A Multi-table Image Recognition System Based on Deep Learning and Edge Detection

EasyChair Preprint 947

Abstract

In order to recognize the information in multitable images, we use deep learning and edge detection and propose a multitable image recognition system based on deep learning and edge detection. After the image is preprocessed to the standard image, the table area in the image sample is located by Hough transform, morphological function and edge detection technology, and the cells in the table area are extracted from the image sample. The projection analysis method separates and extracts a single character in the cell, and finally uses a recognition model based on the deep learning convolution neural network to recognize each single character image. After experiments, we find that the system has the advantages of fast and accuracy, which brings great convenience and benefit to the companies engaged in information field

Keyphrases: Hough transform, Morphological Function, Table Image, deep learning, edge detection, projection analysis

@booklet{EasyChair:947,
  author    = {Dong Xiao and Haiyang Sun and Zhonglin Bao and Peng Yang and Wenrui Zhang},
  title     = {A Multi-table Image Recognition System Based on Deep Learning and Edge Detection},
  howpublished = {EasyChair Preprint 947},
  year      = {EasyChair, 2019}}