SLAMisoneofthemostimportantcomponentsinrobotnavigation.ASLAMalgorithmbasedonimagesequencescapturedbyasingledigitalcameraisproposedinthispaper.Bythisalgorithm,SIFTfeaturepointsareselectedandmatchedbetweenimagepairssequentially.Afterthreeimageshavebeencaptured,theenvironment’s3Dmapandthecamera’spositionsareinitializedbasedonmatchedfeaturepointsandintrinsicparametersofthecamera.Arobustmethodisappliedtoestimatethepositionandorientationofthecameraintheforthcomingimages.Finally,arobustadaptivebundleadjustmentalgorithmisadoptedtooptimizetheenvironment’s3Dmapandthecamera’spositionssimultaneously.Resultsofquantitativeandqualitativeexperimentsshowthatouralgorithmcanreconstructtheenvironmentandlocalizethecameraaccuratelyandefficiently.
