A Detection Method of Garbage Collection Status from Sound of Garbage Trucks

As part of services for citizens in smart cities and the Sustainable Development Goals, there is a service that visualizes the status of garbage collection and allows citizens and government officials to check the status. In Japan, household garbage is put into designated garbage bags and dropped off at designated collection points or in front of homes, and garbage trucks collect them. Therefore, it is not possible to introduce the well-known method of detecting the presence or absence of garbage by attaching sensors to garbage containers. This paper proposes a method to determine the garbage collection status from the sound emitted by garbage trucks utilizing edge computing. A spectrogram of the garbage truck sound is generated and a Convolutional Neural Network is used to identify the characteristic sounds generated during garbage collection. The proposed method was applied to 14 days of garbage collection, and it was confirmed that the proposed method could determine the garbage collection with 90 % accuracy.