Abstract: Amonocularvisionbasedrelativepositionandattitudemeasurementalgorithmwaspresented，whichcombinedthereconstructionoftargetswiththemeasurementofrelativepositionsandattitudefornoncooperationtargets.Firstly，aseriesofpointsonthetargetswerereconstructedusingtheimagesobtainedintheearliermeasurementwithafastandgloballyconvergentstructureandmotionestimation.Secondly，therelativepositionandattitudewerecalculatedbyusingaKalmanfilterbasedonrecursivedepthparametrization，inwhichnewpointsobservedfromnewimageswererecoveredandaddedtothe3Dpointsettoimprovetheaccuracyandrobustnessofthemeasurement.Experimentresultshowsthattheattitudemeasurementcanachieveanaccuracyhigherthan1°,andthepositionmeasurementcanachieveanaccuracyhigherthan2cm，whichindicatesthehighprecisionoftheproposedalgorithm.

Key words: targetreconstruction, measurementforrelativepositionandattitude, monocularvision, noncooperationtarget