11:15 AM - 11:30 AM
[ACC25-08] Highly efficient snowfall detection using a visual IoT system
We propose a snowfall detection method using a moving object detection algorithm based on visual IoT technology. Visual IoT is a type of IoT sensor installed outdoors and can acquire high-resolution and high-time resolution video. We use commercially available IP cameras, which offer excellent cost performance and can be easily installed in various positions. By using a single board computer (Raspberry Pi) on the edge side along with the IP camera, flexible operations such as acquiring a few seconds of video every few minutes are possible. With this system, we were able to acquire data on the size and quantity of snowflakes from 1080p, 25fps videos. The snow area is detected in the screen from the image difference between the video frames, and the tendency of snowfall is obtained. The shape and distribution of the snowflakes captured in the camera images vary with the time of day (visible light and infrared light) and the camera's installation position. In this study, we obtained data on the size and quantity of snowflakes from video images and summarized their temporal transitions.