Research On Real-time 3d Measurement Method Based On Structured-light And Light-filed Imaging

Optical three-dimensional(3D)measurement method,as one of the most important fields in computer vision,has been widely used in many areas,such as industry,medicine,security and so on.Optical 3D measurement methods can be classified into active methods and passive methods,depending on whether the structured-light illumination is applied or not.Generally,active methods achieve better measurement results than passive methods at the cost of multiframe structured-light projection and more measurement time,which makes the accurate realtime 3D measurement system difficult to design.Based on the combination of structured-light and light-field imaging,an accurate real-time 3D measurement prototype system is designed in this paper,which needs only single-frame structured-light projection and less measurement time.This paper is focused on the design of the accurate real-time 3D measurement prototype system,including the design of the system structure,the algorithm for system parameters calibration,the algorithm for 3D measurement,and the improvement of the prototype system.Detailed contents are described as follows.(1)The design of the system structure.Compared to the traditional camera,light-field camera records the depth information into the four-dimensional(4D)light-field,which provides necessary geometrical constraints to reduce the frames of structured-light projection and makes single-frame structured-light projection possible.(2)The algorithm for system parameters calibration.Light-field camera calibration and projector calibration are mandatory to the system parameters calibration.Available lightfield camera calibration algorithms are implemented depending on its own models,which makes the calibrated parameters physically meaningless and difficult to apply.Therefore,a new light-field camera calibration algorithm is designed,comprising of three steps.Firstly,the main lens is modeled by the pinhole model and calibrated.Then,the 4D light-field information and the depth of all calibration features are analyzed to calibrate all light-field camera parameters.Finally,all calibrated parameters are refined by a non-linear optimization algorithm.As for the projector,available calibration algorithms in traditional structured-light 3D measurement system are accurate and adaptive.(3)The algorithm for 3D measurement.A new 3D measurement algorithm is designed,based on the designed system structure,comprising of three steps.Firstly,a difference filter is designed and applied to the traditional fourier transform profilometry,which removes image background appropriately and acquires accurate wrapped phase.Then,phase variance computation is designed to remove the vignetting and improve radiance variance computation,by which the initial depth is also calculated with single-frame structured-light projection and corresponding light-field image.Finally,the wrapped phase is used again to refine the initial depth and acquire 3D point cloud.(4)The improvement of the prototype system,including how to improve the wrapped phase and how to deal with projector distortion.The wrapped phase is improved by windowed fourier transform profilometry and modified fourier transform profilometry for different sences.As for the projector distortion,a traditional camera is integrated into the designed system,the 3D measurement algorithm is also re-designed correspondingly to deal with the projector distortion.