Plant Leaf Disease Detection and Classification Using MobileNetV3 and Wavelet Kernel Extreme Learning Machine

Detection of plant disease is the most important aspect in agriculture, aimed at identifying and diagnosing disease that affect plants. If the plant disease detected earlier, that helps to avoid risk from damaging plants and the soil. However, there is a difficulty to differentiate the diseased and healthy leaves due to similarity of disease symptoms. To solve this problem, an effective Wavelet Kernel Extreme Learning Machine (WKELM) classifier is proposed for distinguishing diseased and healthy plants. Images from the PlantVillage dataset are preprocessed using Bilateral filter and then segmented by Adaptive Gaussian Threshold (AGT) to separate the diseased portion from leaf. MobileNetV3 a Convolutional Neural Network (CNN) used for effective feature extraction with the hyper parameters which are tuned using effective Emperor Penguin Optimizer (EPO) and classification is done by the WKELM for accurate results of plant disease. The experimental results of the proposed model performed by the evaluation metrics achieved accuracy of 99.7
^%
, recall of 99.6%, precision of 99.6%, F1-score of 99.7% which is greater than other existing deep learning models like CNN, Transfer learning with Deep CNN models, CNN - Visual Geometry Group (VGG).