Harmful Animals Detection System Utilizing Cooperative Actuation of Multiple Sensing Devices

In recent years, there have been several incidents of crop damage and injury caused by harmful animals in various areas of Japan each year, amounting to about 15.8 billion yen in 2018. In order to reduce the damage, a number of existing studies have been conducted on camera-based systems. However, this existing system requires that the sensing devices should always be running, which makes it inappropriate for installation in mountainous areas where electronic power is difficult to be supplied to the system. Therefore, in this research, we propose a new harmful animals detection system that can detect not only the approaching of animals to the traps and the fences but also their species and postures by combining various sensing technologies (i.e., beacon sensing, laser radar, and depth camera). The beacon sensing attempts to detect the passage of moving objects by analyzing changes in received signal strength caused by reflection, diffraction, and absorption of radio wave beacons by the object. After detecting the passage of the moving object, a small computer is activated to measure one-dimensional distance to the target object using a laser radar. The time-series data of the measured distance is analyzed by the machine learning technology to estimate the type of the moving object (e.g., human, animal). If the moving object is judged as a harmful animal, the small computer activates the depth camera to acquire two-dimensional distance data of the target animal. The acquired distance data is analyzed by the machine learning technology to estimate the posture of the harmful animal. As explained above, by gradually activating the sensors with higher power consumption, the proposed system achieves power-saving.