The　traditional　luminance　meter　takes　too　much　time　to　detect　the　backlight　luminance　so　that　it　is　hard　to　take　bulk-production　in　the　engineering　practice.　To　solve　these　problems,　the　paper　puts　forward　an　evaluation　method　of　backlight　luminance　uniformity　based　on　ε-support　vector　regression.　A　backlight　luminance　uniformity　testing　platform　is　designed.　First　of　all,　a　CCD　image　sensor　is　used　in　a　single　shot　to　get　a　digital　image　of　several　backlights,　and　the　image　is　corrected.　Secondly　uses　ε-support　vector　regression　(ε-SVR)　in　the　condition　of　nonlinear　to　test　the　luminance　values　of　check　points　in　backlights,　and　forecast　the　real　luminance　values　of　backlights.　At　last,　it　is　compared　with　the　testing　standard　to　judge　whether　the　backlights　are　qualified.　The　results　show　that　the　fusion　of　parameters　can　be　tested　in　the　constructed　support　vector　regression.　And　their　accuracy　shows　good　similarity.　RMSE　of　the　best　prediction　is　43.28,　BR　is　1.21%.　The　error　between　the　predicted　results　and　these　results　of　the　measurements　of　the　traditional　luminance　meter　is　much　less,　and　it　is　easy　to　operate.　The　evaluation　results　of　luminance　meter　are　accorded　with　the　method　of　the　objective　evaluation　of　the　backlight　luminance　uniformity.　©　2015,　Science　Press.　All　rights　reserved.