Analysis of motion detectors evolved by Genetic Programming
Qiao Shi, Wei Yin, Andy Song · 2012
Genetic Programming (GP) is reputable for its power in finding creative solutions for complex problems. However the downside of it is also well known: the evolved solutions are often difficult to understand. This interpretability issue hinders GP to gain acceptance from many application areas. To address this issue in the context of motion detection, GP programs evolved for various detection tasks are analyzed in this study. Previous work has shown the capabilities of these evolved motion detectors such as ignoring uninteresting motions, differentiating fast motions from slow motions, identifying genuine motions from a moving background, and handling noises. This study aims to reveal the behavior of these GP individuals by introducing simplified motion detection tasks. The investigation on these GP motion detectors shows that their good performance is not random. There are contributing characteristics captured by these detectors, of which the behaviors are more or less explainable. This study validates GP as a good approach for motion detection.