An image recognition-based premix packaging label automatic verification method and system

Abstract

The present application belongs to the technical field of image recognition, and discloses a premix packaging label automatic verification method and system. The method first collects label images, and after multi-scale semantic segmentation and optical character recognition, a label layout hypergraph is constructed according to a key field set; then the field topological structure and semantic features in the hypergraph are analyzed, and different fields are subjected to differential semantic normalization, combined with field variation rules, historical legal change samples, and field variation concept lattices are constructed by form concept analysis; subsequently, a verification candidate set is generated in combination with production batch standard information, a partial order relation of candidate explanations is established, a maximum candidate explanation is searched to complete global consistency determination, and a verification result containing a processing strategy and an abnormal classification report is output. The present application relies on hypergraphs and concept lattices to uniformly express field relationships, organize standard and legal variant equivalence relationships, distinguish between reasonable variations and substantive errors, and realize strict consistency of key fields, flexible fault tolerance of non-key fields, and automatic verification of labels.

Images