This research studies the machine vision system and how it may be integrated to assist a robot system sith artificial intelligent (AI). This research focuses on building a vision-based feedback system for robotics application that consists of image processor and two vision-based sensor devices. A robot manipulator controller will drive a single arm industrial robot according to the input from vision system. The feedback systemalso feeds the Artificial Intelligent program necessary information to make the right decision, which is based on rules of a popular game, Tic-Tac-Toe. One of the advantages of this research is that it only uses a low resolution camera and image processing software generated by the algorithms itself without additional sensors such as sonar or IR sensor. This research developed an improved technique on object recognition and space occupancies determination which not affectes by the orientation of the subject. This project also implements colored object recognition technique by its color and size without edge detection process along with a self-calibration technique for detecting object location without any parameter of the camera by using only two referance points. Finally a set of experiments conducted to confirm the vadility of the proposed algorithms. The algorithms succesfully functioning with successful rate from 74% up to 100% and could handle the orientation of tilted object up to 45 degrees. The result from this research may be used in manufacturing plant for a robot system equipped with machine vision and artificial intelligent.