SmartGCM: An LTE UE-Category 1-Based Smart Garbage Classification and Management System

Garbage classification and management(GCM) has become one of the key issues in the development of smart city. Due to single functionality and expensive deployment, many existing garbage classification and management systems are limited in some practical application scenarios. In this paper, we design SmartGCM, a smart GCM system based onLTE UE-Category 1(Cat.1), which can detect real-time status information from community trash cans and classify different types of garbage. Specifically, SmartGCM is based on the embedded microcontroller to remotely monitor the real-time status of trash cans. Then, we designed an optimization algorithm based on the genetic path algorithm to solve the problem of garbage cleaning path planning; meanwhile, combined with the incentive mechanism, a lightweightconvolutional neural network(CNN) based on edge computing is deployed and implemented. To improve the accuracy of SmartGCM classification, we perform edge extraction and smooth noise through the Sobel operator combined with LeNet-5, a deep convolutional neural network. In experiment, we collect 2900+ real data of six types of garbage images, and obtain the optimal garbage classification model by designing reasonable convolutional kernel size and convolutional layer number. The experimental results demonstrate that SmartGCM achieves a classification accuracy of 92.75% and Macro-F1 of 93.15%. SmartGCM can not only improve citizens’ environmental awareness, but can also contribute to the development of smart cities.