Design and Development of End to End AIoT Based Solution to Improve Crop Yield

Abstract

newline newlineAbstract newlineThe fast growth of the agricultural sector requires innovative keys to address prob newlinelems such as low productivity, resource inefficiency, and the need for timely decision newlinemaking support for farmers. Motivated by these challenges, this research introduces newlinetwo novel systems AIoTST-CR and VotTomNet to enhance crop recommendation newlineand disease diagnosis, thereby contributing to sustainable and precision agriculture. newlineTheproposedAIoTST-CRsystemintegrates soil testing hardware with AI models to newlinerecommend the most appropriate crops based on soil parameters. This approach signifi newlinecantly reduces input costs and fertilizer usage while improving crop yield. Performance newlineevaluation revealed that Bayes method and Random Forest attained the maximum level newlineof accuracy, making sure that crop suggestions are trustworthy. A dedicated mobile ap newlineplication, Soil2Crop, was also developed to provide farmers with an intuitive platform newlinefor accessing actionable insights. newlineIn parallel, the VotTomNet framework addresses tomato leaf disease diagnosis by newlineensembling six up-to-date pre-trained deep learning models ( InceptionNet, VGG16, newlineDenseNet, MobileNet, ResNet, and EfficientNet). The ensemble strategy, employing newlinetransfer learning and soft voting, achieved a robust classification accuracy of 98.61%, newlineexhibiting better performance than individual models. newlineOverall, the research establishes that the proposed systems not only surpass current newlinehigh-tech solutions in terms of precision and usefulness but also empower farmers with newlinecost-effective, technology-driven tools for decision support. These contributions hold newlinesignificant promise in improving agricultural productivity and advancing sustainable newlinefarming practices. newline