3D models of objects in motion may be generated using depth imaging data and visual imaging data captured at different times. A 3D model of an object in motion at a first position may be generated at a first time and projected forward to a second position corresponding to a second time. Imaging data captured from the object at the second position at the second time may be compared to the projected-forward 3D model of the object at the second position. Differences between the imaging data and the projected-forward 3D model may be used to modify the 3D model, as necessary, until an accurate and precise 3D representation of the object has been derived.