Evaluación de Algoritmos de Tracking 3D para la Simulación de un Brazo Robótico, Mediante kinect.

Abstract

The research consisted of evaluating Tracking algorithms 3D, applied to the simulation of a robotic arm, using Kinect; in the Electronic Engineering School of Escuela Superior Politécnica de Chimborazo. The evaluation was realized to three algorithms, CAMShift, Algorithm by Color and thresholding algorithm, with them the application of the object monitoring (tracking) was implemented. Writing the program in C++ and using the resources of the Software OpenCV and Libfreenect. The results were evaluated applying statistical methods as measurement and error. When the results were tabulated the Algorithm by thresholding was chosen as the most adapted for the application. A mathematical study on the cinematic of a robotic arm was the following step. This study was realized in the Software Matlab, the same that gave the parameters researched for the robot use. Then, the robotic arm simulation was implemented printing the results on the image captured in real time. The physical system par is the Kinect, a sensor composed of camera RGB, infrared sensor. Its main function is to capture in real time a scene and obtaining the depth dimensions of all the components that belonging it. All programming is developed in the Linux Ubuntu 14.04 operating system. The programming language C++ was used, working as a whole with the bookstores OpenCV and Libfreenect, which serve for the use of Kinect and artificial vision. It was corroborated that the system has the 98,83%, of precision in the information. It is concluded that the thresholding algorithm is the most adapted for a Tracking application using Kinect. It is recommended to the researches from Electronics Engineering School, to take the results of this work as a starting point for physical implementations of robotics arms that will perform subsequently.