IoT-based waste management: hybrid optimal routing and waste classification model.

Internet of Things (IoT) makes connectivity between physical devices which are embedded with sensors, software, and connectivity that let them to communicate and transfer data. This technology makes it possible to collect and transfer data from a vast network device, opening the door to the development of automatic and more efficiency systems. The term "waste management" refers to all of the responsibilities essential to regulate trash, from the point of gathering through reusing and monitoring. Reducing the hazardous consequences of such garbage on the environment and human health is the goal of waste management. By considering these hazardous consequences, this research work is interested in working on an efficient waste management system. The utilization of IoT devices enables municipalities to optimize waste management operations by leveraging data insights. This information aids in scheduling waste collections more effectively and planning optimal routes. Therefore, the research work proposes an IoT-based waste management system with two vital processes such as IoT routing and waste management. At first, routing in IoT is done by proposing hybrid optimization algorithm named Snake Optimization Updated Beluga Whale Optimization algorithm (SOUBWO) under constraints such as distance, energy, link quality, delay, and trust. Secondly, waste management is worked on by following steps including pre-processing, segmentation, feature extraction, and classification. The waste images collected by IoT devices are transmitted from source node (SN) to destination node (DN) by optimal routing. Those transmitted waste images are pre-processed by Wiener filtering process. Consequently, the pre-processed images are segmented by employing proposed Balanced Iterative Reducing and Clustering Using Hierarchies-Altered Distance Metrics (BIRCH-ADM) algorithm. Subsequently, features such as multi-text on histogram feature, proposed Local Gabor XOR Pattern (LGXP)-based feature using novel image processing techniques, and statistical features are extracted. Finally, these extracted features are efficiently classified by hybrid classification model which is formed by integrating conventional deep maxout and Bidirectional-Long Short Term Memory (Bi-LSTM) networks. The effectiveness of the proposed approach is validated through various analyses, including performance and statistical analyses. Moreover, the proposed scheme demonstrates minimal energy consumption, with a recorded value of 0.123. In contrast, conventional methodologies exhibit higher energy consumption, with values such as SOA = 0.237, BWO = 0.146, BES = 0.183, SMO = 0.158, CHOA = 0.174, and PSO = 0.189, respectively. By this hybrid classification model, the process of classification on waste is effectively done and moreover its effectiveness is proved by various analyses.