ABSTRACT　　The　invention　relates　to　a　three-dimensional　object　reconstruction　method　based　on　RGB-D　　camera,　　which　includes　the　acquisition　and　preprocessing　of　color　data　and　depth　data,　　the　　acquisition　and　preprocessing　of　point　clouds,　　pairwise　registration　of　point　clouds　from　　different　perspectives,　　the　closed-loop　optimization　of　overall　cycle,　　and　the　point　clouds　　surface　reconstruction　of　complete　object.　According　to　the　invention,　　a　reconstruction　system　　with　low　cost,　　good　real-time　performance,　　easy　operation　and　high　speed　can　be　provided,　　　high-quality　point　clouds　can　be　generated,　　and　the　provided　initial　registration　can　improve　　the　overall　registration　speed.　1/2　　FIGURES　　Acquire　color　data　and　depth　data　　of　object　　Denoise,　　fill　holes　and　remove　invalid　depth　edges　from　the　　depth　data　　Convert　depth　data　into　point　　cloud　data　　Segment　the　point　cloud　data,　　remove　　background,　　ground　or　experimental　platform　　point　clouds,　　remove　outliers　around　object　　data,　　and　down-sampling　the　object　data　　Pairwise　align　the　adjacent　point　　cloud　:　　initial　registration,　　accurate　　registration　　closed-loop　optimization　and　　alignment　of　the　overall　point　cloud　　Carry　out　Poisson　surface　reconstruction　by　　using　complete　object　point　clouds　to　　obtain　a　three-dimensional　model　　Fig.　1　A　flow　chart　of　three-dimensional　object　reconstruction　method　based　on　RGB-D　　camera　according　to　the　present　invention.