IoT-enabled deep learning for weed species classification using MapReduce and optimization algorithm

doi:10.1016/j.compeleceng.2026.111484

IoT-enabled deep learning for weed species classification using MapReduce and optimization algorithm

2026-09-19

0

PRE

AI

DOI：10.1016/j.compeleceng.2026.111484

Original

Original request for help

Save

Abstract

En 中文

Weeds are unwanted plants that damage the growth of crops and affect the health of humans or farmed animals. Classification of weeds is a persistent problem in agriculture due to various reasons, such as inter-class similarity among weed species and intra-class dissimilarity among similar weed species images at various growth stages. Hence, Beluga Whale Ring Optimization (BWRO)-PyramidNet is newly proposed for weed classification. Initially, the Internet of Things (IoT) simulation is carried out, and the corresponding routing process is done by BWRO. Next, the generated weed image is fed to the MapReduce framework, which has both the mapper and the reducer phases. In the mapper phase, the input weed image is pre-processed using an adaptive bilateral filter, followed by feature extraction. Then, extracted features are merged and sent to the reducer phase, where the weed is classified using PyramidNet. Here, PyramidNet is trained by BWRO, which is the integration of Beluga Whale Optimization (BWO) and Mud Ring Algorithm (MRA). Furthermore, the efficacy of BWRO_PyramidNet is evaluated based on overall classification accuracy, True Positive Rate (TPR), True Negative Rate (TNR), F-measure, precision, energy, distance, and delay, with superior values attained as 96.36%, 95.63%, 96.12%, 95.75%, 95.37%, 0.941 J, 0.409 m, and 0.335 sec, respectively.

IoT-enabled deep learning for weed species classification using MapReduce and optimization algorithm

Abstract

Journal

Organization

Cited Papers