| Digital image correlation(DIC)method is the most popular deformation measurement technique in the field of photomechanics.DIC technique has been widely used in both scientific and engineering measurements.With the widening of DIC applications,the measurement accuracy,computational efficiency and applicability of DIC are in urgent needs of further improvements,to meet the continuous needs of scientific and engineering fields.At present,the displacement measurement accuracy of DIC can research 0.01 pixels,which can meet the most of requirements of displacement measurement.However,compared to the traditional stain gauge technique,the strain measurement accuracy of DIC is not enough for measurement requirements in most situations.Meanwhile,affected by many factors,such as quality of speckle patterns,selection of calculation parameters and calibration of measured mechanical quantities,the consistency and correctness of the DIC measurements can't be guaranteed.Moreover,with the development of DIC technique,DIC is increasingly used in industrial online detection and medical measurement.For industrial online detection and medical measurement,the computational efficiency of DIC is vitally important,the requirement of real-time measurement needs to be met.Furthermore,for the normally used monocular and binocular measurement systems,their applicability still can't meet the practical measurement requirements,such as measurements in extreme environment,multi-scale deformation measurement and large-scale measurements.The aim of this dissertation is to achieve the goal of high-accuracy and real-time deformation measurement using DIC,from the perspectives of measurement accuracy,computational efficiency and system calibration.The major achievements of this dissertation are as follows(1)The statistical error formula of the inverse compositional Gauss-Newton(IC-GN)algorithm is analyzed in terms of noise.We find that the noise-induced bias error of IC-GN algorithm can be eliminated when proper operator is used and the mechanism of elimination of noise-induced bias error is revealed.Through the comparisons to the Newton-Raphson algorithm,the comprehensive advantages of the IC-GN algorithm in computational efficiency and noise robustness are demonstrated.The new statistical error formula of DIC is also given.(2)A seed point-based parallel method is proposed for parallel computation of DIC.With the high-efficiency IC-GN algorithm and lookup table of bicubic B-spline interpolation,real-time DIC is realized.The seed point-based parallel method was further extended to three-dimensional(3D)DIC,the temporal matching,stereo matching and 3D reconstruction are implemented step by step in child thread,the real-time 3D-DIC is realized at a frame rate of 10 frames/s with resolution of 5000 points per frame.(3)Based on the close-range photogrammetry and speckle analysis,we propose the intrinsic and extrinsic parameters calibration method.With the proposed method,the calibration of 3D-DIC in large field of view measurement is realized and high-accuracy 3D deformation measurement is achieved.(4)Based on the speckle-based self-calibration method and the reflection stereo imaging device,we propose a self-calibration single-lens 3D video extensometer for high-accuracy and real-time strain measurement.The problems of effects of out-of-plane displacement,stringent synchronization and complicated system calibration are solved.(5)We propose a camera array-based DIC method for high-accuracy full-field strain measurement.Through the high-accuracy calibrated camera array intrinsic and extrinsic parameters,sub-pixel image stitching and displacement field stitching are realized.The one hundred-million-pixel speckle image and the full-field 10 ?? strain measurement resolution are achieved. |
PDF Full Text Request