Color-Based Thresholding for Early and Reliable Detection of Stem Bleeding in Coconut Trees

Countries like India, Philippines, and Brazil rely heavily on coconut trees for food and revenue. However, stem bleeding, which is a disease caused by a fungal pathogen Thielaviopsis paradoxa (TP), poses a significant threat to coconut production. This invasive species, if left unchecked, can spread to young trees nearby, causing widespread infection that can decimate the yield of coconuts. As such, early detection and cost-effective surveillance have become critical factors in the management of this disease. Through our research, we examine the effectiveness of conventional image processing methods in detecting areas infected with stem bleeding. We also propose a color-based thresholding technique, which we found to be highly reliable provided that optimal lighting conditions are maintained. Our method achieves a pixel accuracy of 80.36%, which is 8.5% higher than Adaptive Thresholding, 14.57% higher than Otsu Thresholding, and 12.48% higher than K-Means Clustering.