Leaf image assisted plant health monitoring is one of the common practices to ensure the healthiness of the plant. Recently, the Artificial Intelligence (AI)-based plant health monitoring based on the leaf image is commonly considered by the researchers to develop Automatic Plant Health Monitoring System (APHMS). The proposed research aims to develop a deep learning-based APHMS to examine the rice plant based on its leaf images and this scheme consist the following stages; image collection and resizing, feature extraction using NASNet-model, feature optimisation with Moth-Flame Optimisation Algorithm (MOA), and classification and validation. In this work, the performance of the proposed APHMS is verified using the conventional and the optimized features and the classification outcome of the optimised features is better. The experimental outcome of this study confirms that the K-Nearest Neighbour (KNN) classifier-based method helps to provide the detection accuracy of 100% on the chosen data when the optimised-feature-based detection is executed.