Packaging Recognition Fusion Algorithm Based on Visual Imaging Technology

In the process of online terminal inspection of packaging bags, the traditional packaging inspection method is mainly carried out manually. However, due to factors such as work fatigue and emotional fluctuations, the efficiency of online terminal packaging inspection cannot be guaranteed. Manual testing also affects the surface cleanliness of packaging bags and consumes a lot of labor. At present, the packaging industry generally adopts intelligent packaging technology to solve the problems existing in the production process of traditional packaging products, among which machine vision technology is its core. However, nowadays packaging companies mostly use machine vision to check for labels, color defects, and other issues. This is beneficial at the end of the production line to achieve the integrity and aesthetics of the appearance information of packaging bags, thereby ensuring the production quality of products. However, currently, there are still problems such as skewness and folding in the inhalable plastic bags transported at the end of the assembly line, which leads to uneven packaging and affects the appearance of the product. In severe cases, it can also result in missing quantities. This article explored the research on packaging recognition fusion algorithms based on visual imaging technology, and verified through experiments that the accuracy of the recognition model based on CNN (Convolutional Neural Networks) algorithm was higher than 81% under different experimental times. Compared to CNN algorithm, the recognition accuracy of BP (Back Propagation) algorithm was much lower. The highest accuracy of the packaging recognition model based on BP algorithm was only about 79%.