Advanced Detection of Coconut Stem Bleeding Using an Improved YOLOv5 Model for Precision Agriculture

Coconut stem bleeding is a critical agricultural challenge threatening the health and productivity of coconut trees, a vital cash crop in tropical regions and a primary source of income for many farming communities. Early and accurate detection is essential to mitigate agricultural losses and promote sustainable farming. This study introduces a novel YOLOv5-based framework for real-time detection of stem bleeding in coconut trees, prioritizing accuracy, scalability, and practicality.The methodology integrates advanced preprocessing techniques like histogram equalization and 2D wavelet transformation with a dataset enhanced using Mosaic data augmentation for improved diversity. Adaptive anchor box calculation and grid sensitivity elimination enhance detection precision and localization accuracy. Adaptive image scaling ensures uniform input dimensions, optimizing inference speed without compromising performance. These innovations address challenges such as object size variability, lighting conditions, and real-time efficiency.The YOLOv5 framework achieved 97% detection accuracy, a mean average precision (mAP) of 96.23%, precision of 87.42%, recall of 95.32%, and an F1-score of 83.65%. Comparative analysis with CNN, VGG16, and traditional image processing highlights its superiority in accuracy and real-time applicability.This framework empowers farmers to implement immediate interventions, reducing disease spread and crop losses. Beyond technical contributions, it offers economic and social benefits by reducing yield losses and management costs, fostering financial resilience for farmers. Its adaptability to other crops underscores its global relevance, setting a benchmark for intelligent, data-driven solutions in precision agriculture.