Automatic archiving processing system and method for scanning archive image based on deep learning

• Main Authors: LI LIN, LUAN SHAN, CHEN WENZHENG, WEI XINTING, ZHAN NA, WANG XI
• Format: Patent
• Language: Chinese; English
• Published: 02.07.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides an automatic archiving processing system and method for scanning archive images based on deep learning. The method comprises the following steps: step 1, performing data preprocessing and model training; 2, performing subject recognition; step 3, performing tilt correction; 4, performing automatic threshold value processing and image reconstruction; step 5, processing table picture data; step 6, table segmentation: inputting the data marked in the step 5 into a Unet network for training to obtain a table segmentation model, and performing table segmentation on the input picture according to the table segmentation model to obtain a picture of cells; and 7, text line segmentation: carrying out text line segmentation on each cell segmented in the step 6 by using a CTPN model. According to the invention, the overall accuracy of file picture processing reaches 91.7% or above, the file filing working efficiency and the standardization level can be improved, and the personnel and equipment inv