Geographic object-based image analysis (GEOBIA) with very-high-resolution (VHR) images plays an important role in geographical investigations, but its uncertainty of segmentation scale always affects the accuracy and reliability of its results, e.g., object segmentations and classifications. Therefore, a scale-selection method is needed to determine the optimal scale for GEOBIA, which, however, can be influenced by three factors, i.e., categories, surrounding contrasts, and internal heterogeneities of objects. So if we want to select the optimal scale, the three factors should be totally considered. The existing scale selections including supervised and unsupervised methods partly considered these three factors, but could not resolve all of them, thus this issue is still open and needs further study. This chapter reviews five kinds of scale-selection methods, compares their advantages and disadvantages, and then discusses the future direction of scale selection. We also point out the advantages and limitations of each scale-selection method, and suggest that these methods should be utilized together to finally determine the optimal scales of objects so as to improve the accuracies of object segmentation and classification.