Machine vision techniques for the detection and segmentation of red grape fruits under uncontrolled lighting

Abstract In recent years, the growing demand for automation in grape harvesting has highlighted the need for efficient, reliable methods to reduce labor requirements and improve harvesting operations. Manual grape picking is labor‐intensive and time‐consuming, driving demand for automated solutions. This study presents the design and implementation of a machine vision‐based image processing approach for detecting red grape ( Vitis vinifera ) clusters in outdoor vineyard environments using standard red–green–blue (RGB) images. A dataset of 500 images was collected under natural lighting conditions, with 200 selected for quantitative performance evaluation. Using RGB color space analysis, two algorithms were developed to extract and segment grape clusters based on their color characteristics, effectively distinguishing them from background noise. The first algorithm (Algorithm 1) achieved a detection accuracy of 96.84%, calculated as the ratio of correctly detected grape clusters to the total number of grape clusters in the evaluated images. After enhanced preprocessing and adaptive thresholding, the improved Algorithm 2 achieved 98.11% accuracy using the same evaluation criterion, reducing missed detections to 15 clusters and lowering the average processing time by 0.34 s per image. Both algorithms demonstrate strong performance and adaptability under variable lighting conditions. The results confirm their potential for real‐time integration into automated harvesting systems, offering a practical and efficient solution for reliable fruit detection. This study contributes to the development of practical precision viticulture technologies by providing an accessible alternative for field‐based grape detection.