An Efficient Chilli Leaf Disease Classification with Trichome Pattern Extraction and Intensity Assessment Using DCLHPCLUNN

Chilli is one of the most significant spice crops all over the globe.Moreover, it is vulnerable to numerous leaf diseases, thus degrading the yield efficiency.Therefore, to ensure sustainable chilli production, an accurate Chilli Leaf Disease (CLD) classification is vital.Nevertheless, none of the existing CLD detection studies concentrated on examining the Trichome Pattern (TP), thereby depicting poor disease monitoring.Hence, an enhanced CLD classification with TP extraction and intensity evaluation utilizing Deep Convolutional Log Hessian Penalty Collapsing Linear Unit Neural Network (DCLHPCLUNN) is proposed in this article.Initially, the chilli leaf image is gathered.The collected chilli leaf image is further pre-processed for Noise Removal (NR) and Contrast Enhancement (CE).Afterward, the Background Removal (BR) is carried out, followed by Vein Segmentation (VS) and Non-Vein Region Extraction (NVRE).By performing the Red-Green-Blue (RGB) to Lightness green-red Axis Blue-yellow axis (LAB) color space conversion, BiNarization (BN), and SKeletonization (SK), the trichome is isolated from the extracted nonvein region.Thereafter, the TPs, namely trichome density and length, are extracted.In the meantime, from the segmented vein and background-removed images, the essential features are extracted.Then, by utilizing the Goat Quintic Optimization Algorithm (GQOA), the optimal features are chosen.After that, the chilli leaf is effectively classified into healthy and several disease types by the proposed DCLHPCLUNN.In addition, the disease area is segmented for the disease-affected leaves, followed by intensity evaluation.Hence, the proposed work had superior CLD diagnosis outcomes with an accuracy of 98.58778%, precision of 98.488%, and recall of 98.589%, demonstrating robust disease monitoring capability.