I live in American Samoa where we have large populations of flying foxes, seabirds, and forest birds. There is a plan to build a 120 m met tower with 30 guy lines (10 lines in 3 directions) to study wind related to a potential wind power development on top of the mountain in rainforest habitat. The lanes for the guy lines have been cleared out to 80 m. We plan on installing day and night (LED) bird diverters on the guy lines hoping this will reduce mortality. I believe we will have a very difficult time with mortality searches as most of the animals will end up in the rainforest or will be eaten by cats, dogs, rats, or coconut crabs.
I am looking for a system to record collisions and mortality caused by the tower and lines. I am discussing this with some companies but have not found a system that is suited for this. I am hoping to record collisions day and night (thermal or IR) and use AI to detect events. We are also thinking of using acoustic monitoring as this has been used in other places to detect bird collisions with high power lines. Are there any available systems for this or is this something that could be developed with funding? Thanks for any information that could be helpful with reducing or monitoring mortality, especially of night flying flying foxes and seabirds. Included is a
11 June 2025 6:54am
We have a thermal imaging product that could see the animals at night well.
To detect collisions, you would need a high inference rate a Jetson could do that, a middle range Jetson with using a megadetector type model would be the way to go I think.
In principle we could develop a solution against funding. We have all the hardware and coding to do it. What needs developing is a lab algorithm to read all the annotations on the images to found bounding boxes that card close to each other and maybe needs a suitable megadector model. Likely PyTorch wildlife would be good for that.
30 June 2025 4:55pm
Hi Adam,
Have you connected with Wildlife Imaging Systems? Their thermal system can detect bird and bat activity. My company collaborates with them, and we have previously worked on developing algorithms to distinguish collisions from other activity.
Kim Hendrikse