Industrial Vision Robot with Raspberry Pi using Pixy Camera : Stereo Vision System

Abstract

Industrial robots are not human, they are machines. They are programmable ma-nipulator devices which can move tools or parts via a set sequence of motions. In addition, they can be reprogrammable, that is, the robot’s action can be modified by changing the control settings without replacing the hardware. They add some characteristics of traditional machines likewise as characteristics of machine oper-ators. For an operator, it is easy to be taught to do a new task. But, for a machine, a task can be repeated for prolonged times with great precision.

This project focused at developing a Robot Vision system using a combination of low-cost camera hardware and computer algorithms to enable robots to process visual data from the world. The stereo vision algorithm which consists of two cameras, and the developed application are able to calculate a 3D position from s 2D detected object. In addition, the detection algorithm based on color differ-ences was used by the cameras which enable 2D object tracking and outputted data coordinates of the object being detected. Then, the 3D object position is produced through the calculated 2D object data coordinates, which made ready for robot teaching.

Furthermore, the developed application was based on OpenCV API in C++, which was an interest in the development of this project. The use of this was to treat the image capture by the cameras. TIY software with modification was used to do the object tracking. BOOST is a set of C++ libraries that provide image pro-cessing, and linear algebra functionalities. This library was the appropriate choice because of the reliance of TIY on it, and some other aspects of it that are im-portant to the rest of the project.

Finally, the test results showed that the project was successfully developed. In addition, with the developed project, my expertise in embedded system pro-gramming has been consolidated and I have obtained further knowledge in the field of robotics and computer vision.