Drosophila border cells are a genetically tractable model to study collective cell migration. With recent advances of culture conditions ex vivo, it becomes possible to observe the dynamics of this process in time-lapse image sequences. However, the complexity and heterogeneity of cell behaviors between experiments make it difficult to relate migration phenotypes back to the underlying genotypes. In this study, we propose image analysis algorithms to identify the boundary of border cell clusters, to track boundary movements and map them into a spatiotemporal space called morphodynamic profile, in which the effect of molecular perturbations can be quantitatively characterized. To this end, we design informative numerical features and train a support vector machine classifier to recognize cell migration patterns of different genotypes. Experimental results demonstrate that our technique can effectively quantify the morphological dynamics of border cell clusters, despite the substantial experimental heterogeneity, and identify similar movement patterns among different genotypes. These methods are now being combined with genetic approaches to systematically map out, by epistasis experiments, the interaction of multiple chemotactic pathways.