Vision BasedLanedetection forActive Security inIntelligent Vehicle
Fuqiang Liu, Wen-Hong Zhu · 2006
This paper presents amonochromic camera lane canapproximate normalroadstructures, itstill cannot detection algorithm basedonSnakeforimproving activedescribe somecases, i.e. anSturn. Adeformable template security inintelligent vehicle. Thealgorithm allows the methodwasproposed byoptimizing alikelihood function elimination ofseveral commonassumptions: flatroad,basedonthismodel.However, thisalgorithm cannot constant pitch angle orabsence ofroll angle, whicharethe guarantee a global optimumandtheaccuracy, without limitations ofmanyformeralgorithms. Moreover, compared .. . withtheformer methods, thesnakebased lanemodelisableto requirig huc ationa mresourcs. describe anyarbitrary shape byasetofcontrol points. Weuse Theapproach based onmorphological filterig hasbeen an improved Houghtransform methodto wipeoff suggested (8)(9). Thistechnique usedthemorphological disturbances caused byshadow oredgeofother things. Then watershed transformation tolocate thelaneedges inthe wegettheaccurate vanishing points using amethod ofmutualintensity gradient magnitude image, although this technique comparison todesert somebadones. Themethodofupdatehastheadvantage ofnotrequiring anythreshold forthe andtracking isalsoreferenced atlast. Theresults of gradient magnitudes, ithasthedisadvantage ofnotimposing experiments, whichhavebeendoneinurbanroad, showthat anyglobal constraints onthelane edgeshapes. thealgorithm isadapted tomanyroadconditions. Evenina Inthis paper, wepresent analgorithm that isable todetect complex driving environment, italsohasagoodperformance. andfollow theroadsquickly androbustly. Inthefirst step, Keywords-Cannyedgedetection; Houghtransform; weuseCannyedgedetection togettheedgeoftheimage. Snake; lane detection andtracking Thenwefind thecenter oftheroadboundaries tolocate the road. Thirdly, animproved Houghtransform isusedtofit theroadineachsection withstraight lines. Compared with