Measuring Motion

6D-Vision forms objects by grouping compatible motion vectors that are nearby. The motion is determined by averging the 6D motion information. For vehicles and cyclists, the measurements are combined in an additional Kalman filter using a suitable model to determine the rotational speed besides position, size and speed [11]. This is necessary to obtain accurate predicitons over 1-2s in order to distiguish between collision-critical and uncritical traffic situations. The image shows that the driver of the oncoming car turns his steering wheel and is about to make a turn into the path of the ego-vehicle. This situation exhibits a high collision risk!