An Edge Detection Algorithm for Photoelasticity Analysis

Abstract

Photoelasticity has become a modern tool of stress analysis which is capable of competing with other tools employed currently, including finite element analysis. Improved model production and automated fringe analysis allow us to perform investigations of complex models, speeding up the rate of analysis and reducing the action by users, consequently automating the whole process. However, before automated fringe analysis, the mask of the model should be extracted. The authors discuss the development of a new algorithm to detect the mask of the model by analysing isochromatic fringe patterns used in photoelasticity. It is important to know the mask of the model for its analysis and to obtain a stress map. Unlike the available edge algorithms or any other techniques used to detect a model's mask, the proposed algorithm was developed to minimise user action, allowing the process to be automated. There is a major difference between the area of the background and area of the model from the point of view of image processing. Grey level of points inside the background region are distributed along the tilted plane with low total variance, and those points inside the model regions are distributed along the isochromatic fringes having the shape of a wave. The variance of certain areas is measured with respect to the approximated plane created over such area from the grey level of each point. Areas having low variance are then selected and extended to true boundaries based on the fact that edges are characterised by a huge jump in the grey level. The proposed method is validated experimentally for a plate with multiple cutouts in a dark field and a circular disc under diametric compressive load with frozen stress in white field.