Intelligent GIS has been developed for the developer of Shakhty aircraft systems

2020: Geodata classification application development

Norbit (part of the LANIT group of companies) and its machine learning partner N-Systems have developed a geodata classification application for Shakhty, a developer of unmanned aerial systems and data processing and storage software. This was announced on March 11, 2020 by LANIT. The implementation of the project is aimed at eliminating violations in the field of urban planning and land legislation, and as a result, increasing the tax base.

The project is part of the national program of digital transformation of regions and includes the collection of data through the flight of territories and the compilation of high-precision orthophofuels, the creation of geoinformation systems and their integration with other data sources at the level of federal and regional structures.

The application allows you to automatically determine the coordinates of buildings and structures captured by unmanned aerial vehicles. By comparing the coordinates of objects with images and data from cadastral registers, the company with high accuracy detects violations in the field of land legislation. The developed machine learning model recognizes over 80% of all buildings and structures, regardless of the time of year.

The development of the application used open source technologies, the Python development language, the GeoServer geo-information system, the Keras neural network library, the Mask-R CNN neural grid, OpenCV, TensorFlow. Model training was performed on nVidia equipment.

The application solves an important task to eliminate violations in the field of land issues, including helping to regulate the replenishment of regional budgets with taxes. We plan to apply the developed geoinformation systems in many areas of the regional economy. For March 2020, a "roadmap" is being formed for the implementation of such projects in several regions, noted Dmitry Dobrovolsky, General Director of Shakhty

Together with the NORBIT team, we made a project whose peculiarity lies in the use of machine learning algorithms to transform spatial data. During the project, we used not only orthophofuels, but also dense point clouds and an altitude map. So, in a short time, our customer received an effective technology for classifying buildings and structures, said Dmitry Demidov, CEO of N-Systems