Scholarlink

<p>The deployment of smart waste management systems is crucial for protecting the environment against pollution through proper recycling or landfilling. Nonetheless, Deep Learning (DL) and Internet of Things (IoT) promise a solution in automating waste classification. Meanwhile, low image quality and lesser classification accuracy are still posing a challenge. This paper presents a novel smart waste management framework that integrates edge computing with IoT sensors and DL technologies to allow classification of biodegradable waste. The sensors and cameras are used to capture images of wastes, which are pre-processed to remove noise and resized to feed into the DL model. Key features of the wastes are then derived using L2-Hysteresis normalization with Histogram of Oriented Gradients (L2-Hys-HOG). Subsequently, biodegradable materials are accurately classified using a modified version of a Recurrent Neural Network with Gated Recurrent Unit (RNN-GRU) model. This model deploys and works on edge devices to take decisions and give feedback to users in proper waste disposal. As a result, the model achieves an accuracy of 97.79%, precision of 97.80%, recall of 97.79%, and F1 score of 97.78%. Moreover, an edge computing solution with a computational capacity of 120&#xa0;GHz saves energy up to 29.7 Joules, which can enhance waste sorting efficiencies and support real-time decision-making in waste classification tasks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Edge computing based deep learning framework for biodegradable waste classification in smart waste management systems

  • 期刊论文

  • 发表日期：

    2025年11月26日

查看全文
  • M. Thanjaivadivel,
  • T. Rajasekaran

摘要

The deployment of smart waste management systems is crucial for protecting the environment against pollution through proper recycling or landfilling. Nonetheless, Deep Learning (DL) and Internet of Things (IoT) promise a solution in automating waste classification. Meanwhile, low image quality and lesser classification accuracy are still posing a challenge. This paper presents a novel smart waste management framework that integrates edge computing with IoT sensors and DL technologies to allow classification of biodegradable waste. The sensors and cameras are used to capture images of wastes, which are pre-processed to remove noise and resized to feed into the DL model. Key features of the wastes are then derived using L2-Hysteresis normalization with Histogram of Oriented Gradients (L2-Hys-HOG). Subsequently, biodegradable materials are accurately classified using a modified version of a Recurrent Neural Network with Gated Recurrent Unit (RNN-GRU) model. This model deploys and works on edge devices to take decisions and give feedback to users in proper waste disposal. As a result, the model achieves an accuracy of 97.79%, precision of 97.80%, recall of 97.79%, and F1 score of 97.78%. Moreover, an edge computing solution with a computational capacity of 120 GHz saves energy up to 29.7 Joules, which can enhance waste sorting efficiencies and support real-time decision-making in waste classification tasks.