Research on classification method of food packaging character defects

Delong Lu,Tao Jia,Qingguo Lin,Guoqing Wang,Zonghua Zhou,Suyu Jin

DOI: https://doi.org/10.1117/12.2605916

2021-09-15

Abstract:The production date on the food package is an important safety information. It is widely used in the production line to print the production date with inkjet code printer. However, character defects such as fracture and displacement may occur due to machine failure and offset of packaging materials. In order to improve the printing quality of packaging characters of production line, this paper studies the related classification algorithm for the common printing defects of production date. In order to classify defects, a 17-dimensional feature vector is designed to describe defects based on the geometric features of defects. Because of redundancy and interference information, linear discriminant analysis (LDA) is used to reduce the dimension of the eigenvectors. The classification mode and parameters of SVM are experimented on the classification section. According to the experimental data, the classification accuracy is more than 94%.