Ball Identification and Localization using Regression for Wheeled Robot Soccer

Abstract

Robot soccer is a smart robot mostly use in a soccer game. One of the crucial abilities that robot soccer must have is to correctly identify and locate the ball. This paper presents a real-time ball identification using color Altering in the HSV color system. Once the ball is detected, the localization is performed by regression method to find the function to calculate the real position of the ball by entering the ball position in the camera frame. The Cartesian coordinate then convert to polar coordinate, and then error analysis is performed using MAE and MSE. The result shows that for y coordinate, the quadratic function works better than the exponential function, and for the x, cubical regression is performed. The result is 14,43 MAE and 248,07 MSE.