Digital image correlation method based on hybrid genetic algorithm

作者：Cheng, Zhen Jin; Wang, Quan Bao^*; Chen, Ji An; Zhao, Hai Tao

来源：Optics and Precision Engineering, 2016, 24: 710-717.

DOI：10.3788/OPE.20162413.0710

• Algorithm accuracies
• Contrast experiment
• Digital image correlation methods
• Hybrid genetic algorithms
• Image measurements
• Principle analysis
• Real-code genetic algorithms
• Subarea matching

In consideration of features presented of relatively weak local search ability, relatively low algorithm accuracy, etc with Genetic Algorithm(GA) applied to digital image correlation method, Conjugate Gradient(CG) method with strong local search capability was introduced;with the principle of complementarity, hybrid genetic algorithm was designed to improve the accuracy and stability of digital image correlation method. Firstly, according to different binding modes of CG and GA, three kinds of effective CG-GA hybrid algorithm were presented: embedded CG hybrid algorithm, elite CG hybrid genetic algorithm and parallel CG hybrid genetic algorithm. Then, principle analysis for these three algorithms was carried out. Next, based on simulated speckle pattern of preset displacement, contrast experiment between the above three algorithms with same parameter and standard RGA(Real-code Genetic Algorithm) for subarea matching search was designed. Lastly, Experimental results taking u direction as an example show that: mean error of embedded CG hybrid genetic algorithm is reduced by 69.52% compared with standard RGA algorithm with error fluctuation also reduced and relatively big calculation consumption;solution accuracy and stability of elite CG hybrid genetic algorithm and CG hybrid genetic algorithm are greatly improved;mean error between the two is reduced by 98.43% compared with RGA with high accuracy. In the digital image correlation method, the latter two hybrid algorithms are recommended for subarea matching search.

• 出版日期2016
• 单位上海交通大学