基于差分进化的数字图像相关方法  

Digital Image Correlation Method with Differential Evolution

作　　者：潘兵[1] 谢惠民[1] 

机构地区：[1]清华大学工程力学系,北京100084

出　　处：《光电子．激光》2007年第1期100-103,共4页Journal of Optoelectronics·Laser

基　　金：国家重点基础研究发展计划资助项目(2004CB619304);国家自然科学基金资助项目(10232030;10121202;10472050);教育部优秀青年教师资助计划资助项目(20020003025)

摘　　要：提出一种基于差分进化算法的数字图像相关(DIC)方法。通过将图像子区的转角包含在相关函数中,利用具有超强全局最优化特性的差分进化算法可准确地进行相关搜索,再将相关搜索结果作为Newton-Rapshon(N-R)方法的迭代初始值进一步精确计算位移、应变信息。该方法可对存在任意刚体转动情况下的试件表面变形进行准确测量,克服了通常的DIC方法在测试对象存在稍大转角(通常认为转角大于7°)的情况下即不能使用的缺点,在被测对象既有平动、又有转动时仍然能进行准确可靠的相关计算。A new intelligent digital image correlation（DIC） method that uses differential evolution is presented, which can ac curately measure the surface displacements and strains of the deformed object,even which has a rigid body rotation with ar bitrary angle. The rotation angle is contained in the correlation function considering the influence of rigid body rotation of the interrogated subset. Due to the extraordinary ability of global optimum of differential evolution, the proposed method can complete reliable and accurate correlation-matching searching in the deformed image. Then the search results are employed as the initial guess of Newton-Rapshon（N-R） method to further improve the registration accuracy. The proposed method overcomes the shortcoming of the general digital image correlation method, which can work under the condition that the tested obiect has a rotation angle larger than seven degree.

关 键 词：数字图像相关(DIC) 差分进化 Newton-Rapshon(N-R)方法 转动 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]