We are creative, ambitious and ready for challenges! Hire Us
ENHANCED FLIGHT SAFETY: EDGE BASED AI BIRD SURVEILLANCE
Client’s Need
Reduce risk of bird strikes near airfield approach airspace. Bird surveillance system capable of real-time detection using offline and inexpensive edge hardware processing.
Goals of the project
- Bird surveillance system capable of real-time detection.
- Edge based solution operating standalone basis with no internet or cloud access needs.
- Process HD Video at 25 frames per second in real time.
- Adhere to strict power limit of 15 watts for the solution.
Our Approach
- Divide each frame into 12 sections for enhanced bird detection accuracy.
- Apply YOLOv5 deep learning model to each segment to detect birds more accurately.
Utilize Nvidia Deep stream for efficient batching and processing of segments.
Outcome
- Enhanced flight safety at the airfields.
- Achieved real-time bird detection near airports.
- Full HD video processing at 25 fps on an edge-based system.
- Leveraged YOLOv5 and Nvidia Deep stream for efficient processing.
- Strict adherence to hardware and power limitations, allowing integration with various deployment modules including UAVs.
- Reduced potential accidents and financial losses for the aviation industry.
Published:
September 28, 2024
Category:
Edge Computing
Client:
Oceanthemes
