Automatic Disease Detection of Basal Stem Rot Using Deep Learning and Hyperspectral Imaging

Basal Stem Rot (BSR), a disease caused by Ganoderma boninense (G. boninense), has posed significant concern for the oil palm industry, particularly in Southeast Asia, as it potential to cause substantial economic losses. The breeding programme is currently searching G. boninense-resistant planting materials, which necessitated intense manual screening nursery track progression of development response different treatments. combination hyperspectral image and machine learning approaches high detection BSR. However, feature selection still required construct model. Therefore, objective this study establish an automatic BSR at seedling stage using pre-trained deep model images. aerial view divided into three regions order determine if there any spectral change across leaf positions. To investigate background images affect performance detection, segmented plant have been automatically generated Mask Region-based Convolutional Neural Network (RCNN). Consequently, models are utilised detect BSR: convolutional neural network that 16 layers (VGG16) trained on image; VGG16 RCNN both original results indicate with 938 nm wavelength performed best terms accuracy (91.93%), precision (94.32%), recall (89.26%), F1 score (91.72%). This method revealed users may without having manually extract attributes before detection.