Complex Three-dimensional Point Cloud Model Automatic Registration Technology Research

With the development of computer graphics, computer vision and surveying, theaccess to information of three-dimensional model have been used widely,three-dimensional point cloud as a new way to describe the expression of surfacefeatures three-dimensional objects gets more attention. At the same time, processingtechnology for point cloud has also become one of today’s most popular fields ofstudy.The three-dimensional point cloud registration is critical in reverse engineering andplays a very crucial role in the reconstruction of three-dimensional model andpreliminary processing point cloud data. The quality of registration affects thematching accuracy and speed directly and has always been a difficult and hot researchof three-dimensional graphics technology. This paper does a thorough research oncomplex three-dimensional point cloud model of automatic registration technologybased on an aviation turbine engine blade geometry quality assessment project.In order to resolve the disadvantages of classic ICP (Iterative Closest Point)algorithm that it has High requirements on the initial position to avoid a local optima,and it has a low efficiency when processing a mass points, on the basis of classicalICP algorithm, this paper combines with initial registration algorithm of PrincipalComponent Analysis (PCA),and it uses a data reduction method of random samplingand K-D tree searching corresponding point set to reduce complexity of computingtime and improve the efficiency and precision of the traditional ICP algorithm. Thispaper also completes geometry quality assessment with error between thecorresponding points after the registration, and uses the RGB color model to visualize.The experiment shows that this algorithm can be a very good make up for theshortcomings of pure rough registration and classical ICP registration with goodmatching accuracy and speed.And the use of registration errors in two clusterscorresponding points of the point cloud between the geometric quality assessment iscompleted, generate an error file, use the RGB color model to visualize.With the contrast of experiments, this algorithm can compensate for thedisadvantage of the rough registration and the classical ICP registration with a goodmatching accuracy and speed. This algorithm can achieve the registration of complex surface parts, and improve the accuracy and efficiency of processing with complexsurface features. This algorithm can cover a wide range of automotive applications ofprecision parts, ship propellers and other parts developed in terms of research andindustrial applications for reverse engineering has important theoretical and practicalsignificance.