Hydropower plants must be passable for fish in order to enable fish migration. According to the Water Protection Act, the effectiveness of such fish passes or fish lifts must be proven. For this purpose, fish were previously recorded in a manual procedure. To improve efficiency and scalability, BKW has collaborated with Helbling to develop an algorithm for automated fish presence detection with determination of size and number of fish. The algorithm generates a statistical report on a daily basis with number and size of detected fish.
Services
Industries
Technologies
Key figures
- Increase of efficiency and scalability
- Fish detection with 96% sensitivity and 98% specificity
- Automatic generation of daily statistics
- Problem understanding, data preparation, algorithm development, evaluation and deployment
Our Contribution
- Selection of appropriate image processing method (rule based algorithm vs. artificial intelligence)
- Assessment of vision system and recommendations for improvement
- Manual annotation of >4000 images for training and testing
- Development of neural network based on YOLOv5 architecture
- Statistical evaluation of detection performance
- Support for integration on target system
Outcome
- Machine vision replaces tedious manual inspection
- The presence of fish in an image is detected with a sensitivity of 96% and specificity of 98%
- Detected fish are classified into 6 size groups
- Results are compiled in a file and daily statistics are automatically generated
- Fish detection algorithm is deployed on a Linux system
- List of recommended hardware modifications, incl. type of illumination, camera arrangement, and imaging parameters
Contact
Cases
Internationales Unternehmen der Papierindustrie – Analyse des Dampfsystems
Industries:
Services:
Cases
6-Streifenausbau Aarau Ost – Verzweigung Birrfeld – Bauherrenunterstützung
Industries:
Services:
