A FastMethodforVanishingPointEstimation and Trackingand ItsApplication inRoad Images
Zhibo Guo, Jingyu Yang · 2006
A fastmethodforvanishing point (VP)estimation andtracking inroadimagesanditsapplication forlane detection areinvestigated inthispaper. Afterdetermine the region ofinterest oftheroadimages, weuse'Sobel' operator to detect boundaries. Thentheboundary imageistransformed to theHoughSpace byHoughTransform. We proposed aGaussian Predication Modelonthebasis ofthelatest vanishing point to predicate thecurrent VP,andwe constructed an objective function including thelinear segmentintensity and the predicated VP,subsequently theoptimal position oftheVP is estimated through theleast square method. Thecomputational complexity oftheestimation algorithm is0(n),whichcan operate inrealtime. We test ouralgorithm onmanystructured roadandsemi-structured roadimages, theresults showthatit can robustly determine theVP position regardless the disturbance ofaroundpedestrian, crowdandvehicles. Moreover, experiments onPETS2001dataset showitsapplication tothe roadlanedetection forthedriver assistant system.