Artificial intelligence to improve the detection of animals in isolated areas?

A recent study examined the feasibility of using artificial intelligence to improve camera traps that facilitate the detection of animals in isolated areas.

The authors of the study, from the Institute of Microelectronics of Seville (IMSE) and the Doñana Biological Station (EBD), have determined that it is viable to use artificial intelligence to improve these camera traps and have devised the best way to bring this technology closer. to places where there are hardly any communications infrastructures, and where detecting fauna is not an easy task because the programs are not optimized.

The IMSE is a joint center of the Higher Council for Scientific Research (CSIC) and the University of Seville, both entities in Spain. The EBD depends on the CSIC.

“Our experience in the design of microelectronic hardware has allowed us to face the challenge of capturing these images in complex scenarios and processing them in the smart camera itself. The work of Delia Velasco Montero is very notable, which has allowed the algorithms to be adapted to very demanding operating conditions. It is enough to consider, for example, the changes in lighting throughout the day, or due to the weather,” says one of the main researchers of the project at the IMSE, Ricardo Carmona.

In addition to considerably reducing the manual work of reviewing the images taken, this system allows the presence of animals to be detected in real time, facilitating rapid response actions. Another advantage of integrating artificial intelligence into the camera itself is the reduction in the data stored to transfer and analyze, as it filters the information, discarding that which is not of interest. All of this allows the biologist to focus directly on data analysis, such as behavioral patterns, population monitoring, etc.

This innovative method has been integrated into a hardware prototype developed at IMSE and has been put into practice in the Aracena National Park. Regarding its application, although it was initially in a local environment, it has sufficient potential to adapt to other similar geographical areas.

For its part, the EBD’s carnivore research group studies the population trends of several species in their natural habitats. To do this, they use strategically placed camera traps, which end up generating a large volume of images and video sequences, most of them of no interest. Often, that amount of visual material is too much to analyze and categorize manually.

Capturing images of animals in isolated places within forests and other natural areas is not an easy task, but it can be thanks to the help of artificial intelligence. (Photo: Amazings/NCYT)

Specifically, for this study images collected by Ariadna Sanglas and Paco Palomares in the Sierra de Aracena Natural Park have been used, within the framework of a project to evaluate the population of feral cats. The work of manually reviewing images involves a large amount of time. You must first discard the images in which no animal appears and then correctly identify the species or species that appear in the remaining photographs.

The study is titled “Reliable and efficient integration of AI into camera traps for smart wildlife monitoring based on continuous learning.” And it has been published in the academic journal Ecological Informatics. (Source: CSIC)