Testing the performance of a single pole detection algorithm using the confusion matrix model

Abstract

This study aims to examine the single pole detection algorithm using a confusion matrix model which is a specific table that makes it easy to visualize the performance of an algorithm. The algorithm tested is the YuRHoS pole detection algorithm, a new algorithm developed by researchers to detect pole objects with not poles. Methods used is by calculating three aspects of algorithm performance in machine learning, namely sensitivity, specificity, and accuracy. The value of the three aspects of performance depends on four variables, namely true positive, true negative, false positive and false negative. The calculation process is done by matching the pixel detection region with the ground-truth region. The test results for 4 (four) different single pole images found that the YuRHoS pole detection algorithm is better than other algorithms on two measurement aspects, namely specificity, and accuracy. Excellence aspects of specificity obtained because of its ability in detecting the object instead of a pole. Excellence aspects of accuracy indicated because more accurate in detecting a pole. As for sensitivity aspects, both the detection algorithms are having the same reliability in correctly predicting a pole.