One of the major challenges in multiple particle tracking is the capture of heterogeneous movements of objects in crowded scenes. This scenario is particularly prominent in bioimaging, where intracellular structures undergo instantaneous switches between cytoplasmic diffusion and motor-mediated, fast displacements We propose here a piecewise-stationary motion model (PMM) for the particle transport along an iterative smoother that exploits recursive tracking in multiple rounds in forward and backward temporal directions. By fusing past and future information, our method, coined PMMS, can recover fast transitions from freely or confined diffusive to directed motions with linear time complexity. We complemented recursive tracking with a robust inline estimator of the search radius for assignment. We demonstrate the improvement of our technique on simulated data. On biological applications, our algorithm allows us to quantify the extremely small percentage of motor-driven movements of intermediate filament precursor particles along microtubules in a dense field of unbound particles. We also show that our algorithm can cope with a strong reduction in recording frame rate. We will conclude this presentation with preliminary results in dynamic quantification with light-sheet microscopy for 3D subcellular imaging in the context of mitosis. While these techniques allows for complete imaging of rapid intra-cellular phenomenon, new challenges arise to validate and interpret the data.