Over 10,000 Species of Birds Exist
You may only need to focus on one. We can help.
Identifying a Species in Real Time
The overwhelming advantage of IdentiFlight over other approaches is the ability to determine species in real time. This is accomplished using a proprietary hybrid of machine vision and neural network technologies.
Machine vision techniques consist of analyzing images against a set of rules to determine the degree to which an image matches the features sought. For example, is the bird of the right size? Does it have color characteristics matching the target bird? Are the wing profiles in flight similar to the target bird—for example, is it a flat glide profile of an eagle or the dihedral shape of a vulture?
In addition to machine vision techniques that analyze specific pre-determined attributes of an image, IdentiFlight also utilizes convolutional neural network technology. In this technology, millions of images gathered from IdentiFlight’s data set are used to train a neural network. The neural network is trained with images known to be the target species and tuned to reduce errors in predicting the species in a test set of known bird images. IdentiFlight finds that with appropriate training, the neural network classification is superior even to machine vision techniques. Using the convolutional neural network data allows the system to continuously learn and improve as the data set grows.
Need to Protect Another Species?
IdentiFlight technology is easily adapted to any species. For example, a species may be incorporated into the machine vision algorithms by collecting a suitable library of images of the target species. The library ideally would include images in a wide variety of conditions, such as high contrast, backlighting, and different flight presentations.
AI is the heart of species classification using convolutional neural network technology. Just as a personalized search engine performs better as more data are collected, so too does IdentiFlight’s classification of species. IdentiFlight has developed processes that collect and analyze data to improve the classification of species as site conditions vary, such as weather or even forest fire smoke.
By the very nature of endangered species, occurrences are rare. However, IdentiFlight can address this issue by making conservative species classification decisions early and improving the library over time. As more data are collected, the classification algorithms are updated with improved results. Internet search engines, for example, use a similar process for improvement.
Wedge-Tail Eagle
Aquila audax
Native to Australia
Bald Eagle
Haliaeetus leucocephalus
Native to North America
Red Kite
Milvus milvus
Found throughout Europe
Condor
Gymnogyps californianus
Native to North America
Honey Buzzard
Pernis apivorus
Found across Europe
White-Bellied Sea Eagle
Haliaeetus leucogaster
Native to Australia
Black Stork
Ciconia nigra
Found across Europe
Golden Eagle
Aquila chrysaetos
Found across the Northern Hemisphere
White-Tail Eagle
Haliaeetus albicilla
Found across Europe and temperate Eurasia
The IdentiFlight Advantage
Alternative technologies have not proven to be as effective as IdentiFlight. The first and most common approach is the use of human “biomonitors” to scan the skies for protected birds and then order curtailment of specific turbines. Numerous studies have shown that IdentiFlight is capable of detecting many times more birds than people can. Further, being an automated solution, IdentiFlight can continue to work through inclement weather, such as lightning, when biomonitors have to leave the site for safety reasons.
Other technologies use wide angle camera systems to detect birds and activate noisemakers to chase all birds away so that the turbine does not have to curtail. Unfortunately, there is insufficient evidence that these deterrents are effective. Further, the camera systems lack the ability to know the bird distance, so confusion among species, including insects, is common. An IdentiFlight advantage over such systems is the ability to detect birds at sufficient distance that curtailment is a viable conservation strategy. Furthermore, noise of deterrents may complicate permitting and public acceptance, and the birds may become accustomed to the noise and any protection value is lost.
Radar has been offered as another possible solution. IdentiFlight’s detection technology is superior to radar for diurnal species, as radar has no ability to determine species effectively. Though radar wavelengths receive return energy from bird-sized objects, the ability to detect birds from clutter – including rotating turbine blades – has been highly problematic. IdentiFlight’s masking algorithms allow it to see through rotating turbines with minimum confusion, continue to track birds while the high resolution stereoscopic cameras are dispatched to examine specific targets, and discriminate from unimportant flocks of birds that can be troublesome for radar.
The overwhelming advantage of IdentiFlight over other approaches is the ability to determine species in real time using a proprietary hybrid of machine vision and neural network technologies.
