Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden.
powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden.
powered by
Abstract
Goals. The present study was designed to build a prototyping and develop algorithms that allow the detection, classification, and movement of objects of a robotic arm of 4 DOF with the following technologies: ArmUno arm structure, Raspberry Pi 3 B+, PiCam 2.1, driver PCA9685 for servomotors, Opencv3, and python. Another goal was to measure the effectiveness of prediction and classification of objects photographed by the robotic arm, using machine learning with the KNN classifier method.
Methodology. The generation of a dataset of 800 photographic images was proposed, in 4 categories: volumetric geometric shapes conformed by 200 images each one of them. With this, processing techniques were applied to the image captured by the camera to detect the object in the image: Grayscale filtering, Gaussian filtering, and threshold.
Then, the characteristics of the object were obtained through the first two invariant moments of HU, and finally, the machine learning method KNN was applied to predict, that the image captured by the robotic arm belongs or not to a certain category. In this way, the robotic arm decides to move the object or not.
Results. According to the plot of the obtained data described in the results section; the level of correct answers increases markedly by using the techniques described above. The prediction and classification using KNN were remarkable, For all the tests carried out The average effectiveness of KNN method was 95.42%. Once the scripts were integrated, the operation of the robotic arm was satisfactory.
Anzeige
Bitte loggen Sie sich ein, um Zugang zu Ihrer Lizenz zu erhalten.