s p ectraldimensionreductionmethodb ybuildin go pp onentsi g nals

therearedefectsinthetraditionals p ectraldimensionreductionmethodbasedonp rinci p alcomp onentanal y sis;reconstructeds p ectracurvesareonl yamathematicala pp roximationcomp aredtoori g inals p ectracurves,butvisualcolorre p roductionisnoto p timal.anewmethodcalledbuildin go p-p ositesin g leforreducin gs p ectraldimension;borsd,isp resented.thenewmethodcombineshumanvisualcharacteristicsandas p ectrallowdimensionlinearmodel.thefirstthreebasisfunctionsarebuiltb yo pp onenttheor yandusedtocharacterizecolorimetricinformationofreconstructeds p ectra，theremainin gbasisfunctionsarederivedb ytheresiduals p ectrabasedonp rinci p alcomp onentanal y sisandusedtoimp roves p ectralaccurac yofreconstructeds p ectra.ex p erimentsshowthemodelbuiltb yborsdimp rovescolorimetricaccurac yofreconstructeds p ectraandcolorstabilit yunderchan g in gli g htsourceswherethecases p ectralaccurac yofthemodelhasnotbeenlowered.