For a long time, wildlife monitoring has been a difficult area for biodiversity monitoring, especially for birds, songbirds and other animal groups that cannot be easily detected by infrared cameras. Acoustic recognition provides a new way to monitor these hidden animals. Especially, the accumulation of data from authoritative scientific databases and the wide application of deep learning in acoustic recognition in recent years have accelerated the development of this field. The biodiversity informatics research team of the Institute of Zoology, Chinese Academy of Sciences has long been devoted to biodiversity data construction and data mining analysis, releasing various data products such as the Chinese Biological Species List, Chinese Biological Map, Chinese Animal Theme Database and Biographies app, which have been widely used, and accumulating rich experience in species image and sound recognition. Based on the authoritative list and distribution database, the team collects and integrates animal sound data, combines deep learning with ecological models for joint modeling, and builds an intelligent recognition model of sound patterns mainly for birds; and using IoT technology, the team has developed field sound pickup sensors and edge computing devices to collect environmental sounds in real time for edge computing and return recognition results to the monitoring system in real time for analysis, constituting A complete automatic wildlife audio monitoring system.
At present, the wildlife acoustic intelligent monitoring system is adopted by the Ministry of Ecology and Environment, and is the only equipment system based on acoustic recognition for biodiversity monitoring. The system was first deployed in the mobile monitoring vehicle and ground monitoring station in Tongliao, Inner Mongolia for demonstration and verification. The system has been working stably for more than 2000 hours so far, and nearly 10,000 valid recordings have been returned, and more than 60 types of environmental sounds have been monitored and recognized, and the system performance has been effectively verified. Up to now, the system has collected nearly 70,000 effective sound data and monitored and recorded the songs of more than 200 species of birds, which will strongly support the biodiversity background survey and long-term monitoring activities in these areas, and also provide a continuous stream of animal sound and distribution data for the CAS species diversity database to support the long-term research work on the spatial pattern and changes of biodiversity.
The system is led by Dr. Lin Congtian. The test equipment was provided by Habitat (Beijing) Science and Technology Co., Ltd. and the Institute of Microelectronics, Chinese Academy of Sciences provided suggestions for equipment improvement.