Developed the Gansu Province Nature Reserve in China, In cooperation with academic bodies, A smart system that uses facial recognition technology to remotely monitor giant panda bears to study them and find better methods and strategies to protect them from extinction.
The system depends, which was developed with Sichuan Normal University and Nanyang Singapore University of Technology, On a network of camcorders and a host of artificial intelligence techniques that can distinguish bears from each other, Which contributes to the identification of each individual to form a file about him including age and gender, Then study his behavior, track his life and keep his safety like never before.
Obtaining accurate statistics about animal species in their habitats, updated data on their movement, spread and threats, is essential in the process of developing effective national policies and strategies for their protection. But preparing studies to collect that data in the field is a laborious, expensive, lengthy and error-prone process. This is in addition to the additional efforts required to update the data continuously. For example, In nearly the past fifty years, China has not conducted more than four field surveys to determine panda populations. The latest was in Sichuan Province, which has the largest population of giant panda bears in China. However, due to the spread of pandas over vast patches of bamboo forests and rugged mountains (an average of less than seven bears per 100 square kilometers, As the study later revealed, officials hired about 700 scientists and field researchers from more than 100 institutions and organizations to complete the study. These and other challenges have underscored the need to find innovative, more feasible and effective methods of managing reserves and protecting giant pandas.
But facial recognition technology has also been challenging when applied to animals such as pandas. Their faces are similar and look free from differences to the human eye. The beginning of the project was two years ago when the Gansu Reserve deployed 300 high-resolution and sensitive video cameras equipped with infrared night photography in strategic locations throughout the reserve. It was linked to an artificial intelligence system for image processing. These cameras took thousands of photos (2,896 in 2021 alone) and captured videos of giant pandas – as well as other rare wild species – in various aspects of their daily lives (diet, mating, relaxation, fighting). There were 110 bears in the reserve. The researchers also used other images from the nearby Sichuan Reserve — home to 1,387 bears — to create a rich database of 12,000 photos and about 10,000 videos with the researchers' observations.
Over the course of two years, the researchers used a range of advanced techniques to process these images. It included computer vision that captures the smallest details and differences in visual data and analyzes it similarly to complex thought processes within the human mind, but at high speed to extract relevant information, and machine learning techniques including deep neural network technology, Which, when linked to the cameras, enabled the system to monitor the behavior of bears and follow their activities in real time and analyze them automatically to extract useful information. But most importantly, the algorithm developed by the researchers showed that there are indeed visible milestones. albeit minor, The system can identify them in the faces of panda bears in order to distinguish them from each other, The most important of which is limited to the black spot around the eyes and ears, In addition to differences in the way the panda closes its jaw, and his body posture, And the shape of the shadow on his cheeks. According to officials in the reserve, The success rate of panda face recognition and discrimination from each other is up to 98%. The system's success rate in identifying other wild animals in the reserve has reached about 80%.
Thus, researchers in the reserve can collect the necessary information and data about the pandas from their offices in real time without the need to take field trips that used to consume longer than their tasks and which may negatively affect the pace of life and stability of bears when humans are in their habitats. The system replaced field inspections in which DNA samples of bears were collected to determine their sex, age or other characteristics – either by analyzing their excrement or directly examining the animals. It is interesting to note that the smart system also detected errors that had occurred in previous classifications of bears, This further boosted confidence in its accuracy.
The system now provides a comprehensive view of the state of the bears at any given time. Whether it's her wellness and her behavior, or any environmental or biological developments that may occur in or pose a threat to their habitats. It will also contribute to the development of more effective policies and strategies in protecting giant pandas. In addition, such a system could be widely used to protect other rare or endangered animal species.
It is noteworthy that the technology of visual recognition of animals in nature reserves through the face or body characteristics, It had previously been used in Kenya to create a database of wild lions to determine their numbers, locations, mobility and protection. Successful experiments were also conducted to automatically identify some spotted or striped animals such as tigers, giraffes and zebras. Which are distinguished from each other within their species in the unique shape of the stripes on their fur. Studies are currently underway to take advantage of facial recognition technology in several applications, including remote sensing, And identify livestock such as cows and others in the precision agriculture sector. On the other hand, Environmental activists are using artificial intelligence algorithms to search social media for illegally trafficked chimpanzees.
References:
