从理论上推导了数字图像相关(DIC)方法中应变精度的影响因素, 指出应变计算精度会随着窗口尺寸的变小而下降。为了提高图像有效分辨率, 针对细长平面试件提出了一种二维多相机全场DIC方法。该方法根据特征点检测及匹配算法定位特征点对的亚像素位置, 通过DIC方法对特征点对进行高精度配准, 利用发展的逐步优化单应矩阵方法求解图像变换关系, 得到变形前后的无缝拼接图像。分别实施了纯平移和橡胶梁三点弯两组实验。在纯平移实验中, 该方法计算得到的应变均值误差及均方根误差均在50 με以内, 验证了该方法的有效性; 采用橡胶梁三点弯实验对比该方法与三维多相机全场DIC方法, 并基于实验结果对该方法的优点与缺点进行了分析。

The factors impacting strain precision in the digital image correlation (DIC) method is derived theoretically. It is pointed out that the precision of strain calculation will decrease as the calculation window gets small. In order to improve the effective image resolution, a novel perspective to two-dimensional multi-camera full-field DIC method is proposed for slender specimen. The sub-pixel position of matching feature point pairs is found based on feature point detection and matching algorithm. The matching feature point pairs are registered with high precision using the DIC method. Finally, a gradually optimized homography solution method is developed to solve the image transformation to get the seamless stitched images before and after deformation. Two sets of experiments including pure translation and three-point bending of rubber beam are conducted. In the pure translation experiments, the mean error and root-mean-square error of calculated strain are within 50 με, which verifies the validity of this method. Two-dimensional and three-dimensional multi-camera DIC methods are compared through the three-point bending experiments. Based on the results, merits and demerits of the proposed method are discussed.