Abstract : Golgi outposts (GOPs) that transport proteins in both the anterograde and retrograde directions play an important role in determining the dendritic morphology in developing neurons. To obtain their heterogeneous motion patterns, we present a data association based framework that first detects the GOPs and then links the detection responses. In the GOP detection stage, we introduce a multi-scale Markov Point Process (MPP) based particle detector that uses multi-scale blobness images obtained by Laplace of Gaussian (LoG) for GOP appearances. This reduces the number of missed detections compared to the use of image intensity for GOP appearances. In the linking stage, we associate detection responses to form reliable tracklets and link the tracklets to form long, complete tracks. As such, high-level information (e.g., motion) is encoded in building the affinity model. We evaluate our approach on the microscopy data sets of dendritic arborization (da) sensory neurons in Drosophila larvae, and the results demonstrate the effectiveness of our method.