Theatrical And Experimental Research On Points Cloud Registration Based On 3D Laser Scanning System

On-site measuring technology of large forgings plays a key role in open die forgings process. Non-contact measuring instruments for forging measurement have the advantage of fast online measurement, high precision and stable repeatability over contact ones. Laser range sensor is a type of non-contact measuring tools, which can be applied to varying industrial fields with respects to scene reconstruction, registration and navigation, object recognition and tracking, etc. And it's especially suitable for fast online measurement of large objects in complicated working condition such as the open die forgings process in hot status.To implement multidirectional and wide-range measurement of forgings, it's necessary to collect points cloud datasets of forging surfaces from multiple views. So the registration of different points cloud datasets, which contributes effectively to the complete surface reconstruction, plays an indispensable role in measuring application. Therefore, researches on feature point/line primitives'extraction, matching and registration of points cloud datasets are essential procedures to surface reconstruction and dimensional measurement. With regards to points cloud registration of 3D laser measurement system, the dissertation is focusing on:1,Points cloud registration algorithm based on matching axis of cylinder. With respect to the geometry that any cross section of axial forgings or vessel course is in the form of ellipse, the segmented dataset in the scanning plane can provide the centroid of ellipse located on the axis, which is the feature line to be matched in the proposed registration method. So by working with forging manipulator, points cloud of axial forgings without the surface features can be registered around the axis.2,Points cloud registration algorithm based on the feature points extracted from scanning planes. On the analysis of 3D laser measurement system's scanning both from bottom to top and from right to left, the advantage of geometrical topological information of datasets can provide the feature points extraction method in both vertical and horizontal scanning planes using hierarchical clustering method, which is easier than surface fitting procedure. Thus all the extracted points belonging to the feature lines of edges are the foundation of points clouds registration after being matched. And the overlapped datasets between the edges can be accurately registered by ICP algorithm with higher precision.3,Points cloud registration based on differential geometries of moving least surfaces(MLS). The differential geometry of surface implicit functions provides geometric feature invariants of points cloud datasets. By introducing the constraints of sequenced distance vectors between features, the point primitives can be mapped in space without information of forging manipulator's working conditions, which is effective and flexible in the on-site measuring application.Considering on-site non-contact measurement of large-scale forgings as research background, three kinds of points cloud registration algorithms in the context of both intrinsic and extrinsic feature properties have been proposed and verified by experiments, in the framework of scanning row by row or column by column. The proposed algorithms provide the foundation for points cloud merging and surface reconstruction of 3D laser measurement system and broaden its application with widely scientific value and practical significance.