Bird detection at wind farms
Research project for bird protection at wind farms with Recovery Fund support.
With the support of the Recovery Fund, Elektrum Group and Latvijas Mobilais Telefons SIA are implementing research project No. EP.01 "Automatic detection and identification of birds in wind turbine parks using computer vision and machine learning methods".
Research project No. EP.01 “Automatic detection and identification of birds in wind turbine parks using computer vision and machine learning methods” is being implemented within the framework of LEO Pētījumu centrs SIA project No. 1.2.1.2.i.2/1/24/A/CFLA/007 “Competence Centre of the Latvian Electrical and Optical Equipment Manufacturing Sector for Improving Energy Efficiency”.
The aim of the project is to research and develop a method for automatic detection and identification of birds in wind turbine parks, focusing on large soaring birds that are particularly at risk of colliding with wind turbine rotor blades.
The project plans to use a combined technology approach, integrating radar technology with high-resolution camera systems to enable bird observation also in the dark hours of the day. The solution developed in this way will be usable autonomously for avifauna monitoring.
The project provides for the creation of two technological solutions. The first solution, using machine learning algorithms, will be able to perform real-time bird detection at a sufficient distance and species identification. The second solution, using an ESA radar system, will be able to perform bird detection, trajectory analysis and distance estimation in order to effectively prevent collision risks according to turbine dimensions.
During the project, it is planned to combine both developed technologies in a single product prototype, providing broader and more complete functionality.
Project implementation deadline: 30 September 2027
Total project funding: EUR 400,780.91
Legal basis: Cabinet of Ministers Regulations No. 33 of 9 January 2024 "Implementation regulations for measure 1.2.1.2.i.2 "Development of innovative products and technologies" of investment 1.2.1.2.i "Improving energy efficiency in entrepreneurship (including the transition to the use of renewable energy resource technologies in heat supply and research and development activities (incl. in the bioeconomy))" of reform and investment direction 1.2 "Improving energy efficiency" of the Latvian Recovery and Resilience Facility Plan"
Research project implementation 04.2025 – 09.2026
Within the activity, a laboratory prototype for collecting bird observation data was created and improved by installing high-resolution camera systems at several observation sites. The cameras were placed on the base station infrastructure of Latvijas Mobilais Telefons SIA, ensuring long-term acquisition of image and video data in various weather conditions and seasonal periods of bird activity. The data obtained form the basis for developing a representative dataset for the development of bird detection and species recognition solutions.
During the activity, test software for the camera system was developed, which ensures data acquisition, storage and initial processing. At the same time, software algorithms and tools for annotating image and video data were created in order to systematically label bird observations and prepare high-quality training data for the subsequent training of machine learning models. The operating steps of the software algorithms for detecting bird objects in images were also defined, analysing various computer vision approaches and assessing their suitability for work in real environmental conditions.
In parallel with data collection, the collected image and video materials were structured, their quality was assessed, and they were prepared for the annotation process.
Research project implementation 10.2025 – 03.2026
Within research project No. EP.01 “Automatic detection and identification of birds in wind turbine parks using computer vision and machine learning methods”, the first activity of the project, “Development of a laboratory prototype of a technical solution with high-resolution cameras and radar for bird detection in wind turbine parks”, was carried out from October 2025 to the end of March 2026.
As part of the activity, an in-depth study of existing bird monitoring technologies was carried out, analysing visual-based systems, radar systems and hybrid systems. Based on this analysis, the main technical requirements for the laboratory prototype were defined, especially regarding the necessary spatial coverage, image resolution and the ability to detect small, distant objects. In parallel, a detailed study of algorithmic approaches was carried out – object detection, tracking, classification and decision-making – and the most significant challenges were identified, such as reducing false positives and accurately identifying species in real time.
During the activity, an initial system architecture was also developed, based on a combined approach in which radar provides early detection of flying objects and determination of trajectories, while high-resolution cameras provide visual information for bird identification. The principles of sensor data fusion, synchronisation in a unified coordinate system and the use of edge computing for real-time data processing were defined.
As a result, a technical specification of the laboratory prototype was prepared, which serves as a basis for further data collection, development of machine learning models and validation of the solution in real wind farm conditions.
State aid for the implementation of the projects is provided by Attīstības finanšu institūcija Altum AS.
Source: protecbird.com