Recognizing Hazards

6D-Vision is a „track before detect“ algorithm, i.e. a motion is determined on an image point level, before an object is formed by grouping. This avoids typical errors of other stereo methods that first form objects and then try to determine motion by tracking the object. Especially partially occluded objects, objects close together in space, and difficult viewing conditions are situations where the power of 6D-Vision becomes apparent. Based on the 6D information, one can determine the collision risk for every single image point, independent of an object detection step , as seen in the image on the right. Red arrows depict imminent collision risk, yellow a close encounter, and green no collision risk. In that scenario an automated emergency brake was initiated to avoid a collision .