Research on tobacco debris recognition system based on YOLO model

This study aims to use machine vision technology to design and implement an intelligent recognition and removal system for impurities in tobacco production, in order to improve the purity of tobacco and the quality of cigarettes. By collecting image data on the conveyor belt and combining it with deep learning algorithms, rapid recognition of tobacco debris is achieved, and real-time removal is carried out through automatic devices. In this study, we provided a detailed introduction to the overall design concept of the system, hardware device design, recognition methods, dataset construction, model validation, and research on exclusion methods. Firstly, we chose the YOLO v8 framework as the core of the software algorithm, which has excellent performance in debris recognition due to its efficient image analysis and processing capabilities. By conducting five fold cross validation on a dataset containing various impurities, we verified that the designed deep learning model achieved an accuracy of up to 98.2% in detecting impurities and had good real-time performance, which could meet the real-time requirements of the production line. Secondly, in terms of the research and application effectiveness of the removal method, we have designed corresponding removal methods for the identification results of tobacco debris, and achieved comprehensive removal of tobacco debris by converting digital signals into level signals to drive the removal device. The actual application effect shows that the system reduces the quality control time of wind sorting debris from 30 minutes to 5 minutes, significantly improving production efficiency and product quality. In summary, this study designed and implemented a tobacco debris recognition and removal system based on machine vision, which has made a positive contribution to improving tobacco production efficiency and product quality.