Ball Detection Based on Color and Shape Features Captured by Omni-Directional Camera

Abstract

The ability to detect the ball on a soccer robot is the key to victory in a match. However, in reality, there are several things that can become a distraction in the detection process. These include the color and shape of the ball which often change. This study aims to create a system that can detect balls based on the color and shape of the ball by comparing the level of accuracy and computation time process. The research process was carried out by capturing real-time video on an Omni-Directional camera and through image processing, including image segmentation by taking HSV colors at a certain range, noise image restoration, feature extraction using contours with optimal roundness, and analysis of detection results. From the application of this method to video images of random balls, an accuracy value of 90,61 percent is obtained. This is obtained with an average computing time of 0,025 seconds.