included Python 3.x along with supporting libraries such as OpenCV and NumPy for image processing and numerical computation.

      A lightweight storage approach was used for maintaining attendance records. Student attendance data was stored in a CSV file, which allowed easy recording and retrieval of attendance information without requiring complex database configurations.

      The experimental setup demonstrates that the proposed system can be implemented using commonly available hardware and software resources, making it suitable for educational institutions with limited technical infrastructure.

   2. Dataset Collection

      To train the face recognition model, a dataset of facial images was collected from enrolled students under controlled classroom lighting conditions. Multiple facial samples were captured for each student to ensure reliable training and accurate recognition.

      The dataset collection process involved capturing images of students using the webcam while they faced the camera directly. These images were stored in grayscale format to simplify feature extraction and reduce computational complexity.

      The dataset characteristics are summarized as follows:

      • Number of students: 25-40
      • Images per student: 20-30 samples
      • Total dataset size: Approximately 600-800 images
      • Image format: Grayscale
      • Image resolution: 200 × 200 pixels

        Each student’s images were labelled with a unique identification number corresponding to their roll number. This labelled dataset was then used to train the LBPH face recognition model.

   3. Training Phase

      During the training phase, all collected facial images were preprocessed before being used to train the recognition model. First, the images were converted into grayscale format to simplify processing. The Haar Cascade classifier was then applied to detect facial regions within the images.

      The detected facial regions were cropped and resized to maintain uniformity across the dataset. These processed images were then used to train the Local Binary Pattern Histogram (LBPH) algorithm, which extracted unique texture features from each face.

      The training process generated feature vectors corresponding to each registered student. These vectors were stored as part of the trained model, allowing the system to recognize faces during real-time operation.

      The total training time required for the dataset was approximately 3060 seconds, depending on the number of images and processing capability of the system.

   4. Recognition and Attendance Marking

      During the recognition phase, the trained LBPH model is used to identify students in real time using live video input from the webcam. The webcam continuously captures video frames from the classroom environment, and each frame is processed by the face detection module.

      The Haar Cascade classifier detects facial regions in the captured frames. Once a face is detected, the system extracts the facial region and passes it to the LBPH recognition model. The model then compares the extracted facial features with the stored feature vectors generated during the training phase.

      If the confidence value of the prediction falls within the predefined threshold, the system identifies the corresponding student and displays the students name on the screen. At the same time, the attendance management module records the students attendance in the database along with the current date and time.

      If the detected face does not match any stored dataset entry or if the confidence value exceeds the threshold, the system labels the face as Unknown and no attendance is recorded.

      The average recognition time for each student was observed to be approximately 12 seconds, making the system suitable for real-time classroom applications.

   5. Case Study Execution

      To evaluate the practical feasibility of the proposed system, a case study was conducted in the Diploma in Computer Engineering classroom at Shri Gulabrao Deokar Polytechnic, Jalgaon.

      The system was tested during regular classroom sessions where students were seated in their normal positions. The webcam captured live video frames of the classroom, and the system attempted to detect and recognize the faces of registered students automatically.

      The testing environment included normal indoor classroom lighting conditions and both single-face and multi-face detection scenarios. During testing, the system successfully detected and recognized registered students and automatically

      recorded their attendance without requiring manual intervention.

      However, certain challenges were observed during the testing phase. For example, recognition accuracy slightly decreased when lighting conditions were poor or when a student’s face was partially covered. Additionally, minor delays were observed when multiple faces appeared simultaneously in the camera frame.

      Despite these challenges, the system demonstrated reliable performance and confirmed the feasibility of using face recognition technology for automated attendance management in educational institutions.

   6. Performance Metrics

      To evaluate the performance of the proposed attendance system, several standard biometric evaluation metrics were considered. These metrics help assess the effectiveness, reliability, and efficiency of the face recognition system.

      The following parameters were used to evaluate system performance:

      • Recognition Accuracy (%) Percentage of correctly recognized students
      • False Acceptance Rate (FAR) Percentage of unauthorized individuals incorrectly recognized as registered students
      • False Rejection Rate (FRR) Percentage of registered students incorrectly classified as unknown
      • Average Processing Time Time required to recognize a face and record attendance

        These performance metrics provide a quantitative evaluation of the system and help determine its suitability for practical classroom deployment. The results obtained from the evaluation are discussed in the next section.

5. RESULT AND DISCUSSION
   1. Performance Evaluation

      The proposed face recognition-based attendance system was evaluated in a real classroom environment to analyze its operational accuracy and feasibility. The system was tested using a dataset of 32 students, with approximately 25 facial images per student used during the training phase.

      During testing, a total of 320 recognition attempts were recorded across multiple classroom sessions. The performance of the system was evaluated using standard biometric evaluationmetrics such as Recognition Accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR), and Average Processing Time.

      Table I

      Performance Metrics of Proposed System

      Parameter	Observed Value
      Recognition Accuracy	92.4 %
      False Acceptance Rate (FAR)	3.1 %
      False Rejection Rate (FRR)	4.5 %

      Average Processing Time

      1.4 seconds

   2. Recognition Accuracy

      Recognition accuracy represents the percentage of correctly identified student faces out of the total number of recognition attempts.

      Accuracy = (Number of Correct Recognitions / Total Recognition Attempts) × 100

      The system achieved an average recognition accuracy of 92.4%, demonstrating that the LBPH algorithm performs reliably in controlled indoor classroom environments. The accuracy remained stable when lighting conditions were consistent and students faced the camera directly.

   3. False Acceptance and Rejection Analysis

      False Acceptance Rate (FAR) refers to the percentage of unauthorized faces incorrectly identified as registered students. The observed FAR of 3.1% indicates that the system maintains good identification reliability with minimal chances of proxy attendance.

      False Rejection Rate (FRR) represents the percentage of registered students incorrectly classified as unknown. The observed FRR of 4.5% occurred mainly due to factors such as:

      • Partial face occlusion
      • Sudden change in lighting conditions
      • Minor variations in head pose

        These observations indicate that environmental conditions can influence recognition performance.

   4. Processing Time Analysis

      The average time required for the system to recognize a student and record attendance was approximately 1.4 seconds. Compared to traditional manual attendance systems, which may take 510 minutes for a class of 3040 students, the proposed system significantly reduces administrative workload.

      This improvement demonstrates the efficiency advantage of automated face recognition-based attendance systems.

   5. Comparative Discussion

      Compared with traditional biometric attendance systems such as fingerprint-based devices, the proposed face recognition system offers several advantages.

      • It operates in a contactless manner, improving hygiene and convenience.
      • No additional biometric hardware devices are required.
      • The system is cost-effective and easy to deploy using standard webcams.
      • Maintenance requirements are minimal.

        Although deep learning-based models such as CNNs can provide higher accuracy, they require large datasets and high computational resources. In contrast, the LBPH algorithm provides a good balance between accuracy, computational efficiency, and implementation simplicity, making it suitable for diploma-level educational institutions.

   6. Observations from Case Study

      During real-time classroom testing, several practical observations were recorded.

      • The system performed effectively under normal indoor lighting conditions
      • Recognition accuracy was highest when students faced the camera directly
      • Minor delays were observed when multiple faces appeared simultaneously
      • Recognition accuracy slightly decreases under poor lighting conditions

   Overall, the system demonstrated practical feasibility for real classroom deployment and successfully automated the attendance recording process.

6. CONCLUSION

This paper presented the design, implementation, and case study evaluation of a face recognition-based automated attendance system using the Local Binary Pattern Histogram (LBPH) algorithm. The proposed system was developed using Python and OpenCV and tested in a classroom environment of the Diploma in Computer Engineering department at Shri Gulabrao Deokar Polytechnic, Jalgaon.

The experimental results demonstrated that the system achieved a recognition accuracy of approximately 92.4% with relatively low False Acceptance Rate and False Rejection Rate. The average processing time of around 1.4 seconds per student indicates that the system significantly reduces the time required for manual attendance marking while maintaining reliable performance.

The case study confirmed that lightweight machine learning algorithms such as LBPH can be effectively implemented for real-time applications in educational institutions without requiring high-end computational infrastructure. The proposed system provides a contactless, cost-effective, and efficient solution for attendance management.

However, the system performance may be affected under conditions such as poor lighting, face occlusion, or significant pose variations. Future improvements may include the integration of advanced deep learning-based face recognition

models, cloud-based attendance management systems, and additional security mechanisms such as mask detection and anti-spoofing techniques.

Overall, the study demonstrates that face recognition technology has strong potential to modernize traditional attendance systems and improve administrative efficiency in academic institutions.

ACKNOWLEDGMENT

The authors would like to thank the Department of Computer Engineering, Shri Gulabrao Deokar Polytechnic, Jalgaon, for their support in carrying out this research work.

REFERENCES

1. M. Turk and A. Pentland, Eigenfaces for Recognition, Journal of Cognitive Neuroscience, vol. 3, no. 1, pp. 7186, 1991.
2. P. N. Belhumeur, J. P. Hespanha, and D. J. Kriegman, Eigenfaces vs. Fisherfaces: Recognition Using Class Specific Linear Projection, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 7, pp. 711720, 1997.
3. T. Ahonen, A. Hadid, and M. Pietikäinen, Face Description with Local Binary Patterns: Application to Face Recognition, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 12, pp. 2037 2041, 2006.
4. G. Bradski and A. Kaehler, Learning OpenCV: Computer Vision with the OpenCV Library, OReilly Media, 2008.
5. R. Szeliski, Computer Vision: Algorithms and Applications, Springer, 2011.
6. W. Zhao, R. Chellappa, P. J. Phillips, and A. Rosenfeld, Face Recognition: A Literature Survey, ACM Computing Surveys, vol. 35, no. 4, pp. 399458, 2003.
7. S. Z. Li and A. K. Jain, Handbook of Face Recognition, Springer, 2011.
8. H. Moon, P. J. Phillips, Computational and Performance Aspects of PCA-Based Face Recognition Algorithms, Perception, vol. 30, no. 3, pp. 303321, 2001.
9. OpenCV Documentation, Open Source Computer Vision Library,

   Available: https://opencv.org

10. I. Goodfellow, Y. Bengio, and A. Courville, Deep Learning, MIT Press, 2016.