基于深度学习的发票识别模型在经费报销中应用研究

Application of invoice recognition model based on deep learning in expense reimbursement

随着社会经济的迅速发展,发票的使用日益增多,作为财政支出中经费报销的一种有效凭证,在经济活动中占据了很大的比例。当前的发票报销还采用传统的人工审核,由此导致效率比较低,资源比较浪费。基于此,研究提出了一种基于深度学习的发票识别模型,并验证了其在经费报销中的有效性。实验结果表明,在发票图像检测流程中,其准确率最高达到95.18%,而在发票识别中最高达到97%;在运行速度上,文字识别最低达到0.06ms。综合来看,研究提出的识别模型表现出了较高的识别性能,并可以做到实时检测,从而提升了审核效率,在实际的经费报销中具备实用性。

With the rapid development of social economy,the use of invoices is increasing.As an effective voucher for financial expenditure reimbursement,invoices account for a large proportion in economic activities.The current invoice reimbursement also uses traditional manual review,which leads to low efficiency and waste of resources.Based on this,this paper proposes an invoice recognition model based on deep learning,and verifies its effectiveness in expense reimbursement.The experimental results show that the highest accuracy rate is 95.18%in the invoice image detection process,and 97%in the invoice recognition process;In the running speed,the minimum character recognition is 0.06ms.In a word,the recognition model proposed in the study shows high recognition performance and can be detected in real time,which improves the audit efficiency and is practical in actual expense reimbursement.