A Preliminary Method for Tracking In-Season Grapevine Cluster Closure Using Image Segmentation and Image Thresholding

Figure 2

Examples of original images (a–c) in the dataset exhibit complexities such as varying zoom levels, the presence of unwanted objects (e.g., a hand), or similar objects in terms of shape and colour (e.g., canopy and clusters other than the intended one). These complexities have the potential to affect the segmentation process. Correspondingly, images (a1–c1) represent the segmented cluster boundaries generated using PSPNet, demonstrating its robustness in overcoming these challenges.