Sina Shiri

Python Developer

Electronic Engineer

Writer

Cryptocurrency Trader

Sina Shiri

Python Developer

Electronic Engineer

Writer

Cryptocurrency Trader

Blog Post

Fire and Smoke Detector using Image Processing

April 1, 2020 Projects
Fire and Smoke Detector using Image Processing

This system has the ability to detect fire and smoke in open and covered spaces and can analyze the environment in full time and announce the smallest possibility of a fire or the start of a fire.
The purpose of this system is to solve the problem of limiting the use of fire detection sensors in open environments, as well as increasing the detection distance to several hundred meters and reducing the detection time.
The places that are especially candidates for using this system are:

  • Refineries and petrochemicals
  • Power plants
  • All factories (whether open space or inside sheds and roofed space)
  • The forests
  • Airports and runways
  • Military areas
  • Sensitive security centers
  • Departments, universities and educational centers
  • Business centers
Inside of Explosion-Proof version of hardware

The most important part of the hardware is the processing unit. Because a comprehensive and uniform hardware should be selected according to the basic needs of the system.
Basic system requirements:

  • No a large distance between the camera and the processing unit
  • Access to suitable hardware for heavy processing
  • Access to digital inputs and outputs (not available on personal computers)
  • Light weight due to installation next to the camera and not occupying much space
  • Low power consumption (possibility of connecting to a solar cell)

The Raspberry Pi has been selected to have the ability to carry out heavy calculations of the system’s neural network at the same time as the mentioned items are fulfilled.

Schematic of Device

For Detect Fire and Smoke, In the 1st step a complex image processing algorithm implement by OpenCV and C++ that track a places of picture is fire and smoke candidate. You can see two video of real fire test that All process runs in Raspberry pi and Alarms show in windows via local network.

In the 2nd step a candidate frame send to a Neural Network for 2-Step detection of Fire and Smoke:

The device is designed in two model:

  • Explosion-Proof model
  • CCTV model

In Explosion version the system is placed in an explosion-proof enclosure for use in critical and dangerous situations (in terms of safety issues), and the only connection between the system and the outside environment is made by the power cable and network cable.

The CCTV version of the system includes all the mentioned features and by using the GSM module, it has the ability to establish a stable connection with the mobile phone network and can be used in any environment.

The Device has a Windows based software for monitoring and manage all events, alarms and live-streams in local networks. A screen of software is below that each section is explained in details.

Windows Software of Project

Additionally,  for full access to the device, an application environment has been designed so that the user can make all the settings and at the same time it is possible to monitor the system. The system considers a http virtual server separately from the network at all times and the user can perform the mentioned actions in this way. 
Some of pages of this control panel are in below photo galley.

A simulated video of how the system works is available below:

In Next Video you can see a indoor test of Device that detect Fire in less that 3 seconds.

Feutures
  • Analyze Environment by Image Processing and Custom Neural Network and Detects in less than 5 seconds
  • Raises alarm through: SMS, Website(Network, GSM), Hardware IO, Local Web-View, Local Windows Software
  • No limits in Distance between Devices
  • Complex Software
  • Has Explosion-Proof Model
  • Has Web-View Control Panel for Apply any settings
  • Live stream camera output in Local Network (web-View)
  • No limits in Number of Devices
Software
  • Languages: Python, C++
  • Modules: tensorflow, PyQt, http.server, RPi.GPIO, picamera, pyodbc, OpenCV
  • PCB: Altium Designer
Hardware
  • Raspberry Pi
  • Pi Camera
  • Custom PCB Board with SIM800
  • Custom Explosion-Proof Design

Related links of Project:

Tags:
Write a comment