Applied Research Of Partial Differential Equations For Skeletonization Of Optical Interferometry Images

Modern optical measurement techniques that combine laser technique and computer technique have the advantages of high sensitivity, non-contact, non-destructive, real-time and whole-field observation, and now, are becoming widely used on scientific researches and projects. Modern optical measurement techniques show physical attributes of objects in the form of optical interferometry fringe patterns, the key which obtains information of deformation and displacement from fringe patterns is to extract phase accurately. At present, phase shifting method, Fourier transform method and fringe skeleton method are three most important fringe analysis methods. Among these methods, the fringe skeleton method may be the most straightforward approach. This paper mainly researches the skeleton method for optical interferometry fringe patterns.When traditional skeleton methods extract skeletons, it is necessary to do some pretreatments for initial images, such as filtering, enhancing, binarization and so on, so as to get to better contrast. Against to the disadvantage, this paper describes a skeleton algorithm based on gradient vector flow field (GVF), and it can extract skeleton directly from initial images without pretreatment. In this algorithm, GVF sources from the external force of Snake model which is applied to edge detection. Under working together with internal force and external force, Snake model locks the targets to interested image features nearby and extracts needed data accurately. Designing external force in the form of GVF can expand the scope of mapping to make Snake model towards to expected image features. In this paper, the skeleton algorithm also applies the characteristics of GVF.In this paper, we derive GVF of initial images by calculating partial differential equations, then, determine locations of skeleton points based on the topological analysis of the GVF. We propose various partial differential equation models for calculating the GVFs of holographic interferometry fringe patterns, moiréinterferometry fringe patterns and electronic speckle pattern interferometry fringe patterns. Finally, by comparing with results of traditional methods---Hilditch thinning method and peak tracking method, we can see that the results of the method of this paper all achieve better effects. The skeleton method of this paper could extract the skeletons from initial images directly due to overcome impact of noise effectively, and guarantee accuracy of skeletons.