Turbine blade is the core part of aero-engine. It is necessary to register the measured point cloud with CAD model quickly and accurately for NC machining after precision casting. In view of the problem of the initial posit ion and orientation of the measured point cloud, two kinds of pre-alignment algorithms are introduced and implemented, which are the fast alignment based on principal component analysis and manually associated pre-alignment. As for the low efficiency of searching match point, this paper introduces and implements three searching algorithms, which are grid method, normed space projection method and fast search method based on kd–tree. Through comparative analysis, it is pointed out that the fast search method based on kd–tree has better computational efficiency. On this basis, the registration scheme of manual association pre-alignment, match point search based on kd-tree and ICP method is determined. The difference interval between the measured point cloud and CAD model after registration is [–0.06 mm, +0.1 mm], which is accord with the turbine blade casting accuracy.