Sensor Network System for Condition Detection of Harmful Animals by Step-by-step Interlocking of Various Sensors

In recent years, damage to crops and injuries to humans by harmful birds and animals have increased in areas of Japan. In order to support the coexistence of humans and wildlife, technologies for observing the ecology of wildlife and detecting harmful birds and animals are attracting attention. In our previous study, a method of detecting animals by utilizing a radio beacon has been proposed. In the system, the reception signal strength of radio waves transmitted between radio beacon devices is continuously measured and analyzed to identify the existence of the wildlife near the devices by using machine learning technology. However, the existing method of analyzing the radio wave strength cannot estimate the posture of the animal which can be utilized to identify the detailed behavior of the animals (e.g., the point where feeding behavior was taken) and to understand the situation of damage caused by them in more detail. Therefore, in this study, we propose a new sensor network system that detects the presence of wild animals by analyzing images taken by the thermal camera that can measure the body temperature of the organism day and night. In addition, in order to reduce the power consumption of the entire system that is assumed to be installed in mountain areas where the power supply is difficult, the thermal camera is activated only when the doppler sensor with low power consumption detects the moving object. After that, by analyzing the captured images using a machine learning technology, the system attempts to estimate not only the type and the number but also the posture of the animals.