Plant leaf disease detection utilizing advanced learning methods can minimize significant losses in quantity and quality of crop production. However, the current detection methods have drawbacks in handling noisy and blurred images for accurate plant disease detection. A novel Weighted Myriad Filterative Regressed Gradient Descent Stacked Ensemble Classification (WMFRGDSEC) method has been developed to overcome the existing limitations and provide accurate detection. In this method, the initial pre-processing is performed on the collected wheat leaf images to eliminate noise and image enhancement using the Weighted Myriad filtering technique. Then, the image regions are segmented using Prevosti’s concordance correlative Range segmentation, and the features are extracted by the Kriging regression technique to reduce the detection time. Finally, the accurate classification is done using a Gradient Descent stacked ensemble classifier through several weak learners. Evaluated over 4000 public wheat leaf images from LWDCD2020 dataset, this WMFRGDSEC method increased the disease detection accuracy to 90.7% with a reduced false-positive rate of 8.2% and a detection time of 61.11 seconds.

Plant Leaf Disease Identification, Weighted Myriad Filtering, Prevosti’s concordance correlative Range segmentation, Kriging regression, Gradient Descent stacked ensemble classifier.