Multi-Model Machine Learning for Automated Identification of Rice Diseases Using Leaf Image Data

Abstract

AbstractPurposeRice is grown almost everywhere in the world but is notably prevalent in Asian nations where it serves as the main food source for nearly half of the world’s population. Yet, enduring agricultural problems like various rice diseases have been a problem for farmers and planting specialists for ages. A fast, efficient, less expensive, and reliable approach to detecting rice diseases is urgently required in agricultural information since severe rice infections could result in no harvest of grains. Automated disease monitoring of rice plants using leaf images is critical for transitioning from labor-intensive, experience-based decision-making to an automated, data-driven strategy in agricultural production. In the modern era, Artificial Intelligence (AI) is being widely investigated in various areas of the medical and plant sciences to assess and diagnose the types of diseases.MethodsThis work proposes a hybrid deep-machine learning system for the automated detection of rice plant diseases using a leaf image dataset. Benchmarked MobileNetV2 architecture is employed to extract the deep features from the input images. Obtained features are fed as input to various machine learning classifiers with different kernel functions using a 10-fold validation strategy.ResultsThe developed hybrid system attained the highest classification accuracy of 98.6%, specificity of 98.85%, and sensitivity of 97.25% using a medium neural network. The results demonstrate that the established system is computationally faster and more efficient. The proposed system is ready for testing with more databases.ConclusionsThe suggested technology accurately diagnoses various rice plant illnesses, reducing manual labor and allowing farmers to receive prompt treatment. Future research topics include incorporating cloud-based monitoring for leaf image capture in non-connected farms, as well as building mobile IoT platforms for continuous screening.