2D Molecular Tagging Velocimetry

Problem Statement:

Given a 2D image of a laser sheet captured, accurately capture the coordinate of the intersections. From a pair of 2D images, resolve the correspondence problem between the intersections from each image and produce a 2D velocity field.

Current method utilized the template matching algorithm to detect the intersection. Although the method provides an accurate method of detecting the intersection, it lacks versatility in that if the image is rotated or scaled, the algorithm will fail to detect the intersections. Therefore, a roboust method is being investigated to improve detection task for varying magnification images resulting from stereoscopic imaging setups.

Experiment

Hough line transformation is preferred over template matching because of the robustness to noise of the algorithm with respect to the template matching. Even in signal to noise ratio (SNR) of 1, the Hough line is able to detect all of the lines; however, the limitation of the method is equally prevalent in that the algorithm in not well suited if there exist curvatures in the lines within the image. Since the magnification factor will vary across the spatial domain of the image, template will be created for each intersection and the templates will be sequentially mapped to the displaced image to find region of interest. The assumptions made for this combination of Hough line transform and template matching will only work if the displacement in 2D and 3D is much smaller than the distance between any two adjacent intersetion points.

Work in Progress

This project is still work in progress. Will be updated as more progress is made.