Sistema picking para el mejoramiento de la distribución logística a través de visión artificial para piezas plásticas

Abstract

This study aimed to implement a picking system for improving logistics distribution through artificial vision, for which plastic parts were classified, stored, and shipped according to their shape and color. It was based on a bibliographic study for the determination of parameters in a picking system. Using Computer-Aided Design (CAD) software, the design and structural analysis of the prototype were performed for validation. The hardware used was an Orange Pi 3 board, electronic components, and electric actuators to generate the movement, with stepper motors, drivers, and an Arduino MEGA. Subsequently, in Integrated Development Environment for Python (IDLE), the algorithm for artificial vision, storage, dispatch, and database of plastic parts was coded with Arduino, supported by a graphical interface combining resources for process control and monitoring. In the classification of parts, image conversion, thresholding, and filtering methods were used which, once implemented, resulted in an algorithm efficiency of 80.55% in color detection and 100% in shape in a study at 30 FPS. The time of storage and dispatch of the pieces on the shelf is 17.40 and 13.05 minutes respectively, where a difference was observed due to the fact that the dispatch does not consider the sorting stage. In addition, a total energy consumption of the system of 216 W was obtained, which does not exceed the capacity of the installed source, allowing the system to be scaled for future improvements. Finally, it is concluded that the implemented system simulates the operating conditions of a picking system and thus enables a test environment. It is recommended to include this process in future work to set up complete production lines, thus increasing the research scope of this thesis.