Object recognition and robot arm control by intelligent methods

Automation is a working matter, which has a very long historical background, for human and always arouses interest. Also one of the working areas related with the automation is robots, and their applications. If it is thought that parts of a robot act for executing several tasks, collecting the environmental data and taking them in to account become important. Visual sensors like cameras are often used for collecting these data. Stereovision process, which provides human three dimensional sensing, can be formed for computer vision by using more than one camera. Therefore the depth information of scene in the field of view of stereovision system can be determined. The acts as object grasping, and taking them to the certain localization or doing certain works with attached tools can be possible for robots by using the depth information in robot automation.In this thesis, it is aimed to recognize certain objects without regarding its location, orientation, and scale in a video frame and to determine object distance to a certain reference point. For this aim, an experimental system is implemented in the thesis concept. This system includes an active camera system, Zernike moment method, which is used in feature extraction, an Artificial Neural Network (ANN), used for recognition by using these features, a stereovision algorithm for determining the objects distance from a reference point, a robot arm and its controller unit. Also three ANN?s are used for determining the angles of robot?s first three joints. Adaptive control of electric motors, which actuates the joints of robot, is important from the viewpoint of robot?s reaching the target quick and accurately. So a Fuzzy Logic (FL) controller simulation of the robot?s base motor is done for comparing it with the performance of PID?s, used in the experimental system.In this system a DSP-FPGA based embedded card for implementing the necessary algorithms very fast. Normally, this card is designed as a video processor card, but it makes the card possess the features of stereovision process and capability of transmitting necessary information for robot arm after some efforts, and through circuits, prepared.It is hoped that applications of this thesis can be used in different applications. Because this system executes several activities such as stereovision process, position estimation, and solving the inverse kinematic problem of the robot arm by neural network together.