Ball position estimation in goal keeper robots using neural network

Many developments have been done in object tracking for humanoid robot soccer, but no specific research has been done in goal keeper robots in the wheeled robot soccer category. This paper makes a goal keeper robot that can map or transform ball position information from omni-camera images into cartesian images using a neural network. This paper can also divide the area of the flat field transformed into a grid area based on the size of the robot. The support system uses image enhancement and kalman filter in order to obtain a better image. Kalman filter has been used to maintain the existence of the ball. This system has been created with real ball data on the cartesian field. It has been called a data set, divided into training data and testing data. The percentage combination of data sets and the number of hidden nodes affects the accuracy of the transformation results. This new proposed model has been useful for omni to cartesian images transformation and cartesian to grid images conversion that can be used to determine the blockade area of ball.