Development of a Raspberry Pi based pest detection device for use in livestock feed storage systems

摘要

Stored feed is frequently damaged by rats. In this study, we made use of the compact "Raspberry Pi" computer to develop an animal damage countermeasure. We also attempted to make the automatic photographing apparatus for dark environments. This deviceincluded a Raspberry Pi, infrared camera module, MOSFET. infrared LED floodlight, and pyroelectric infrared ray sensor. The operation of the apparatus is as follows: When the pyroelectric infrared sensor is triggered by the presence of an animal and thesignal becomes HIGH, a signal of HIGH is output from GPIO of the Raspberry Pi and a voltage of 12 V is applied to the MOSFET to make it function as a relay circuit. The infrared LED projector then lights up. At the same time the device starts taking a still image or video using the infrared camera that is connected to the Raspberry Pi. The apparatus succeeded in shooting a movie of a rat, and it successfully took still pictures. We installed the machine-learning library "TensorFlow " on the Raspberry Pi and performed image recognition of murine images using a deep learning program. When the target animal occupies a small part of the image, the image recognition results tend to be erroneous; however, enlarging the area of interest results in more accurate detection. In addition, erroneous judgment was often made in infrared camera images. The time required for the determination was 15 to 45 seconds for the Pi 3 Model B. These results show that it is possible to create a low-cost pest countermeasure device or robot controlled by a Raspberry Pi, and capable of various responses and input processing.