基于时序变形预测的数字图像相关加速方法

利用数字图像相关求解连续变形物体的位移场和应变场时,会遇到处理速度非常慢的问题,原因是相关算法计算量大且忽略了物体在时间轴上的变形规律。本文提出了一种基于变形预测的数字图像相关方法,该方法利用物体在时间轴上的变形规律,通过已经得到的变形值来预测后面时刻的变形初值。首先分析了物体在时间轴上的变形规律,然后结合实际应用对其进行修正,得到物体下一步变形的初值,最后通过NR方法(Newton-Raphson Method)得到物体的位移场和应变场。通过计算机模拟和金属试件拉伸实验并结合GPU(Graphic Processing Unit)编程验证了该方法的有效性。计算机模拟和金属试件拉伸实验的计算结果表明,该方法能够在计算精度保持不变的情况下,使计算速度提高4~7倍。

Acceleration Method of Digital Image Correlation Based on Timing Deformation Prediction

When digital image correlation is adopted to solve the displacement and strain field of an object in continuous deformation, processing speed is very slow due to the large calculation amount of correlation algorithm and the neglect of object deformation on the timeline. In this paper, a new digital image correlation method is proposed based on deformation prediction. By using the object deformation on the timeline, this method can predict the initial values of next step deformation based on obtained deformation values. The object deformation on the timeline is analyzed firstly, then combined with practical application, the deformation is revised to obtain the initial value of object deformation in next step, finally the displacement and strain field are calculated through Newton-Raphson method (NR method).Computer simulation and metal tensile experiment were conducted to verify the validation of this method combined with GPU (Graphic Processing Unit) programming. Results show that keeping the same computation accuracy, this new method can improve processing speed 4 to 7 times.