Random bounce algorithm: real-time image processing for the detection of bats and birds Algorithm description with application examples from a laboratory flight tunnel and a field test at an onshore wind energy plant

Wind energy plants generate an impact on wildlife with significant fatality rates for various bat and bird species, e.g. due to a collision with the rotor blades. Monitoring approaches, such as vision-based systems, are needed to reduce their mortality by means of an optimized turbine control strate...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Signal, image and video processing Jg. 10; H. 8; S. 1449 - 1456
Hauptverfasser: Scholz, Nikolas, Moll, Jochen, Mälzer, Moritz, Nagovitsyn, Konstantin, Krozer, Viktor
Format: Journal Article
Sprache:Englisch
Veröffentlicht: London Springer London 01.11.2016
Schlagworte:
Computer Imaging
Computer Science
Image Processing and Computer Vision
Multimedia Information Systems
Original Paper
Pattern Recognition and Graphics
Signal,Image and Speech Processing
Vision
Video object segmentation
Camera-based detection of bats and birds
Random bounce algorithm
Wildlife monitoring
Wind energy plants
ISSN:1863-1703, 1863-1711
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Wind energy plants generate an impact on wildlife with significant fatality rates for various bat and bird species, e.g. due to a collision with the rotor blades. Monitoring approaches, such as vision-based systems, are needed to reduce their mortality by means of an optimized turbine control strategy as soon as flying animals are detected. Since manual analysis of the video data is ineffective, automatic video processing with real-time capabilities is required. In this paper, we propose the random bounce algorithm (RBA) as a novel real-time image processing method for vision-based detection of bats and birds. The RBA is combined with object tracking in order to extract flight trajectories. Its performance is compared with connected components object detection. Results from a laboratory flight tunnel as well as from a field study at a 2 MW wind energy plant in Southern Germany will be presented and discussed. We have successfully detected and tracked objects both in laboratory experiments with many animals and in field experiments with individual animals at a frame rate of 10 fps.
ISSN:1863-1703
1863-1711
DOI:10.1007/s11760-016-0951-0