Advanced Paddy Seed Segmentation Using Unified Edge Fusion Techniques: Enhancing Precision in Agricultural Image Processing

Abstract

Accurate segmentation of paddy seeds in agricultural imagery is vital for various agricultural tasks such as crop monitoring, yield forecasting, and disease detection. This paper presents a novel approach, Unified Edge Fusion Techniques for Advanced Paddy Seed Segmentation (UEFPaddySeg), aimed at improving the precision of paddy seed segmentation. By combining several edge detection strategies and fusion techniques, the proposed method enhances the accuracy of seed boundary detection and reduces segmentation errors. The UEFPaddySeg framework integrates edge information from diverse sources to refine segmentation, effectively addressing challenges such as seed shape variation, overlapping seeds, and inconsistent lighting conditions. Experimental evaluations, comparing UEFPaddySeg with traditional methods like Kernel Graphcut and FRFCM, reveal its superior performance in key metrics, including specificity, sensitivity, Jaccard Similarity (JS), Dice Coefficient (DC), and overall accuracy. The robust results highlight UEFPaddySeg’s potential as a powerful tool for agricultural image processing, facilitating more precise data for precision farming and contributing to greater automation in agriculture.