The Moscow Institute of Physics and Technology (MIPT) has adapted neural networks for state land monitoring, according to the press service of the university's Center for Scientific Communication.
The neural network segments and classifies data about detected objects in the studied areas.
Instead of spending a whole day walking around land plots, we launch a drone with lidar and take a survey. We clean the data from noise and send it to the neural network.
Experiments were conducted with a neural network based on the PointNext algorithm. It was created for processing data from lidars (laser radar).
Scientists prepared several virtual landscape models using a generative algorithm. Based on this data, they trained the neural network to recognize objects from lidars with 99.98% accuracy.
Earlier, www1.ru reported that makes the "Digital Earth" project.
Read more on the topic:
It became known how many farmers in Russia are ready to use drones for land cultivation
An exoskeleton of the arm with integrated AI was developed at MIPT