An Open Digitization Tool for Extracting Scientific Curve Data in Portable Documents
Shichao Zhou, Jun Lü · Advances in transdisciplinary engineering · 2022
Extracting original data in scientific figures embedded in documents precisely and efficiently remains a challenge, especially when it needs to be performed in high throughput way. To solve the automatic curve recognition problem, this paper proposes an integrated tool for extracting digital data from curve figures in portable documents, which mainly consists of picking/recognition/summary parts. During each extracting process, trained neural network firstly picks up figure pieces; the X-Y axes are then located by horizontal and vertical image projection and their labels are read using character recognition, which is followed by curve data recovery point by point; and finally the recognition result are summarized and sent back to the requester. This open tool is accessible and testable by anyone around the world via email, with open source on Github.