The Analysis Of Unmarked 3D Point Cloud Automatic Registration Technology

With the increasing development of computer and 3D technology, it is of great significance to establish 3D model of really objects. However, it’s impossible to obtain all the point clouds of the complex object in one time due to the restrictions of observing environment, instruments and the object’s shape. In order to obtain the complete 3D object model, the key technique of jointing multi-view point cloud data is introduced in 3D scanner researching in the paper.The existing unmarked 3D surface auto-registration methods mainly include human-assisted methods and unassisted methods. The former typically depend on auxiliary equipment or mark points, thus limiting their functionality. And the unassisted registration methods, including geometry-based registration method and texture-based registration method would result in unstable registration results when measuring objects with different surface feathers.Aiming to solve this problem, we give deep analysis for the geometry-based and texture-based registration method. By analyzing the advantages and disadvantages of the two methods, an adaptive 3D surface auto-registration mothed based on both geometric and photometric features is proposed, the main contents as following:1. The feature points are used to complete 3D surface auto-registration instead of mark points.2. An automatic data registration model is established to synthetically evaluate the complexity of surface geometry and texture. Based on this model, an appropriate registration strategy can be adaptively selected to promise a reliable registration result.3. In order to avoid the mismatching correspondences, the RANSAC algorithm is introduced to guarantee a robust correspondence.Finally, we design experiments to verify the method we presented. Experimental results with various objects indicate the effectiveness of the proposed method.