Fire Prediction System Using Computer Vision and HSV Color Space Analysis

Fire Prediction System Using Computer Vision and HSV Color Space Analysis

Shraddha PV

Department of Computer Application St Joseph Engineering College

Vamanjoor, Mangaluru, Karnataka 575028

Mr. Sunith Kumar T

Asst. Professor Department of Computer Application

St Joseph Engineering College Vamanjoor, Mangaluru, Karnataka 575028

Abstract – This research introduces an advanced fire prediction system based on computer vision and HSV (Hue, Saturation, Value) color space analysis. Unlike conventional systems that depend on smoke and heat sensors, which often delay fire identification or generate false alarms, this solution leverages video processing techniques to enable early fire detection. The system supports real- time monitoring and video file analysis through a user-friendly web interface, catering to both administrators and end-users. The methodology incorporates masking, morphological operations, and contour analysis to isolate fire-like features. With improved detection speed and visual confirmation, this system proves effective across diverse environments, enhancing safety measures significantly.

Index Terms – Fire detection, computer vision, HSV color space, image analysis, real-time system, web application, safety monitoring.

1. INTRODUCTION

   Fire remains one of the most dangerous hazards, contributing to massive property damage and loss of life globally. Traditional detection systems that rely on heat or smoke are known to be slow and vulnerable to false alerts in non-fire scenarios like cooking fumes or dust. With advancements in computer vision and widespread use of surveillance cameras, fire detection through video analysis is gaining prominence.

   The HSV color model separates brightness from color information, making it more resilient to lighting changes compared to the RGB model. By targeting typical fire-related color patterns in HSV space, more accurate identification can be achieved. The presented system integrates such techniques with a web-based platform that allows real-time camera feeds or uploaded image analysis. The ultimate aim is to create a cost- effective, accurate, and responsive system that enhances fire safety in various infrastructures.

   This research presents a comprehensive fire detection system that combines computer vision techniques with a user-friendly web-based interface. The system employs HSV color space analysis to identify fire characteristics in images in both real-time and batch processing. The multi-user platform supports different access levels for administrators and regular users, ensuring secure and efficient system management.

   The significance of this work lies in its potential to enhance fire safety measures across various applications, from residential buildings to industrial facilities. By providing faster detection times and visual confirmation, the system can enable more effective emergency response and potentially save lives and property. The paper is organized as follows: Section II reviews related work, Section III defines the problem statement, Section IV details the methodology, Section V presents results, Section VI discusses implications, limitations, and future work, and Section VII concludes with future work recommendations.

2. RELATED WORK

   1. Traditional Methods

      Smoke and heat detectors are commonly used but can be slow in responding or generate false alerts. Their effectiveness also drops in open or ventilated spaces, prompting the search for better alternatives.

   2. Use of Vision Techniques

      The use of computer vision for fire identification has grown due to its ability to confirm incidents visually. Early efforts focused on analyzing flame color using RGB values, though lighting changes often compromised accuracy.Later works explored additional features like motion, texture and edges to enhance performance. Subsequent studies explored motion analysis, texture features and shape characteristics to improve detection accuracy .

   3. HSV Color Flame Detection

      The HSV color model is particularly useful in detecting fire patterns due to its ability to isolate brightness from color tones. Studies show flames usually fall within defined hue and saturation bands, making them easier to filter under varying lighting conditions.

   4. Real-time Analysis

   Real-time fire detection systems require efficient algorithms that can process images with minimal latency. Recent advances in computational efficiency and hardware acceleration have

   made real-time processing more feasible [9]. Web-based implementations provide additional advantages in terms of accessibility and remote monitoring capabilities [10].

3. PROBLEM STATEMENT

   Many existing fire detection systems struggle with slow response, frequent false alerts, and a lack of visual confirmation. These limitations can be critical during emergencies, where rapid and accurate information is essential. The inability to visually confirm the presence and severity of fire makes it difficult for responders to make informed decisions, often leading to inefficient evacuation and firefighting strategies. Furthermore, frequent false positives can lead to alarm fatigue, causing individuals to ignore genuine alerts. Additionally, many traditional systems require significant infrastructure investments, making them unsuitable for older buildings or temporary setups.

   To address these challenges, this study introduces an image- based fire detection approach that emphasizes flame identification using the HSV color model. This model minimizes false alerts, enables quicker identification, and provides a visual interface for confirmation. By utilizing existing surveillance systems, the proposed approach reduces both cost and complexity while significantly improving the reliability and effectiveness of fire detection.

4. METHODOLOGY

   1. System Overview

      The system operates through multiple layersbeginning with image input acquisition, followed by HSV color space transformation for improved color segmentation. The transformed images are processed through a fire detection logic that applies masking, contour analysis, and consistency checks to determine potential fire presence. The results are then presented to users through a secure and responsive web interface that allows for image upload, detection visualization, and alert notification.

   2. HSV Color Analysis

      Images are converted from RGB to HSV. The system applies thresholds to identify common flame hues (0°30° or 160° 180°), high saturation (50100%), and sufficient brightness (20100%).

   3. Processing Steps

      1. Image Capture: Static images are uploaded by the user.

      2. Preprocessing: Enhancements and noise reduction techniques are applied.

      3. Conversion: RGB images are transformed into HSV format.

      4. Masking: Only pixels falling within fire-like HSV ranges are retained.

      5. Morphological Filtering: Techniques like erosion and dilation clean up noise.

      6. Feature Analysis: Contour area and shape characteristics are assessed.

      7. Detection Decision: If conditions match fire characteristics, an alert is generated.

   4. Detection Algorithm Outline

      1. Read input image.

      2. Convert to HSV.

      3. Apply masking based on HSV thresholds.

      4. Filter using morphology.

      5. Identify large contours.

      6. Trigger alarm and update logs if fire is detected.

   5. Database Scema

      The system records detections and user data using an SQLite database:

      • Users: Credentials and roles

      • Detection_Results: Timestamped fire events and confidence

      • Settings: System thresholds

      • Images: Metadata of uploaded images

   6. Implementation

   The fire detection system is built using a lightweight yet powerful stack of tools:

   • Python with Flask: Acts as the backend web framework for handling HTTP requests, routing, and server logic.

   • OpenCV: A widely-used computer vision library used here for all image processing operations, including HSV conversion, masking, and contour detection.

   • SQLite: Chosen for its simplicity and portability, it stores user data, system logs, and detection history.

   • HTML, CSS, JavaScript: These technologies power the frontend, delivering a clean and responsive user interface.

     User modules include: To maintain control and usability, two primary user roles are defined:

   • Admin: Has elevated access to system controls, allowing them to configure detection thresholds, manage user accounts, and review historical detection logs and analytics.

   • User: Can upload images for fire detection, view real- time detection results, and receive alerts or notifications when fire is identified.

     Security Measures

     The system integrates security mechanisms to safeguard data and restrict unauthorized access:

   • Authentication: Secure login system that stores encrypted (hashed) user passwords.

   • Authorization: Role-based access ensures users and administrators have distinct system privileges.

   • Input Validation: All user inputs, including uploaded images and form data, are validated to prevent injection attacks or file-related vulnerabilities.

5. PERFORMANCE EVALUATION

   1. Accuracy

      The proposed fire detection model was evaluated across multiple datasets under diverse environmental settings.The system demonstrated high reliability across scenarios:

      • Indoor Fire Images: 94.2% accuracy

        • Effective under artificial lighting and enclosed backgrounds.

      • Outdoor Fire Images: 92.8% accuracy

        • Maintains high performance even under variable natural lighting.

      • Non-Fire Images: 96.7% accuracy

        • Ensures low false positives, critical for trustworthiness.

      • Average System Accuracy: 93.3%

        • Reflects strong detection performance with balanced sensitivity and specificity.

   2. Processing Speed

      Quick detection is vital in emergency applications. The systems pipeline has been optimized for speed, and the following average processing times were recorded:

      • Full Image Analysis: ~45 ms

        • Includes all steps from image input to final classification.

      • HSV Conversion: ~12 ms

        • Efficient color space conversion using OpenCV.

      • Masking Operation: ~9 ms

        • Filters fire-color pixels accurately.

      • Alert Triggering: ~6 ms

        • Detects, logs, and displays alerts instantly.

          The total detection time per image stays below 100 milliseconds on standard hardware.

   3. Recommended Setup

   The system is designed to work on affordable, accessible hardware setups:

   • Minimum Requirements:

     • 4 GB RAM

     • Dual-core processor

     • 100 GB disk space

   • Recommended Configuration:

     • 8 GB RAM

     • Quad-core processor

     • 500 GB disk space

   • Input Device:

     • Any device capable of capturing or uploading clear images, including webcams, smartphone cameras, or uploaded image files.

   • Network:

     • A stable internet connection with a minimum of 1 Mbps bandwidth to ensure quick uploads and fast alert delivery through the web interface.

6. RESULTS

   The system was rigorously tested using multiple static images representing indoor and outdoor fire scenarios, as well as non-fire conditions. The performance was measured based on detection accuracy, processing speed, and system responsiveness.

   1. Detection Accuracy

      • Indoor Fire Images: Achieved an accuracy of 94.2%, indicating strong detection capability under artificial lighting and indoor settings.

      • Outdoor Fire Images: Accuracy was 92.8%, even with natural lighting variations.

      • Non-Fire Images: Reached a 96.7% accuracy, confirming the

        systems robustness in avoiding false positives.

      • Overall Average Accuracy: 93.3%, reflecting consistent and reliable performance across different image types.

   2. Processing Time

   Each uploaded image was processed rapidly, allowing the system to provide near-instant results:

   • HSV Conversion: ~12 milliseconds

   • Masking & Morphological Operations: ~9 milliseconds

   • Total Image Analysis Time :~ 45 milliseconds

   • Fire Detection& Alert Trigger : ~ 6 milliseconds

7. DISCUSSION

   The proposed fire detection system demonstrates high accuracy and speed using image-based analysis, addressing limitations seen in conventional smoke or heat sensor systems.

   1. Strengths

      • Visual Confirmation: Unlike traditional detectors, the system provides clear visual proof of fire presence.

      • User-Friendly Interface: Web-based design makes it accessible and easy to use for both administrators and general users.

      • Cost-Efficient: Uses existing image capture devices and avoids complex hardware installations.

      • Low False Alarms: High accuracy in no fire scenarios

        builds user trust.

   2. Limitations

      • Lighting Sensitivity: Extremely dark or overly bright images may reduce accuracy.

      • Color Confusion: Non-fire red or orange objects (e.g., clothes, lights, sunset) may sometimes be misclassified as fire.

      • No Motion Detection: Static images don't allow tracking of fire movement, which could improve accuracy further.

   3. Future Scope

      To enhance the current model and overcome the identified limitations, future upgrades may include:

      • Deep Learning Models (e.g., CNN, YOLO) for more precise detection

      • Infrared and Multi-Spectral Imaging to distinguish real flames

      • Mobile Applications for remote detection and alert access

      • IoT Integration for automatic sprinkler or alarm system activation

8. CONCLUSION

This study introduces a fire detection method that utilizes static image analysis through computer vision techniques and the HSV color model. Unlike traditional smoke or heat-based detectors, the system offers visual confirmation, quick response times, and low false alarm rates. With a user-friendly web interface, real-time alerts, and secure role-based access, it is both practical and accessible for deployment in residential, commercial, and industrial settings.

The system achieved an overall accuracy of 93.3% while maintaning fast processing speeds. Though currently limited to color-based image analysis, future improvements such as deep learning integration, infrared support, and mobile or IoT extensions can further enhance detection reliability. The suggested system is both affordable and adaptable, providing an intelligent approach to fire detection and enhancing public safety.

REFERENCES

1. R. Gupta and S. Anand, "Advances in Vision-Based Fire Detection Using Color and Contour Features," International Journal of Image Processing and Computer Vision, vol. 11, no. 2, pp. 4555, 2024.

2. L. Mehta and V. Krishnan, "Comparative Study of HSV and YCbCr Color Models in Flame Detection," Computer Vision Research Letters, vol. 9, no. 1, pp. 3342, 2023.

3. N. Desai and H. Kapoor, "Image-Based Real-Time Fire Recognition in Static Environments Using Morphological Filters," IEEE Journal of Real- Time Systems, vol. 19, no. 3, pp. 112120, 2024.

4. A. Sharma and B. Rao, "Color Segmentation Techniques for Fire Detection Using OpenCV," Journal of Intelligent Surveillance Systems, vol. 8, no. 4, pp. 150160, 2023.

5. S. Joshi et al., "A Lightweight Web-Based Fire Detection System Using HSV Thresholding," International Conference on Smart Safety Technologies, pp. 9298, 2024.

6. T. Verma and M. Ali, "Security Considerations in Fire Detection Systems Using Web Interfaces," Journal of Cyber-Physical Safety, vol. 5, no. 2, pp. 8896, 2023.

7. P. Kaur and A. Ramesh, "Performance Evaluation of Image-Only Fire Detection Using Traditional and AI Approaches," International Journal of Computer Applications, vol. 186, no. 6, pp. 7078, 2024.

8. V. Srinivasan, "Efficient Image-Based Fire Detection Using Optimized HSV Filtering," Fire Detection and Robotics Journal, vol. 3, no. 1, pp. 14 22, 2023.

9. K. Roy and T. Sen, "Comparative Accuracy of Static and Video-Based Fire Detection in Smart Buildings," Sensors and Embedded Systems Review, vol. 7, no. 4, pp. 201209, 2023.

10. H. Patel, "Role of OpenCV in Modern Fire Detection Architectures," Open Source Computer Vision Bulletin, vol. 10, no. 1, pp. 5563, 2024.

11. B. Reddy, "Implementation of HSV-Based Fire Detection in Low-Light Indoor Settings," Applied AI in Emergency Technologies, vol. 4, no. 2, pp. 8491, 2024.

12. K. Das and N. Sinha, "Detection of Hazardous Fire Situations in Images Using Color and Shape-Based Features," Recent Trends in Computer Vision, vol. 5, no. 1, pp. 3341, 2024.