Speed up, accurate and applicable to all "quadrupeds"! The latest progress in animal monitoring technology has come.
When it comes to wildlife investigation, we will think of infrared cameras, which are called "undisturbed encounters", that is, the collection of images of wild animals is completed without their awareness. At the same time, the infrared camera also has the characteristics of "passive", that is to say, its shooting function will only be triggered when wild animals pass by.
UAV monitoring system helps wildlife survey "speed up"
Recently, the team of teachers and students of Northwest University has made great progress in wildlife observation technology through the integration of drones and various high-tech means.
In the picture, the research team of Northwest University is debugging drones and thermal imaging observation equipment and importing preset routes. They are going to conduct remote sensing monitoring and sampling on the distribution position and population size of Sichuan golden monkey, a star species living in the reserve, by means of low-altitude aerial photography.
Before the flight, the scientific research team had set a number of routes that could cover the whole reserve through many discussions. During the flight in strict accordance with the established routes, the team members flexibly used the alternate flight mode of high and low altitude, and completed the sampling of Sichuan golden monkey population within about 300 square kilometers of the whole reserve in less than 52 hours.
He Gang, Senior Engineer of Northwest University:We used a total of 170 routes in this area, spent about 3116 minutes, flew 977 kilometers, nearly 1000 kilometers, and completed the investigation of the whole area, covering about 92% of the area. Eight groups of monkeys were found in the investigation, with a total number of 647 to 755. We located each group of monkeys, and each group of monkeys was individually identified.
In fact, it is only the preliminary work of this investigation to complete the sampling with infrared thermal imaging equipment carried by UAV. Although the sampling results can collect the wild animals hidden in the dense forest into the imaging picture without omission, it is not easy to accurately screen out the Sichuan golden monkey from it. From the image of animal outline collected by thermal imaging equipment, we can see that there are not only Sichuan golden monkey, but also many wild animals such as giant panda, takin, Asian black bear, including macaques with almost the same shape as Sichuan golden monkey. In order to distinguish Sichuan golden monkey accurately, more professional ecological interpretation and data correction are needed to finally confirm the investigation results.
He Gang, Senior Engineer of Northwest University:UAV monitoring wildlife is a systematic project, which involves several modules. You must be able to fly, be safe, be able to see clearly, and be able to identify the target, and sometimes there is interference. Finally, you have to correct and judge its results, so many technical modules should be put into a comprehensive application.
According to experts, this technology is composed of unmanned aerial vehicle (UAV) equipped with many high-tech modules such as thermal imaging, and takes UAV as the main carrier, turning traditional passive shooting into active shooting, thus achieving the goal of standardized, full coverage and repeatable monitoring of wild animals in the target area. The economic cost and time cost required for animal monitoring are greatly reduced.
Animal behavior characteristics can also be used for individual identification.
A "monkey face recognition" technology developed by Northwest University has played an important role in the investigation and monitoring of Sichuan golden monkey.
Now, this technology has broken through again. Not only the facial features of Sichuan golden monkeys can be used for identification, but also their body movements and behavior habits can be used as the basis for individual identification.
The technology of "monkey face recognition" evolved from the widely used intelligent system of "face recognition". The technology of "face recognition" based on human facial features has actually broken through the limitation of only relying on faces.
Professor Guo Songtao of Northwest University:Face recognition has a high accuracy, but it doesn’t always capture the face. What should I do at this time? In fact, the development direction of artificial intelligence technology is to use a variety of conditions, which we call multimodal data to process in parallel. When it has no face, we use body or other features to identify it.
Monkey face recognition relies not only on monkey face, but also on "multimodal data" composed of animal behavior and other physical characteristics.
According to experts, the "monkey face recognition" technology can identify species based on not only the objective "shape" but also the subjective "state". This technology seems to be a big step from "perceptual" intelligence to "cognitive" intelligence, and its accuracy has also been greatly improved. In fact, it relies on more professional and in-depth observation and data collection of animals beyond the external features of the body, such as skeletal features.
"Multi-modal data monitoring" will be used for multi-region and multi-species
Based on the original "monkey face recognition" system, researchers innovatively developed a multi-modal data monitoring system. Then why did the research team use monkeys as the species used in the initial research and development of this technology, rather than other wild animals? What is the application prospect of this technology?
Sichuan golden snub-nosed monkey, a national first-class protected wild animal, belongs to primate and is biologically classified into the genus Rhinopithecus under the primate family. Yunnan snub-nosed monkey, a national first-class protected wild animal belonging to Sichuan snub-nosed monkey, will soon be the first batch of users of this monitoring technology. Researchers will expand the pilot area from Qinling Mountain in Shaanxi Province to Baima Snow Mountain in Yunnan Province.
Professor Guo Songtao of Northwest University:There are 27 species of primates in our country. We used to work in Sichuan Golden Monkey and Qinling Mountains. Recently, we will carry out a similar pilot project in Baima Snow Mountain in Yunnan Province. The wild monkeys in Baima Snow Mountain, like the wild monkeys we studied in Qinling Mountain, will be automatically identified and tracked.
Once a monitoring system can realize multimodal data recognition for primates, it means that this technology can be used for almost all quadrupeds.
This monitoring technology is applicable to almost all quadrupeds, including the takin which has been tested in Qinling Mountain, Shaanxi Province. It will also be applied to many kinds of wild animals in different protected areas in China.
(CCTV reporter Wang Shantao Zhang Yongfu)