A Novel Invoice Automation System

OCR has been employed since the inception of the machine learning era, but it charged a lot of effort and money to develop. OCR models with improved accuracy have been produced numerous benefits to recent advancements in machine learning and deep learning that involved models and methodologies. Receiving invoices using a software tool to read and analyse them, record the processed data in an accounts payable system, match it with purchase orders and receipts created in connection with the invoice, process the invoice, and release payments to the appropriate party. Optical character recognition (OCR) is a method for extracting text from scanned documents and images. The receipt or document must be clear and language should be readable in order to extract such data effectively and more accurately. Here the proposed model is primarily focusing on pre-processing of the receipts so that the information from these can be easily comprehensible by the model and extraction accuracy will be escalated. The techniques associated with the proposed model are OpenCV, tesseract and google OCR to compare the accuracy after proposed pre-processing methods. The accuracy of OCR extraction after enhancement of the receipts is 20-30% higher than normal OCR output.