As　one　of　the　core　components　of　a　turbine,　the　quality　of　the　blade　manufacturing　has　a　strong　impact　on　the　energy　conversion　efficiency　of　the　turbine,　where　the　key　technology　of　quality　evaluation　of　blades　is　point　cloud　registration.　However,　with　the　application　of　structured　light　three-dimensional　measurement　technology　in　full　profile　measurement,　the　typical　point　cloud　registration　methods　only　focus　on　the　minimization　of　surface　profile　error,　ignoring　the　position　error　relative　to　reference　datum,　which　can　easily　lead　to　the　misjudgment　of　qualified　blades.　In　this　paper,　a　new　blade　error　evaluation　method　is　presented　to　register　the　point　cloud　data　scanned　from　a　physical　blade　to　its　theoretical　CAD　model,　which　fits　the　two　surfaces　based　on　parameter　priority.　Firstly,　qualified　blades　are　quickly　selected　after　global　fine　registration　using　the　best-fit　algorithm.　Subsequently,　based　on　the　priority　of　position　error　parameters,　the　coordinate　descent　algorithm　combined　with　the　minimum　zone　criterion　is　adopted　for　local　fine　registration,　which　guarantees　accurate　evaluation　results.　Finally,　the　shape　and　position　error　of　the　actual　blade　is　obtained　accurately　by　calculating　transformation　parameters　of　registration　and　the　deviation　between　the　registered　point　cloud　and　its　CAD　model　of　the　blade.　Experimental　results　show　that　compared　with　state-of-the-art　registration　methods,　the　presented　method　gives　higher　priority　to　the　parameters　which　are　difficult　to　finish　or　repair　by　machining,　and　the　position　errors　are　controlled　in　the　tolerance　area,　which　effectively　reduces　the　misjudgment.　In　addition,　evaluation　results　of　blade　errors　with　the　method　are　mainly　reflected　in　profile,　which　is　valuable　for　guiding　the　blade　finishing　or　trimming　in　practice.