Image Recognition Method of Import and Export Ledger of Catering Industry Based on Table Recognition
Aiming at the problem that the current table recognition algorithm has weak text and table recognition ability for the food and beverage import and shipment account pictures in natural scenes,this paper refers to the previous table recognition algorithm architecture.This paper presents an algorithm for table recognition of ledger picture based on im-proved DBNet text detection algorithm,SPIN scene text recognition algorithm and TableMaster table recognition algorithm.The TableMaster recognition algorithm is used instead of the PLANet table recognition algorithm,which enhances the recog-nition capability and accuracy under lightweight conditions.In this paper,text detection and form recognition model are combined with series inference,and the recognition rate is 94.7%in test scenarios,which has high practical value.
table identificationDBNetSPINTableMasterledgercatering industry
万方数据
维普
