In recent years, industrial automation has made it possible to improve the quality of control systems of mobile robotics to reduce time and increase accuracy in applications such as object detection for product classification. In this work, a multi-object detection for classification and packing system has been developed using image processing, machine learning, and a robotic arm model KUKA KR10. Objects are detected using an adaptive threshold binarization and morphological operations to smooth and fill regions. Machine learning algorithms have been trained and evaluated using Neural Networks and Random Forest using color in RGB space as a feature to classify objects. Finally, objects are moved from the current to the assigned position through the robotic arm. The robot movements are executed by RoboDK and Python with a controller using a local area network and the trajectories are calculated using inverse kinematics. For tests, a controlled environment with 9 objects with cube shapes of different colors has been used to detect colors and classify those of blue and red only. The results show the vision system has an accuracy of 94.50% in color detection. The time that the robotic arm takes to classify all the objects is on average 116.65 sec. at 10% of its maximum speed. Future work focuses on object detection in industrial environments using a conveyor belt and other object features to classify such as shape or size.
A Multi-object Detection for Classification System Using Machine Learning and Robotic Manipulator
Lect. Notes in Networks, Syst.
Proceedings of the Latin American Congress on Automation and Robotics ; 2023 ; Universidad Don Bosco, San Salvador, El Salvador November 15, 2023 - November 17, 2023
2024-03-31
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Manipulator System for a Robotic Vehicle
Online Contents | 1998
Real-time surface detection and reconstruction using a robotic manipulator
Kraftfahrwesen | 1997
|Real-Time Surface Detection and Reconstruction using a Robotic Manipulator
British Library Conference Proceedings | 1997
|Object Classification for Robotic Platforms
TIBKAT | 2020
|Ocular Manipulator For Robotic Vision
SPIE | 1984
|