Government Post Graduate College Rajanpur
SOFTWARE REQUIREMENTS SPECIFICATION
(SRS DOCUMENT)
for
<Fire Detection System >
By
Muhammad Adeel Asif
GCBR-20-24-127
Supervisor
Mr. Muhammad Rawal Gulzar
Bachelor of Science in Information Technology
Table of Contents
Table of Contents............................................................................................................................................................2
Revision History.............................................................................................................................................................1
Application Evaluation History......................................................................................................................................2
1
1 Introduction..................................................................................................................................................................3
1.1 Purpose......................................................................................................................................................3
1.2 Scope.........................................................................................................................................................3
2 Overall description.......................................................................................................................................................3
2.1 Product perspective...................................................................................................................................3
2.2 Operating environment.............................................................................................................................3
2.3 Design and implementation constraints....................................................................................................3
3 Requirement identifying technique..............................................................................................................................4
3.1 Use case diagram......................................................................................................................................4
3.2 Use case description..................................................................................................................................4
4 Functional Requirements.............................................................................................................................................6
4.1 Functional Requirement X........................................................................................................................6
5 Non-Functional Requirements.....................................................................................................................................7
5.1 Usability....................................................................................................................................................7
5.2 Performance..............................................................................................................................................8
References.......................................................................................................................................................................8
2
Revision History
Name Date Reason for changes Version
1
Application Evaluation History
Comments (by committee) Action Taken
*include the ones given at scope time both in doc and
presentation
Supervised by
< Mr. Muhammad Rawal Gulzar>
Signature
2
1. Introduction
This section provides an overview of the Software Requirements Specification (SRS) document for the
Forest Fire Detection System, detailing its purpose, scope, and key features.
1.1 Purpose
The purpose of this document is to specify the functional and non-functional requirements for the Forest
Fire Detection System, a machine learning-based application designed to identify and classify forest
fires using image data. The document is intended for stakeholders, developers, and testers to ensure a
clear understanding of the system’s functionality, design, and goals. This SRS serves as a reference
throughout the development lifecycle, ensuring that all requirements are met and deviations are
minimized.
The product aims to assist environmental agencies, firefighting teams, and decision-makers by providing
an efficient, automated solution for early detection and classification of forest fires, enabling quicker
responses and minimizing damage.
1.2 Scope
The Forest Fire Detection System is a software application that uses machine learning models,
specifically Convolutional Neural Networks (CNN), to analyze image datasets and detect patterns
indicative of fire or smoke. The system operates by:
Processing Visual Data: Images are preprocessed and normalized to improve detection accuracy.
Training a Model: The software uses labeled datasets to train a CNN that can differentiate between fire,
smoke, and non-fire images.
Providing Predictions: Users can upload images, and the system outputs classifications (e.g., "Fire",
"Smoke", "No Fire") along with confidence scores.
Visualizing Results: Detected regions of fire or smoke are highlighted for better interpretability.
The software supports agencies in proactive disaster management by providing:
Early detection of forest fires.
Reliable classification results to prioritize resources.
A scalable design to handle large datasets and real-time processing.
2. Overall Description
This section provides a comprehensive overview of the Forest Fire Detection System, including its
context, operating environment, and any constraints influencing its design and implementation.
2.1 Product Perspective
3
The Forest Fire Detection System is an entirely new product developed to leverage advancements in
machine learning and computer vision for real-time forest fire detection. It does not replace or build on
an existing system but addresses a growing demand for automated solutions in environmental monitoring
and disaster management.
This system integrates machine learning techniques with intuitive user interfaces to deliver accurate and
interpretable results. By processing image datasets and identifying fire-related patterns, it provides timely
insights, aiding in proactive disaster response. Its modular design ensures scalability, making it suitable
for expanding functionalities, such as incorporating satellite imagery or integrating IoT sensor data in
future iterations.
2.2 Operating Environment
The system is designed to operate in diverse environments, accommodating various user needs and
hardware configurations. Below are the key aspects of the operating environment:
Hardware Platform: The system requires a workstation or server with GPU support for model
training and predictions. Minimum recommended specifications:
o Processor: Intel i5 or equivalent.
o RAM: 8 GB (16 GB for training large models).
o GPU: NVIDIA GTX 1050 or higher with CUDA support.
Operating Systems:
o Windows 10/11.
o macOS (latest versions).
o Linux distributions (Ubuntu 20.04 LTS or higher).
Web Browsers:
o Google Chrome (all versions).
o Mozilla Firefox (version 12 and above).
o Microsoft Edge (latest versions).
Database and Servers:
o Cloud databases (e.g., AWS RDS, Google Cloud SQL).
o Optional on-premise database hosting for sensitive datasets.
The system supports users located in various geographical regions, ensuring accessibility through cloud-
hosted services and remote deployment options.
2.3 Design and Implementation Constraints
The design and implementation of the system are subject to the following constraints:
Programming Language:
o The system shall use Python as the primary language due to its robust libraries for machine learning
(e.g., TensorFlow, PyTorch) and image processing (e.g., OpenCV, PIL).
Code Libraries:
o Pre-trained models and libraries like TensorFlow/Keras will be employed to accelerate development
while maintaining state-of-the-art performance. 4
Hardware Limitations:
o Training large datasets will require GPU support. Systems without GPUs may experience reduced
performance or limited training capabilities.
Deployment Constraints:
o Cloud deployment options will rely on platforms such as AWS, Azure, or Google Cloud, which may
incur usage costs.
Security and Privacy:
o User-uploaded images must be processed securely, adhering to data privacy regulations like GDPR.
Examples:
CO-1: The system shall utilize TensorFlow/Keras as the machine learning framework.
CO-2: The application shall support GPU acceleration for efficient model training and predictions.
3. Requirement Identifying Technique
To identify the functional requirements for the Forest Fire Detection System, the use case technique
has been employed. This approach is well-suited for interactive end-user applications like this system,
where users upload images to receive predictions. Use cases provide a clear, user-centric view of the
system's behavior and interaction, ensuring all functional aspects are addressed.
3.1 Use Case Diagram
The following use case diagram illustrates the interaction between the user and the system:
1. Actors:
o Primary Actor: User (e.g., environmental agency staff, researchers).
o Secondary Actors: Cloud storage, Machine Learning model.
2. Use Cases:
o Upload image for analysis.
o Preprocess image data.
o Classify image (e.g., fire, smoke, no fire).
o Visualize prediction results.
o Provide feedback on predictions.
(Include a detailed diagram created in a UML tool or manually. Let me know if you’d like me to generate
this.)
3.2 Use Case Description
Below is an example use case template tailore
d for
the
Forest
5 Fire
Detecti
on
System
:
Use Case ID UC-1
Use Case Name Upload and Analyze Image
Actors Primary Actor: User; Secondary Actors: Machine Learning Model, Cloud Storage
A user uploads an image to the system, which preprocesses the image, classifies it using a trained
Description machine learning model, and displays the classification result (e.g., fire, smoke, no fire) with a
confidence score.
Trigger User uploads an image for analysis.
Preconditions PRE-1: The user has access to the system.
PRE-2: The system is operational and connected to the model and database.
Postconditions POST-1: Prediction results are displayed.
POST-2: Processed image data is stored in the system logs for future analysis.
Normal Flow 1. User uploads an image through the interface.
2. System preprocesses the image (resize, normalize, etc.).
3. System classifies the image using the trained CNN model.
4. System outputs the classification result with confidence scores.
5. User views the results on the interface.
Alternative
1. User uploads an unsupported image format.
Flows
- System displays an error and requests a valid file format.
Exceptions 1. Model is offline:
- System displays an error message and notifies the user to retry later.
Business Rules BR-1: Only images in JPEG or PNG format are accepted.
BR-2: System processes one image at a time.
Assumptions 1. Users have high-quality image data suitable for processing.
4 Functional Requirements
This section outlines the functional requirements of the Forest Fire Detection System, organized by key
features and the associated software functionalities. Each functional requirement is labeled uniquely and
defined in detail, ensuring a clear understanding of the system's capabilities and responses.
4.1 Image Data Preprocessing
FR 4.1.1: The system shall preprocess image data to standardize input dimensions to 64x64 pixels, ensuring
compatibility with the trained model.
Rationale: Standardization enhances model accuracy and processing efficiency.
Dependencies: Relies on input data from cameras or user-uploaded images.
Error Handling: If input data is corrupted or incompatible, the system shall display a user-friendly error
message.
FR 4.1.2: The system shall normalize image pixel values to a [0.0, 1.0] range for consistent model input.
Rationale: Normalization optimizes model performance by standardizing input features.
4.2 Fire and Smoke Detection 6
FR 4.2.1: The system shall classify images into three categories: "Fire," "Smoke," and "Not Fire."
Rationale: Accurate detection is critical for emergency response.
Source: Machine learning model trained on labeled datasets.
Error Handling: For ambiguous or unclear images, the system shall provide a warning stating, "Classification
confidence is low."
FR 4.2.2: The system shall achieve a minimum classification accuracy of 95% on the test dataset.
Rationale: High accuracy ensures reliable detection in real-world scenarios.
4.3 Visualization and Reporting
FR 4.3.1: The system shall visually display the input image along with the detected category and confidence
score.
Rationale: Providing visual feedback enhances user trust and understanding.
FR 4.3.2: The system shall generate a detailed log report for every prediction, including timestamp, category,
confidence, and input image metadata.
Rationale: Logs enable tracking and analysis of system performance.
4.4 User Notifications
FR 4.4.1: The system shall notify the user via visual and audio alerts if fire or smoke is detected.
Rationale: Prompt notifications are critical for timely action.
Dependencies: Requires integration with audio and display devices.
FR 4.4.2: The system shall provide actionable suggestions based on the detected category, such as "Evacuate
immediately" for fire.
Error Handling: If the system fails to generate actionable advice, it shall display a generic safety warning.
4.5 System Reliability and Error Management
FR 4.5.1: The system shall retry preprocessing and classification up to three times if an error occurs during the
process.
Rationale: Retry mechanisms reduce the impact of transient issues.
FR 4.5.2: The system shall log all errors and provide a detailed error report to the administrator.
Rationale: Detailed error logs facilitate debugging and maintenance.
Table 6 Show the functional requirement template
Identifier Title Requirement Source Rationale Business Dependencie Priority
Rule s
FR-001 Image The system System Consistent Image None High
Standardization shall Architect image inputs must
standardize dimensions always 7
all input improve conform to
images to model the
64x64 pixels accuracy and expected
before reduce dimension
processing to preprocessin for model
ensure g errors. processing.
compatibility
with the
model.
FR-002 Image The system System Improves All inputs FR-001 High
Normalization shall Architect performance must be
normalize by ensuring normalized
pixel values inputs fall before
to a [0.0, 1.0] within a feeding
range for consistent into the
consistent range. neural
model input. network.
FR-003 Fire The system Machine Accurate Must FR-001, FR- High
Classification shall classify Learning classification maintain 002
images into Model enables 95%
three timely accuracy as
categories: detection and per
Fire, Smoke, response. validation
and Not Fire, dataset
based on the results.
trained
model.
FR-004 Display After Product Transparent Outputs FR-003 High
Predictions classification, Owner predictions must
the system enhance user always
shall display trust and include the
the input usability. predicted
image, category
predicted and
category, and confidence
confidence score.
score to the
user.
FR-005 Notification If fire or Safety Critical for Alerts must FR-003 High
Alerts smoke is Officer initiating conform to
detected, the timely local safety
system shall response to regulations.
send visual potential
and audio disasters.
alerts to
notify the
user
immediately.
FR-006 Logging and The system Product Enables Logs must None Medium
Reporting shall log Owner performance comply
every monitoring with data
prediction and security
with the debugging. policies.
input image
metadata,
timestamp, 8
and
confidence
score.
FR-007 Error Retry If an error System Reduces the All retries FR-001, FR- Medium
occurs during Architect likelihood of must be 003
image transient logged for
preprocessin issues auditing
g or affecting purposes.
classification, usability.
the system
shall retry the
process up to
three times
before
generating an
error report.
FR-008 Actionable The system Product Provides Messages FR-005 Medium
Advice shall provide Owner, users with must align
Generation actionable Safety clear with
suggestions Team instructions standard
based on to act on. emergency
detected procedures.
categories
(e.g.,
'Evacuate
immediately'
for fire, 'Stay
alert' for
smoke).
5 Non-Functional Requirements
This section details the non-functional requirements of the Forest Fire Detection System, ensuring high-quality
performance, usability, reliability, and maintainability. These requirements are specific, measurable, and
verifiable.
5.1 Usability
USE-1: The system shall provide an intuitive graphical user interface (GUI) that allows users to upload images
and view detection results with minimal effort.
USE-2: The system shall display clear error messages and guidance for resolving input-related issues, such as
unsupported file formats.
USE-3: The system shall enable users to access prediction logs and historical detection data within three clicks.
USE-4: The system interface shall be accessible to users with visual impairments by supporting screen readers
and high-contrast display modes.
Rationale: These usability requirements aim to make the system easy to learn, efficient, and inclusive.
5.2 Performance
9
PERF-1: The system shall process and classify an image within 2 seconds under normal operating conditions.
PERF-2: The system shall handle simultaneous inputs from up to 10 users without significant degradation in
performance.
PERF-3: The classification accuracy of the system shall be no less than 95% under validated test conditions.
Rationale: High performance ensures timely and accurate detection, essential for emergency responses.
5.3 Reliability
REL-1: The system shall maintain an uptime of 99.9% over a monthly reporting period.
REL-2: The system shall provide fallback mechanisms to log errors and notify administrators in case of
component failures.
Rationale: Ensuring system availability and robustness is critical for its reliability.
5.4 Security
SEC-1: The system shall encrypt all input data and logs to prevent unauthorized access or data leaks.
SEC-2: User authentication shall be required for accessing logs or administrative features.
Rationale: These security measures protect sensitive data and maintain system integrity.
5.5 Maintainability
MAINT-1: The system shall allow modular updates to the machine learning model without impacting other
components.
MAINT-2: The system codebase shall include detailed inline documentation to facilitate debugging and future
enhancements.
Rationale: Maintainability ensures long-term adaptability and cost-effective system management.
References
Here are references to relevant papers on forest fire detection systems:
1. Forest Fire Detecting System by Bin Wan (2010)
This system integrates video compression, infrared imaging, visible color recognition, and remote
sensing for automated fire detection and early warning.
2. Forest Fire Detection System by Cao Jian and Li Guozhong (2015)
Describes a system that uses temperature, video, and light signals collected via sensors to detect forest
fires.
3. Forest Fire Detection System and Method by Zeng Hengdong (2015)
Utilizes thermal infrared imagers and video servers to detect and analyze fire outbreaks, focusing on
quick response times.
4. Forest Fire Detection System by M. Gormley Andrew and Chatola Fanny (2024)
Integrates IoT, machine learning, and weather APIs to improve fire detection accuracy, focusing on
environmental conservation and public safety.
5. Forest Fire Detection System Using IoT by Pavitra N, Sania Khan, Seeksha Jain, Anusha Mn, and Y
Pavan Kalyan (2020)
Proposes a fire detection system leveraging IoT devices and image processing, including GPS10 tracking
and SMS-based alerts.
