AI Aids Wildlife Conservation Efforts

Artificial intelligence is a game-changer for wildlife conservation. Tracking wild animals in the field can be a slow and dangerous endeavor. An AI-based image identification solution enables researchers to process vast amounts of data and better understand dozens of wild species and their environment.

Image credit: WildlifeAI

Just as a crime scene footprint can be a critical piece of evidence for a gumshoe, a wild animal’s footprint can be equally valuable for wildlife conservationists.

Every species—and every animal within that species—has a unique footprint. Each footprint and the substrate beneath it tell a story about the animal’s movements and its environment. The information gleaned from a footprint helps conservationists track wild animal behaviors and monitor endangered species. The information also aids in the preservation of biodiversity on the planet.

Put Your Best Foot Forward

Animal conservationists can learn a lot from a footprint, but processing all the information manually is a cumbersome task. The wildlife conservation organization WildTrack created the Footprint Identification Technique (FIT) as a non-invasive means of monitoring animals in the wild. Digital images of animal footprints are measured to create a geographical profile and then analyzed to determine the animal’s species, age-class, gender, and individual identity.

Man photographic animal track

Image credit: WildTrack

FIT is very accurate, but it’s time-consuming. Every image requires manual data input and annotation from field experts. Even when all the data is entered, the analysis is slow because of human and geographical limitations. To speed up the process and to support data collection from across the globe and from citizen scientists, WildTrack sought an artificial intelligence-based image analysis solution.

The conservationists collaborated with a team of researchers at the University of California, Berkeley to automate the FIT process. The resulting tool, which integrates JMP data visualization software and SAS data analytics, employs artificial intelligence, computer vision, and machine learning to identify and analyze animal footprints more quickly and accurately. SAS and Intel® are long-time partners in developing global analytics solutions. WildTrack is currently using the end-to-end solution in the field.

Image Analysis Gets off on the Right Foot

To train the algorithms, researchers lured tigers and other wild animals across a prepared plot of land. They captured multiple images of each foot—four feet per animal equates to four times as much data that can be collected.

Images from the wild have been added, and citizen scientists worldwide can upload photos as well. Researchers, trackers, and citizen scientists can download a smartphone app to upload images from anywhere across the world. The images and related data, such as location, date, and time, are fed into the AI-based platform, which can identify dozens of species with greater than 90 percent accuracy. In addition, the eBeeX fixed-wing drones from senseFly are equipped with an interface to support aerial data collection.

The WildTrack database includes footprints of rhinos, tapirs, and bears. The expanded data collection further strengthens the machine learning algorithms and encourages discourse across researchers. People who upload images can follow the movements of that animal online.

The AI-based solution enabled the WildTrack team to extend its research. What started as footprint identification now includes other biometric traits, including coat color, coat pattern, and vocalization. WildTrack can monitor migration patterns, mating activities, feeding and hunting behaviors, and social groups. Differences in the substrate indicate moisture levels, weather conditions, and gait changes. The latter can indicate speed of travel or signal injury.

Adjustments to the machine learning algorithms will support research in other areas as well. The AI-based solution can be used to count and monitor endangered animals and prevent poaching. Future uses could include tracking turtles based on their distinctive shell patterns, identifying the type and health of flora and fauna, or detecting and monitoring trails used by animals or humans.