In this paper the design, implementation and application of a computer controlled vision system with a robot manipulator for classification of various fruits is introduced. The system, which consists of a RGB Camera, a PC and a five degree of freedom (DOF) robot manipulator, specifically set up to work in real-time applications. At the first stage of study the captured image of objects in RGB scale viewed by the camera is stored via USB link of computer as digital image data in registries for further segmentation process to recognize the objects. A particular Neural Network is designed to classify objects exactly. The moving objects on a belt, driven by a servo-stepper motor, are recognized in size and property via a vertically mounted camera. The recognized objects are then assorted by a computer controlled robot manipulator with five DOF and a gripper at the end-effecter, in order to transport the objects via a predefined trajectory to designated locations. As second stage, the kinematic link table of robot manipulator using inverse kinematics is determined, so that from the given way points of a particular trajectory the link parameters (joint angles) can be calculated. The joint angles are used as reference values for a joint based control algorithm programmed in C++. The joints are equipped with geared DC motors with position feedback. A particular control program is also used to control the motion of servo-stepper belt motor and the pneumatically activated gripper valve of the manipulator. Successful results are evaluated both in assorting of fruit in size, property, in synchronous working of the robot arm and in connection with image processing.
Read full abstract