Adaptive hybrid segmentation combined with meta heuristic optimization in transfer learning for plant leaf disease...

Plant diseases can damage specific parts of leaves for better readability during the farming process. Plants refer to various types of crops, including fruits and vegetables. During the production phase of healthy crops, plant diseases often begin by infecting the leaves. Leaves, being exposed, are more vulnerable to disease than other plant parts. When affected by disease, crop yield decreases, leading to economic loss. Hence, early disease identification model is required and deployed in an automated computerized way. The analysis have shown that multiple approaches were executed to detect the disease, still it suffers from pitfalls like inadequate feature extraction, handcrafted features, computation burden, complexities and so on. To improve the process, an efficient method is developed for detecting various plant diseases by different learning method. Firstly, the different plant leaf data were gathered from UCI, Kaggle web sources and benchmarks. The unwanted noise in the input leaf images are pre-processed by using median filter. Subsequently, the affected or abnormal region was segmented by the adaptive hybrid K-means with fuzzy C-means clustering (AHKM-FCM); the parameter tuning is also done by improved random variable-based water strider algorithm (IRV-WSA). Finally, the segmented region was subjected into the Transfer Learning Network that was processed with Efficient-net, ResNet and Densenet, in which fine tuning of weight was accomplished by using the IRV-WSA. The model was analyzed and computed across divergent measurements. Classification output results from the proposed IRV-WSA-ETLNet model include 94.853% accuracy, 94.750% sensitivity, 94.888% specificity, and 96.068% F1 Score. Additionally, this system uses less computing time 24.378 ms. Compared to previous methods, the findings of the proposed model demonstrate improved classification rates and help the farmer to increase crop production.