Random bounce algorithm: real-time image processing for the detection of bats and birds

Abstract

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.