Study On 3D Measurement Of Single-frame Surface Structured Light Based On Digital Speckle Projection

The 3D data of the object surface can provide effective data support for the evaluation of forming performance,the quality control of forming and the improvement of forming process.Surface structured light 3D measurement is widely used due to its high resolution,high speed and high precision.The 3D measurement method of multi-frame surface structured light is adopted at present,and it has limited application range,and it is difficult to realize real-time measurement of dynamic scene.In order to measure complex objects quickly and accurately,3D measurement of singleframe surface structured light based on digital speckle projection is studied in this paper.This method adopts single-frame projection,so it doesn’t need switch projection pattern,and it is easy to code and decode.It shows a great application potential in dynamic and handheld measurement.The research of this paper is as follows:The principle of 3D measurement technology based on digital speckle projection is studied.This paper introduces the imaging principle of camera,including the frequently-used coordinate system and the perspective projection model of camera.Then the binocular stereo vision system is introduced,and the principles of epipolar geometric,stereo correspondence and triangulation measurement are discussed in detail,and it is analyzed how to retrieve the3 D shape of the object using single frame images.It emphasizes how to acquire the disparity map by stereo correspondence is the key of the 3D measurement based on digital speckle projection,and it is the basis of this paper.In order to solve the stereo matching problem in the 3D measurement based on digital speckle projection,the local stereo correspondence based on digital image correlation is proposed.ZNCC(Zero-mean Normalized Cross Correlation)correlation criterion is utilized as the correlation criterion to measure the similarity of the subset,and then it acquires the initial disparity map using the winner-takes-all strategy.The uniqueness check,L-R check and the connected region method are used to remove the wrong disparity due to noise and occlusion.The IC-GN(Inverse Compositional Gauss-Newton)algorithm based on first order shape function is proposed to improve the accuracy of the disparity.The experimental results show that the spherical fitting standard deviation of the standard sphere reconstructed by this algorithm is no more than 0.05 mm,and the absolute error of the distance between the center of two standard spheres is 0.0709 mm.and it satisfies the accuracy requirement of 3D measurement based on digital speckle projection.This algorithm can reconstruct complex objects accurately and completely,but it takes a long time.The semi-global stereo matching algorithm based on Census transform is studied because the local stereo matching algorithm based on digital image correlation takes a long time.The matching cost based on Census transform is utilized to measure the similarity of pixels,and it is simple to calculate and efficient to calculate.The eight-path cost aggregation is utilized to enhance the robustness of matching cost.The winner-takes-all strategy is adopted for the disparity space image(DSI)to obtain the initial disparity map.The wrong disparity is removed by the above method.IC-GN algorithm based on first order shape function and The curve fitting method are proposed separately to improve the accuracy of the disparity.Experimental results show that the spherical fitting standard deviation of the standard sphere reconstructed by the semi-global stereo matching algorithm based on curve fitting method is higher than0.19 mm,and the absolute error of the distance between the center of two standard spheres is0.076 mm.The spherical fitting standard deviation of the standard sphere reconstructed by the semi-global stereo matching algorithm based on IC-GN algorithm is no more than 0.06 mm,and the absolute error of the distance between the center of two standard spheres is 0.060 mm.Both of these algorithms can reconstruct complex objects completely and densely,and it takes a shorter time with a lower accuracy.Finally,the experimental results of three stereo matching algorithm are compared.These three matching algorithms can solve the stereo matching problem of 3D measurement based on digital speckle projection with high precision.The semi-global stereo matching algorithm based on IC-GN algorithm can not only recover the 3D shape of the object accurately,completely and densely,but also takes less time.For general application,it is recommended to use it as the stereo matching algorithm of single-frame surface structured light based on digital speckle projection.