Research Of Sub-pixel Displacement Digital Image Correlation Algorithm Based On Correlation Coefficient Weighted Fitting

Experimental mechanics is an important part of mechanics. Optical measurement mechanics has the advantages of full filed, high precision, non-contact measurement. It is now developed into an important tool for research in the field of experimental mechanics. The digital image correlation is developed into the most active, the most vitality and the most widely used optical measurement method in recent years. It has received a lot of successful applications and has broad application prospects.Although the current digital image correlation algorithms can achieve higher accuracy, usually have the problems of large amount of calculation and the longer computation time. The current calculation methods of sub-pixel displacement are the use of spatial correlation, and essentially, they are use of the continuity of deformation in space. In fact, the deformation also has the continuity of time. A digital image correlation algorithm is proposed in this paper to solve sub-pixel displacement, which was based on the combination of the spatial correlation and temporal correlation.The main research results are as follows:(1) A fast integer pixel displacement search algorithm based on seed point is employed to calculate the integer pixel displacement. The results demonstrate that it can greatly improve the computational efficiency. (2) The MLS fitting algorithm with a weighted function based on the correlation coefficient is employed to calculate the sub-pixel displacement in DIC for the first time, which is the improvement and development of temporal sequence digital image correlation algorithm. Experimental results show that it is effective to reduce the calculation error and to improve accuracy. The local features are utilized during the fitting process, so simple basis functions can be used to fit the complex deformation, which means that the method is flexibility and adaptability. (3) The calculation accuracy and efficiency of the algorithm are obtained by comparing the experimental results. The average error of the calculation results is about 0.03 pixels and the standard error is around 0.06 pixels. Experimental results show that the proposed algorithm is more efficient than NR algorithm at the same accuracy case. The computational efficiency is improved from minimum 4.2 times to maximum 15.3 times in the whole process. The computational accuracy will be enhanced more obviously as the calculation subset is bigger. The efficiency has been markedly improved by at least 28.8 times and the maximum reached 253.8 times in sub-pixel displacement calculation stage, and the computational can reach 13,000 (points/ second) (4) The experimental speckle pattern is different from the commonly used simulated speckle pattern, while the results of the experiment are applied to generate the deformation results of different experiments. The speckle is the actual spray generated, that has a very good randomness. The images features can be increased due to the speckle sizes are different. Therefore the experimental speckle pattern has a higher correlation.