Digital image correlation (DIC) is a non-contact, full field displacement measurement technique. It is primarily suited to making high precision and high accuracy measurements, and is therefore commonly used in laboratories for experimental work. However, in recent years, improvements to technology and commercial interest in the industrial internet have created the potential for methods such as DIC to be utilized widely in industry.
This thesis makes a preliminary investigation of errors that are likely to occur when DIC is performed in an uncontrolled environment. The characteristics of displacement and strain field measurements affected by camera motion, changing focus and inconsistent illumination are compared. To achieve this, computer graphics software is used to simulate a scene in which a stationary plate is viewed by a stereo imaging system. Animations of systematic changes to camera position, focus and lighting are made, and a ray-tracing render engine is used to produce the resultant photo-realistic images. In later simulations, the plate is substituted for a cylinder in different orientations, to investigate how error characteristics vary with surface angle and distance.
The DIC algorithms are found to be robust, allowing viable measurements even when significant changes are made to the imaging environment. Changes to illumination and focus are seen to produce random noise, most likely resulting from incorrect matching of subsets. Conversely, camera motion is seen to result in systematic error, with each transformation component displaying distinct characteristics. This observation is significant, since it indicates a possibility for identifying and correcting for such errors in industrial applications. Surface curvature is found to have minimal impact on the error characteristics for most camera transformations, but significant differences are observed for camera translations along the y-axis and rotations around the x-axis.