FireGuard

Reimagining how we tackle forest fires using IoT and Machine Learning

Comment 1

Inspiration

The turn of the new year was accompanied by the news of fire ravaging most of Australia. In total, an estimated of over 1 billion animals are estimated to have died along with over 25 million acres of land. The effects of the fire left a lasting impression on our views of how technology can be used for good and helped us discover a newfound sense of purpose. We sought to use our technological prowess to ensure that an issue like this doesn't become a common occurrence. Fireguard is an Internet of Things solution designed to detect and prevent forest fires at the instant they arise. By placing this tool in different places in the forest, we are able to monitor important vitals about the environment and detect sudden changes in temperature, CO2 levels, atmospheric pressure and volume of total organic compounds. We then retrieved this information and created a unique web user interface that responded to the real-time changes in data to relay information across all individuals over watching the situation.

The scope of our hackathon project was:

  1. A hardware device that monitors environment conditions
  2. A dashboard that allows one to monitor the status of all the beacons and alert EMS if they detect sudden changes in environment conditions
  3. Geofencing integration to alert others in the area to leave before the situation gets out of hand

Fireguard Tech Stack

Machine Learning: (time series forecasting) The machine learning aspect one which we believed could play a key role in our solution yet needed a strong use case for. Initially, we wanted to use an attached camera and the google cloud vision API to confirm fires and reduce the number of false positives created by this data. After some thought, we realized that this would be a misuse of the data we collect and decided to use time series forecasting. This model works by detecting sudden changes in data in order to detect anomalies. We used a seasonality change model to account for large variations slowly brought upon over large amounts of time, such as season changes. In order to test, we simulated a fire on our device and tracked the data in a csv file. That data was passed into the model and the anomalies were rendered on our alert dashboard.

Web View: The information on Google Firebase is then retrieved and displayed on a web app user interface updated in real-time taking advantage of the Google Maps API. The web app is built using the Vue.js framework to process the real-time data. Other areas of the web app such as the front-end design are built using the HTML5, CSS3, and Javascript. Additional features were implemented using Bootstrap, and ApexGraphs.

Hardware: To accomplish the goals of FireGuard, we chose to use the NodeMCU which is an IoT (Internet of Things) enabled microcontroller built on top of the Arduino framework. We attached several sensors to the board to help us gather the vital environmental data needed to be sent to Google Cloud's Firebase Database. Some of these sensors included the DHT11 Temperature and Humidity Sensor, CCS811 Air Quality Sensor, VEML 6070 UV Index Sensor, SparkFun Soil Moisture Sensor, and the TCS34725 RGB Sensor. This combination of sensors and an IoT enabled microcontroller enables us to gather and upload data in a lightweight and efficient format, reducing latency and overhead in emergency situations.

Data Base: The data is received by the NodeMCU microcontroller is sent to a real-time Google Firebase database where it is categorized by sensor and type of data.

Impacts of Forest Fires

As mentioned above, over a billion animals have died and 25 million acres of flora and fauna have gone alongside through the Australian fires. However, besides these impacts there are quite a few additional environmental impacts that come with forest fires including the use of chemicals in firefighting and heavy smog can lead to further levels of poisoning in local water systems and such.Increased carbon release that only exacerbates the presence of greenhouse gases and subsequently climate change.

There are also a number of health impacts that come along with forest fires, some of which are quite obvious as the lowered air Quality as fine particulates bring about more respiratory problems. In fact, healthy firefighters can feel impacts for over 4 months after fighting fires. What most people fail to realize unless in a situation similar is that forest fires are quite traumatic and lead to significant levels of PTSD or other related mental illnesses. Evacuation and the amount of stress that come with are correlated to higher levels of mental health issues. The loss of homes, families and people is very traumatic. Moreover, during and after fighting fires there are concerns for food Safety and water quality. High levels of toxins and chemicals exist in the air that can find themselves running into water systems.

What's Next for Fire-Guard

In the future, the data received from the multiple nodes around the world will give us enough training data to improve our machine learning algorithm. We hope to have enough data to better account for particular local climates in order to remove the possibility of false positives. We also hope that this data can be used to migrate to another model which can even be used for predicting forest fires before they arise.

Share this project:

Updates