Fiber Optic Core Image Detection: Comparison of classifiers

Visual inspection of fiber ends is often required during installation or maintenance of fiber optic cabling. Automated analysis first requires accurately determining the location of the fiber core. In this paper, we compare the accuracy and reliability of several different classifiers in finding the fiber core. Classifiers such as naive bayes, perception, and three layer feed forward neural networks have proven to be a reliable way of recognizing items in images. These three classifiers as well as a none learning contrast-based algorithm are applied to the problem of finding the fiber core and results are compared. Our results show that using a three layer neural network is the most accurate of the detection algorithms.