The Improved Genetic Algorithms For Digital Image/Speckle Correlation Technique

Digital image/speckle correlation (DISC) is a photometric technology toaccurately determine displacement and strain. DISC has succeeded in displacementand strain measurements in recent years due to the advantages of automatic,non-contact, and full field, and has found valuable and widespread use in manyresearch and engineering applications.In this paper, the underlying principle of DISC is introduced, and color DISC ispresented. The use of colored random patterns for the calculation of the correlationcoefficient is proposed to improve the accuracy of DISC for displacement and strainmeasurements, and a new correlation coefficient based on RBG color space is definedto express the similarity of the two color speckle patterns.The key of DISC is the correlation operation. A lot of research has beenperformed in this area, and various algorithms have been proposed. However, thetraditional algorithms often require reasonable initial guess of displacement and thespatial derivatives of the digital image, which increase computation complexity. Inthis study, genetic algorithms (GAs) are proposed for DISC.GAs are stochastic optimization algorithms that were originally motivated by thenatural selection mechanism and evolutionary genetics. In this paper, the underlyingidea and application process of GAs are introduced. The performance of GAs dependslargely on the selection of the genetic operators. Various operators are tested andoptimal operators are proposed. GAs are inherently parallel, because theysimultaneously evaluate many points in the parameter space. Considering many pointsin the search space, a GA has a reduced chance of converging to the local optimumand would be more likely to converge to the global optimum. Compared withsimulated annealing (SA), GAs have great capability in global optimum. However,SA has the ability of probabilistic local searching. Therefore, the two techniques arecombined here to produce a hybrid algorithm named genetic simulated annealing(GSA) that has the merits of both genetic algorithms and simulated annealing.A computer simulation scheme is employed to test the reliability and stability ofGAs for DISC in various situations. Use it, the effect of the sizes of speckle andsubset is estimated.At last, the DISC based on improved GAs has been successfully applied tomeasure the strain-stress field in practical experiments. Comparison between the exactvalue and the results determined by the practical experiment shows that the methodpresented in this paper is a new, stable and accurate method for displacement andstrain measurements.