Artificial intelligence is revolutionising wildlife monitoring

It is of high conservation value to monitor species and assess ecological indicators that reflect the health of ecosystems and species most vulnerable to anthropogenic change.

However, challenges arise when protecting species that are elusive in nature, have a large geographical range, or exist at low densities. Current techniques being utilised for monitoring include surveying on-foot, satellite tracking, and camera traps. These are all either highly labour intensive, prone to human error, and may produce limited results at a great expense.

Unmanned aerial vehicles (UAVs), otherwise known as drones, are aircrafts without a person onboard and can operate either autonomously via preprogrammed flight routes or controlled by an individual on the ground. UAVs are becoming gradually less expensive and are increasingly used to monitor wildlife. Although there have been concerns associated with the disturbance of natural behaviour and obtaining permits, this method of data collection generates higher accuracy when calculating the abundance of species than traditional methods. To date, UAVs have been used on many terrestrial and marine fauna, including black bears and elephants, as well as dugongs and turtles. Furthermore, UAVs can now be fitted with thermal imaging cameras and acquire georeferenced images to successfully and geographically pinpoint individuals.

Due to increased footage and images obtained post-flight, the processing time is drastically increased and may negate any reduced survey time. In order for UAVs to be truly efficient, the image processing of conservation tools must be improved, and automated image detection could speed up this process.

Artificial intelligence is intelligence learned by machines allowing them to operate like humans. This may include problem solving, planning, and recognition. Machine learning describes the process by which machines adapt and learn through experience. These two new research fields scientists have been able to ‘teach’ machines to recognise even the rarest of species such as the endangered reticulated giraffe. In a study on images from the Serengeti National Park, AI was able to label six months’ worth of images in just a few hours with 96.6% accuracy. In comparison, it took thousands of volunteers 2-3 months to accomplish the same feat. By training machines to combine optical (by sight) and acoustic sensing (by sound) machine learning to wildlife identification, multi-modal sensing can result even more accurate identifications.

Unmanned aerial vehicles and artificial intelligence are of particular use in marine conservation. Our oceans are a three-dimensional and dynamic environment covering more than 70% of the Earth’s surface. Moreover, extensive data collection forms the framework of our management decisions. Thus far artificial intelligence has overcome the bottleneck of processing large data and the inaccuracies that occur from human error. Through UAV and AI technology, our most remote environments such as open oceans are becoming increasingly accessible, and there is no doubt that the future will see scientists look to the development of underwater autonomous technology with machine learning capabilities.