Radiographic image edge detection techniques : experiments on a rocket nozzle simulator

The purpose of this thesis is to study the application of edge detection techniques to radiographic images. Image edge detection is the first step in enhancing analysis capabilities from visual analysis to radiographic image measurement with known accuracy. Edge detection will make tasks like measurement of rocket nozzle throat erosion or propellant burn rate possible.

Five detection techniques were considered as candidate detectors for radiographic image analysis. These techniques were all tested using the same radiographic images and their results were compared.

This study was focused on measurement of rocket nozzle throat erosion. Since data from an actual rocket firing were not available, a nozzle throat simulator was designed and constructed. X-ray data of the simulator provided exactly known "throat" or hole diameters which were used in evaluation of the different detection techniques.

For the conditions examined here, the K-means combined with Sobel gradient preprocessing edge detection technique was found to be superior to the others evaluated. Results from this procedure produced error levels less than 0.125 inches for hole measurements ranging from approximately 1 to 3 inches