Stereo accuracy for collision avoidance

In a stereo configuration, the measurable disparity values are integral, therefore the measurable depths are discrete. This can create a trap for a safety system whose purpose is to estimate the trajectory of a moving object, and issue an early warning. Accuracy of this estimation is determined by the samples which have different measurable depths. Change in measurable depths becomes obvious for closer regions, but due to the limited extent of the stereo common field of view for these regions, the object might not be in the common field of view. A velocity estimation algorithm has been created, which takes into account the constraints of stereo, while accurately estimating the object´s trajectory. From examination of various scenarios, we show that a stereo system could be misleading while estimating the object trajectory: it could predict that an object on a collision course is `safe´.