<![CDATA[Leaf Area Index (LAI) is an important parameter for describing crop canopy conditions, but LAI estimation based on downward-looking Digital Cover Photography (DCP) strongly depends on the successful segmentation of vegetation and non-vegetation pixels. This study aimed to evaluate the sensitivity of seven thresholding methods, namely Otsu, Yen, Li, Triangle, Isodata, Mean, and Minimum, to Apparent LAI in sweet corn canopy. RGB images that passed quality control were converted into the CIE Lab* color space; subsequently, the a* channel was transformed and inverted before segmentation. Apparent LAI values were calculated using a gap fraction approach, the Beer-Lambert formulation, a 7 x 7 grid subdivision, and clumping index correction. The results showed that Otsu, Li, Isodata, and Mean were the most stable methods because they had complete sample numbers and data distributions that met the normality assumption. In contrast, Yen and Triangle produced extreme values, whereas Minimum showed processing limitations with only 80 valid data points. The sensitivity analysis showed that the moderate scenario had lower variation than the extreme and complete scenarios. The RM ANOVA results showed significant differences among the stable methods, with F = 345.4141, p-GG = 0.0000, and ng^2 = 0.43902. Post hoc testing showed that Li and Isodata were not significantly different, whereas the other method pairs differed significantly. These findings confirm that the choice of thresholding method strongly determines Apparent LAI estimation, with Li and Isodata as the most consistent methods, Otsu as a moderate comparator, Mean as a conservative method, and Yen, Triangle, and Minimum as diagnostic methods.]]>
