Flow Direction Prediction of The Ball Movement for Humanoid Robot Soccer Goalkeeper Ahmad Fuady, Indra Adji Sulistijono, and One Setiaji Department of Mechatronics Engineering, Electronics Engineering Polytechnic Institute of Surabaya (EEPIS), EEPIS Campus Sukolilo, Surabaya 60111, Indonesia. (Email: foe@student.eepis-its.edu ) Abstract In Robocup, The most important sensor on humanoid robot soccer is a camera. The camera serves as the eyes of the robot as well as in humans. This camera is used by the robot to detect the ball. The task of Humanoid Robot Soccer as a Goalkeeper is to keep the goal and must be able to block a ball that came directly to the goal from many directions so the ball is not got into the goal. As already been known, the results of the camera image captured is always later than the pictures that has been taken, moreover if the shooting process is added with image processing such as adjusting contrast, brightness, etc. then the delay in image capture will be even greater. This causes late in reaction to the Humanoid Robot Soccer Goalkeeper when it’s blocking the ball that coming from any direction. Therefore we need a computer vision technique to estimate the direction of movement of the ball so there is no delay in reaction to the robot in blocking the ball. In this paper is discussed the flow direction prediction of the ball movement for Humanoid Robot Soccer Goalkeeper. The processing of ball movement prediction is obtained by comparing the previous ball data detection and the latest data detection to get the direction of the ball movement. This robot is a second generation of humanoid robot called EFuRIO soccer (Fußball EEPIS Robot IO). Keywords : Robocup, humanoid robot soccer goalkeeper, computer vision, flow direction prediction, EEPIS Fußball Robot IO (EFuRIO) I. INTRODUCTION Humanoid Robot Soccer is a human-shaped robot developed specifically for the game of football. As Goalkeeper, robot must have camera to capture image around robot that use for tracking ball. The method is widely used to detect the object is to use a color code on the image to determine the detected [1]. Image according to Webster's dictionary, is a representation, resemblance, or imitation of an object or objects. The image is the picture in two dimensions. Seen from a mathematical standpoint, the image is a function of light intensity continue on two-dimensional plane. An image obtained from the catching power of light / light reflected by objects through optical instruments such as cameras, eye, scanners and so on. While the image processing is a process where the input and output of images [2]. In early development, image processing is done only to improve the quality of the image, but with the development of science and technology world of computing that is characterized by increasing memory capacity and processing speed computers, as well as the development of computer science that allows people to get information from an image then the image processing can’t be released to the field of computer vision. Computer Vision is often defined as one branch of science that studies how computers can recognize objects that were observed / observed through the sensors (cameras, etc.) [3]. Field of science is to develop a variety of approaches by combining the techniques of Image Processing and Pattern Recognition / Object. Computer Vision System (CVS) is expected to have a high level of capability as the Human Visual System (HVS). Research on computer vision has produced a method that can be used to scale, orientation and affine invariant gradient image features to determine the characteristics of the object in the image [4]. Computer vision is closely connected with other fields. To support the task of computer vision then there is some support functions to the system [3]:  Process the image capture / image (Image acquisition)  The processing of image / image (Image processing)  Analysis of the image data / image (Image analysis)  The process of understanding the image data / image (Image understanding) In this paper, computer vision use to predict flow direction of ball movement by comparing the previous ball data detection and the latest data detection to get the direction of the ball movement. This paper is organized as follows. Section II explains the design and dynamic model of humanoid