HYBRID FEATURE EXTRACTION WITH GRADIENT BOOSTING TREE FOR PLANT DISEASE DETECTION AND CLASSIFICATION MODEL ARPITA DEODIKAR

Plant disease plays a vital role in the productivity of the agricultural sector. For the detection of plant diseases at the primary stages, automated disease diagnosis models using image processing and computer vision techniques are beneficial. In this view, this paper presents a new Hybrid Feature Extraction (HFE) with Gradient Boosting Tree (GBT) Classification Model, called HFE-GBT for Plant Disease Diagnosis. The presented HFE-GBT model comprises four major stages namely preprocessing, segmentation, feature extraction, and classification. The HFE-GBT model uses Gaussian filtering (GF) based noise removal and contrast enhancement process. In addition, grabcut based segmentation technique is applied to identify the diseased portions of the image. Followed by, a hybridization of grey level co-occurrence matrix (GLCM) and histogram of gradients (HOG) based feature extraction process is employed. Finally, GBT model is applied as a classifier to assign appropriate class labels of the input plant images. The performance validation of the presented HFEGBT model takes place against citrus and tomato leaf disease dataset. The obtained experimental values proved the improved detection performance of the HFE-GBT model with the higher average accuracy of 96.08% and 97.75% on citrus and tomato dataset respectively