Our team of engineers, FPT development model AI can analyze data, rainfall, geology to localize areas at risk of flooding, and landslides.
In the context of rain flood takes place at many local team of engineers, FPT at the center for Research and application of artificial intelligence, Quy Nhon has built the foundation forecast rain flood, at address antoanmualu.org from 19/11. Born in circumstances of emergency, but this system “can help governments, people actively plan before a disaster occurs”, thanks to the ability to forecast floods, landslides by AI.
According to Vu Hong Fry, director of the center, in the first step, the group find reliable data sources to train the prediction model by AI, using hydrological data and maps flood risk has been the European construction and authentic. This helps the model learn the characteristics, hydrology, standards from the rain water accumulates to how the flow varies with the terrain.
After the pattern is “coaching background,” the group continue to load actual information of Vietnam. Thanks to that, when put into operation, the model has the ability to read precipitation data in real time, combined with the topography and hydrology to predict areas at risk of flooding or landslides.
Forecast results are visually displayed on a Google map. Here, each point of risk is staking, switch colors according to the level of risk to privacy rights, can fast evaluation. “The important thing is for people to see danger, from which actively evacuate, do not wait till the water up before running”, he Fried said.
The system is also used to forecast the risk of landslides, based on geological information and related data that the model was trained. The group is focused on flood and landslide, as these are the two morphological disasters common and dangerous in recent years.
Other than many forecast system traditional need data input from local models WHO of engineers, FPT almost automatically. “The province doesn't need to prepare your data is automatically collected and processed,” mr. Fried said. This helps the model can be extended quickly to the other local. With each new province, the group need approximately 6-7 hours coaching model, mainly due to restrictions on the server by the products created in a short time.
Besides the ability to forecast, the platform integrates two buttons “I am safe” and “I need relief”, which allows people to send GPS location with personal information to force function.
From an idea occurred between the time the flood was coming in fast in Quy Nhon, model ONE of the group of young engineers shows technology can play a large role in disaster response, not only support rapid response, but also can forecast to minimize the damage.
The accuracy of the model is not enough time for a comprehensive review, but the group said it received support from the experts, hydrology and volunteers. They consider the input data, collate outputs and feedback to calibrate the model. In the Future, the platform has integrated data for Dak Lak, Khanh Hoa, Quang Ngai. The group said a number of other local contact was made to deploy the test. In the future, the group intends to shorten the duration of training and preparation available model for the high-risk provinces, to when need be activated immediately.
Source: vnexpress