Sistema de visión artificial para el reconocimiento de productos básicos en la percha de un supermercado

Abstract

The main objective of this project is to create an algorithm assisted by artificial vision to recognize basic products on a supermarket shelf, aiming to optimize time, improve efficiency, and reduce errors when placing products on the shelves. For the visualization of the algorithm, an artificial vision interface was developed to classify consumer products such as fruits, bottled beverages, vegetables, canned goods, among others. YOLO was used as the main tool for training the neural network. YOLO excels in artificial vision applications as it helps optimize key parameters for object recognition, object detection, and object counting. The main components used in this project include a Raspberry Pi 4 with an internet connection and an HD webcam. To execute the code, the RPI terminal is initialized using specific commands to open an interface designed for the project. Once the artificial vision algorithm is started, the system begins detecting basic consumer products. Object detection is performed in real-time through the webcam, operating at a resolution of 640 x 480 pixels. When capturing products, the resolution enhances to 1280 x 720 pixels, allowing for image improvement and comparison with the inventory database. Finally, the interface displays a table showing the results of the product count on the shelf. Keywords: Computer vision, Raspberry Pi, Python, OpenCV, object detection, object counting, image processing, automation, technolo-gical applications.