Searching Ball Around ROI to Increase Computational Processing of Detection

摘要

Wheeled Indonesian Soccer Robot Contest (Wheeled KRSBI) is a national competition focused on the development of wheeled soccer robots. EEPIS Robot Soccer On Wheeled (ERSOW) is one of KRSBI's wheeled soccer robots. ERSOW has Artificial Intelligent (AI) for detecting balls, dribbling, and avoiding opponents. Currently, the object detection system in ERSOW uses image processing, this will be a heavy computation because it is done thoroughly in all pixel area. This research focused on making local detection to reduce heavy image processing. Local detection works by searching for local objects in the area around the detected object. It creates the ROI for local detection. From the results of tests conducted when the ball is detected, local detection system can increase 4 EPS from the full frame detection system, whereas in the test results when the ball is not detected, the local detection system is 0.2 FPS slower than the full frame detection system.