Desarrollo de un sistema de visión artificial para el reconocimiento, clasificación y maquinado de patrones con una tarjeta ARM

The present project contains the development of an Artificial Vision system to give solution to the control of conventional industrial processes, which have the main disadvantage of not self-adaptable in cases of different patterns to be machined. The study of the Artificial Vision was carried out...

Descripción completa

Autor Principal: Andrade Fierro, Luis Gonzalo
Otros Autores: Chulca Simbaña, Luis Alfonso
Formato: bachelorThesis
Idioma: spa
Publicado: 2018
Materias:
Acceso en línea: http://dspace.ups.edu.ec/handle/123456789/15284
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
Sumario: The present project contains the development of an Artificial Vision system to give solution to the control of conventional industrial processes, which have the main disadvantage of not self-adaptable in cases of different patterns to be machined. The study of the Artificial Vision was carried out that helped to acquire the basic knowledge to develop a self-adapting process. The use of Artificial Vision algorithms allowed us to obtain information from the environment through a webcam to later process the information obtained in a model programmed in Matlab Simulink. With the help of Simulink's artificial vision toolkit, the main characteristics of the models studied were obtained, such as the geometrical center, as well as the recognition of patterns using a neural network together with the calculation of Hu's invariant moments. Through the Raspberry support package for Simulink, the block programming was compiled, thus achieving autonomy for the ARM card, which processes the information at 32 bits. Finally, the coordinates data of the geometrical center and the shape of the model are sent by RS-232 serial communication to the robotic arm Mitsubishi RV-2AJ, to perform the machining of the models by making a perforation in the geometrical center of the same. The system was able to recognize the figure of the models as well as their morphological characteristics; and the robotic arm managed to make a hole in the geometric center with an error of approximately 3%.