Improved ground plane detection in real time systems using homography

Advanced Driver Assistance Systems (ADAS) are used to assist driver by providing real-time information about cross traffic, parking space and so on. Such applications require a fair amount of processing to be done on the incoming video feed. It has been proved that the homography approach is efficient for ground plane detection. But, in the case of Road Plane Detection, the lane markings appear as obstacles when the SAD approach [1] is used. We propose a solution to this problem which involves using a Road Model and applying lane masking. We are able to get an accuracy of 88% to 96%. This technique is suitable for real time applications and we also provide benchmarks on a DSP processor, TMS320C6000 for the key functions used in this algorithm.