Optimisation et déploiement de modèles d’intelligence artificielle embarqués sur Raspberry Pi pour des applications en temps réel

Abstract

This project aims to develop an intelligent system based on a robotic arm capable of recognizing various geometric shapes (such as circles, squares, and triangles) through an integrated camera, and then automatically pointing to the desired shape based on its color and geometric features. This system serves as a simplified model of computer vision applications in automated control, combining image processing with motion-based decision-making in an interactive environment. The system is implemented using a Raspberry Pi 4 as the central processing unit, with all functionalities programmed in Python. The OpenCV library is used to process the images captured by the camera, extracting geometric shapes and analyzing color characteristics using the HSV color space. This is followed by edge detection and shape recognition using contour analysis algorithms. Based on the target shape, commands are sent to the robotic arm to move its joints via stepper motors toward the shape’s position, enabling precise and realistic pointing motion.