A seal script recognition method and system based on incep-capsnet network

Abstract

The present invention relates to a method and system for recognition of seal script based on Incep-CapsNet network. The method includes: obtaining a seal script image data set, and making classification labels according to font and font; performing preprocessing and image augmentation operations on the image data set; ‑CapsNet network extracts features and recognizes and classifies seal script images after preprocessing and image augmentation; reconstructs seal script images after preprocessing and image augmentation through the reconstruction network as an effective regularization method. The present invention can extract more and more essential features according to the Inception structure, and overcome the shortcomings of the traditional deep convolutional neural network model, such as the direction of the unidentifiable features and the relative position in space, etc., and has accurate recognition accuracy , and has strong robustness.

Description

technical field [0001] The present invention relates to the field of image recognition, in particular to the research field of image recognition of ancient characters, specifically a seal script recognition method and system based on the Incep-CapsNet network. Background technique [0002] Seal script is the core component of ancient Chinese characters, and it is also the earliest one among the five major calligraphy fonts in my country. It mainly includes big seal script and small seal script. Among them, big seal script mainly refers to ancient scripts such as oracle bone inscriptions and bronze scripts before the Qin Dynasty, and small seal script refers to the unified characters of Qin Shihuang. The font to use afterwards. In terms of the quantity, shape and structure of seal characters, they have all been well developed. As an important information carrier of ancient Chinese culture, seal script has played an important role in the cultural inheritance and development of...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the deficiencies of the prior art, and propose a seal character recognition method and system based on the Incep-CapsNet network, which can extract more and more essential features according to the Inception structure, and overcome the traditional deep convolutional neural network based on the capsule network. The shortcomings of the network model, such as the orientation of features that cannot be recognized and the relative position in space, have both accurate recognition accuracy and strong robustness.

Images