摘要
将整像素位移搜索和基于微区统计性质的亚像素位移梯度算法相结合的数字图像相关方法具有计算简单、求解效率高等优点。该方法的基本假设是变形前后的子区做刚体平移,这则与有位移梯度存在的实际情况相矛盾。首先分析了该基本假设的理论误差,在基于梯度的数字图像相关方法中,得出变形子区做刚体平移的假设和在Newton-Rapshon方法中子区均匀变形的假设所获得的变形子区中心位移在理论上为相同的结论。然后用四组实验来验证该方法在实际实验条件下的计算精度和稳定性,并在铝板试件的单向拉伸实验中,将该方法与N-R方法在有应变存在情况下的计算结果作比较,结果表明该方法计算的位移和理论位移符合得很好,但其计算速度和效率要远优越于N-R方法。
A fast and high accuracy gradient-based digital image correlation method uses only fine search at the integer pixel level resolution and the subsequent algorithm of spatial-gradient-based sub-pixel registration. In this algorithm the basic assumption is the rigid body translation of the interrogated subset, however, this is in contradiction to the real circumstances where strains exist. Theoretical error analysis showed that the assumption of subset rigid body translation in gradient-based DIC and the linear approximation of the deformation mapping in N-R method yield identical subset center displacements. Then four real experiments were pre.sented to explore the feasibility and sensitivity of this algorithm. The experiment of uniform uniaxial tension test of aluminum specimen was investigated as a state when strains exist. The displacements calculated by the proposed method were compared with the ones obtained from N-R method. The results showed that the experimental data are in good agreement with the theoretical solution, but the proposed algorithm was much faster than N-R method.
出处
《光学技术》
EI
CAS
CSCD
北大核心
2005年第5期643-647,共5页
Optical Technique
基金
高等学校博士学科点专项科研资金(20020003025)
教育部回国人员科研启动基金的自然科学基金群体项目资助(101212020)
教育部优秀青年教师资助计划项目