Two-resolution Structured Light System And View Planning

As the most view planning methods are based on single-resolution measurement system,the compromise should be made between measuring speed and measurement fidelity.Thus two cases may appear in the reconstruction of objects:First,it's fast to measure objects by low-resolution measurement system,but the fidelity is low.Second,high fidelity will be met when objects are reconstructed by high-resolution measurement system,but the measurement process is inefficient.In order to prevent the contradiction between the measurement speed and measurement fidelity,a two-resolution structured light system is proposed,which is set up by adding a Short Focal Length(SFL)camera to a structured light system.Based on the two-resolution system,a novel view planning method is proposed for fast object reconstruction.Firstly,the structured light system with the SFL camera is used to acquire the overall morphology of the measured object,for that it has a large measuring area and thus fast measuring speed.Secondly,after the analysis of the surface features of the object point clouds,some important details of the object surface are measured again with high resolution by the structured light system with the Long Focal Length(LFL)camera.Therefore,the measured object surfaces are acquired with an elaborate description to surface features and a fast acquiring speed.The main results of this dissertation are as follows:Firstly,after comparing with the state of the art of the structured light system's key technologies,such as Relative Phase Calculation,Phase Unwrapping,Stereo Reconstruction and System Calibration,a typical structured light system with a LFL camera and a projector is built.Secondly,a SFL camera is added to the typical system,and thus a new structured light system is built with the SFL camera and the existing projector.Thus,a two-resolutions structured light system is built.Thirdly,a novel view planning method is proposed based on the two-resolution property of the measurement system.The view planning method can be described as follows:(1)Next Best Views(NBVs)are calculated by the visibility classification of the surface points and clustering method named Mean-shift algorithm,and then the overall morphology of the measured object can be acquired;(2)Feature regions of object surface are obtained by analyzing the point clouds acquired by the structured light system with the SFL camera.And then these regions will be measured by the high-resolution system with the LFL camera;(3)The two types of point clouds are merged and redundant point clouds are removed to obtain the integrated point clouds of measured object.Therefore,the view planning method could not only reach a fast measurement speed,but also possess a high fidelity to the complex regions of the measured object.Finally,the feasibility and effectiveness of the proposed view planning method is verified by the reconstruction of several objects.