Line-based Landmark Recognition forSelf-localizatio n of
Abdul Baist, Robert SablatnigN · 2005
Thispaperl describes visual landmark extraction for self localization ofan auitonomolus mobile robot inatwellknowndynamic environment. Thegradient basedHough transformprovides thestrongest groupings ofcollineair pixelshaving roughly thesame edgeorientcation. Grouipsof pixels arethen processed tocalcuilate length andendpoints ofline segments,which together with thelength anid direction ofthenormcal completes thedescription oftheline segmnents. Thisisfollowed byclassification offield markinigs which arebright lines andaircs on adarkbackground formingdouble edges ofopposite gradient. Corners, junctions andline intersections aredetermined bytheinterpretation ofdetected line segments.Simulation results illustrate the performance ofthemethodu(sing bothreal andsYnthetic images.