Image data augmentation is a fundamental data pre-processing technique within Artificial Intelligence (AI). It comprises various operations in the image space, such as translation, rotation, reflection, and dilation. The watershed algorithm facilitates image transformation, wherein pixel values are determined. These values (n) are subsequently categorized into three distinct groups: (n≤0), (0<n <2), and (n ≥2). Following this transformation, PCA-clustering post-natural disaster building damage levels into three clusters: lightly, moderately, and severely damaged. These clusters correspond to three different value ranges: group 1 (n<0), group 2 (0≤n <2), and group 3 (n≥2) Our experiment revealed two key findings. First, image augmentation can mitigate the weaknesses of the watershed algorithm. Second, variations in image sizes do not affect the accuracy of building damage assessment after natural disasters if we utilize the watershed algorithm.