ABSTRACT
In this dissertation we developed a fiber optic connector optical return loss (ORL) model from the connector endface image profile of a 200X inspection microscope. The connector endface profiles were histogram equalized and then evaluated with the Canny, Sobel, Prewitt, Laplacian of the Gaussian (LoG), Roberts and Frei-Chen edge detection operators using Matlab 2013 script, these operators yielded edge pixel count averages of 139.1, 49.43, 49.57, 103.2, 45.19 and 50.98 detections per section respectively. The Canny operator emerged the best candidate technique after yielding the maximum edge pixel count of 139.1 detections per section, where the figure of merit is the magnitude of the total number of edges detected. The image segmentation and image similarity measure of the Image under Test (IUT) and the Measurement Quality Jumper (MQJ) were implemented using the cross correlation coefficient and pixel based segmentation method respectively so as to establish the coefficients of the connector profile region reflectance in the ORL model. These tools were implemented using a combination of the Matlab 2013 Graphic User interface (GUI), image processing toolbox and symbolic math toolbox. The model results showed a mean connector ORL accuracy of �0.5dB and �0.55dB for the multimode and Single mode connector ORL respectively. These fall within the TIA/EIA-455-95-A maximum specification of �1dB and �5dB for the Single mode and Multimode optical fibers