Here, we focus on object classification of moving objects (e.g., cars, bicycles and pedestrians).
To achieve this the shape, the movement and the context of an object is used.

In a training phase these features are learned for every possible object class.
For example, a pedestrian can change direction faster than a car.
Furthermore, the context provides additional clues, like objects moving on traffic lanes are more likely to be a car than a pedestrian.