Earlier in 2020, the forest change monitoring subsystem was tested in test mode throughout the entire territory
According to Mikhail Nikitin, head of departmentprotection, protection and supervision in the forests of the Ministry of Natural Resources, Forestry and Ecology of the Perm Territory, the system increases the efficiency of control and supervision activities in the region. In addition, it reduces patrolling costs by streamlining forest surveys. It is much easier and faster to react to specific signals displayed on the map. Thanks to this, more and more objects with alleged violations are found. In the future, these subsystems can be used as an evidence base in control and supervisory activities and in courts.
Innopolis University Developers and CompaniesInnogeotech has created an algorithm that eliminates the problem of missing small objects typical for neural networks: algorithms for determining clearings work with objects with a size of 3 * 3 pixels. The problem of the presence of haze from clouds in the images was also solved - algorithms automatically distinguish haze in the sky from forest changes, previously additional processing was carried out for this. The algorithms work in summer and winter with images from Landsat 8 and Sentinel 2 spacecraft.
"Service for monitoring forest changes automaticallydownloads space imagery data at weekly intervals. Modern technologies of image processing and deep learning make it possible to effectively solve problems that seemed impossible a few years ago - emphasizes Ramil Kuleev, Director of the Institute of Artificial Intelligence at Innopolis University - The direction of development for the forestry industry is very important for us, we see the prospect in solving problems of automatic taxation forests, the integration of various data sources - space imagery, lidar imagery and drone imagery, forecasting the development of negative situations, including emergencies - fires, forest drying up ”.
“We have reduced the minimumarea of detected forest changes. Thanks to the large volume of the reference sample, our neural network currently detects clearings in images with clouds and cloud shadows, explains Dmitry Shevelev, head of the forest industry digitalization project at Innopolis University. “Before this, we had to cut out the clouds in the images, or use cloudless images. We also continue to work on expanding the database of satellite imagery sources. Now the subsystem is being finalized in terms of using data from domestic satellites Resurs-P and Kanopus-V.”
On the territory of the Perm Territory for continuousmonitoring will include 12.4 million hectares of forest resources. Prior to that, the developers of Innopolis University introduced the technology in the territory of the Republic of Tatarstan, in an automated mode it monitors forests on the territory of 1.2 million hectares - 31 forestry districts of the republic. The service, using artificial intelligence technology, analyzes space images received from Earth satellites, preprocesses them and sends the results to neural networks, the networks segment these images and issue a vector with polygons. This service was developed as part of the creation of an Integrated Remote Monitoring System for the Volga Federal District, which also monitors agricultural land, infrastructure and capital construction, and waste treatment processes.
“Work experience on the territory of the Republic of Tatarstanand the Perm Territory gives us the opportunity to test the work of the forest monitoring service over large areas. We see that thanks to the work done and continuous improvement, the service can now be scaled to large areas, in the near future covering the entire forest fund of Russia,” concludes Dmitry Shevelev.
Read also
Abortion and science: what will happen to the children who will give birth
Check out the most beautiful pictures of Hubble. What has the telescope seen in 30 years?
NASA published a photo of the Earth from the moon, which was taken in 1968