Indian Students Use Artificial Intelligence to Predict Wildfires in California
Two students from California, Aditya Shah, and Sanjana Shah share how they developed a machine that can predict areas that are susceptible to wildfire using Artificial Intelligence (AI).
Long ago, Aditya used to capture different pictures of the Big Basin Redwood State Park where the giant sequoias are situated. In 2017, these centuries-old trees were turned into ashes as the most destructive wildfire hit the area burning at least 9, 000 tree.
This phenomenon challenged Aditya to find a solution that could predict the likelihood of wildfire and eventually prevent them. Together with Sanjana, Aditya uses AI Tensorflow – an open source machine learning tool from Google to identify the areas where wildfires can mostly occur.
How it works
Tensorflow utilizes artificial intelligence to analyze the images of biomass and estimates the likelihood of a tree to emit fire. First, it is attached to the tree. Then, it gathers data regarding the moisture content and the size of the tree to determine the amount of dead fuel. It is also used on downed branches, twigs, and fallen leaves.
When the moisture content is 0 percent, it is categorized as dead
fuel. The more dead fuel a certain tree or area has, the higher its potential to ignite fire especially during warm seasons. If the moisture content is high, its likelihood to emit fire is low because most of the fire’s heat energy would be used up to drain the moisture before it burns.Aditya explains that if the area has 0 percent content moisture, they could report it to the fire department for close monitoring. “Since the authorities know which part of the forest is more likely to ignite a fire, they can do something to prevent wildfires from happening,” says Aditya.
Register For International Conference on AI in Biotech & Healthcare
Future application
The two students have now connected their Smart Wildlife Sensor to connect with a network of a sensor. Consequently, the fire department does not have to physically visit the forest areas to collect samples of dead fuels and classify them manually.
The sensor can predict the likelihood of wildfire within 100 square meters in the forest. If the estimation goes well, it can potentially prevent the almost uncontrollable spread of wildfires thus reducing the cost that the government spends to protect lives and properties.
Last summer, Aditya, and Sanjana were able to test their device in California’s forest. With the help of Cal Fire, they were able to capture images of dead fuel which were accumulated on the forest ground.
The two students are going to make further tests this year to train the machine learning model. This time, they are planning to use ground and aerial drones to capture more images of biomass in wildfire-prone zones in California.
Aditya and Sanjana are going to make further improvements to their machine. This means that everyone can expect lesser fire incidents in the future which could be very beneficial not just in California but also to other countries.
Table of Contents
Want to learn more about Artificial Intelligence?
Join Us at AIBTHT Conference on 25th November 2018.
Source: Google Technology Updates
