Text this: View adaptive multi-object tracking method based on depth relationship cues