Research On Image Retrieval And Intelligent Evaluation Algorithm Of Hard-pen Calligraph

Chinese calligraphy is a treasure of world culture.Inheriting and promoting Chinese calligraphy not only helps to enhance cultural confidence but also helps to promote the outstanding traditional culture of the nation to the world.However,the current scarcity and uneven distribution of Chinese calligraphy education resources have increasingly become prominent issues,which have led to difficulties in the development of Chinese calligraphy education.In recent years,with the increasingly mature technology in the field of images,intelligent teaching has become a new way for the development of calligraphy education.However,the research on image retrieval and intelligent evaluation algorithms for Chinese calligraphy has yet to be well developed.On the one hand,existing calligraphy image retrieval methods often perform poorly when dealing with massive data.On the other hand,the current intelligent evaluation algorithm for Chinese calligraphy only involves simple selection of a few artificial features for similarity comparison,which cannot truly help students learn calligraphy.To address these problems,this article focuses on more representative hard-pen calligraphy in daily life,and studies the image retrieval and intelligent evaluation algorithms for hard-pen calligraphy.Based on this,a relatively complete system is constructed.This system can achieve computer-assisted teaching in the calligraphy field to some extent,helping people improve their Chinese writing skills in daily work and study.The main work of this paper includes:(1)To meet the research requirements of this paper,this paper collected and sorted out the pictures of primary and middle school students’ handwritten hard-pen calligraphy works in collaboration with a calligraphy teaching company,and proposed a method of slicing and dividing the pictures of handwritten hard-pen calligraphy works,so as to produce a data set of handwritten hard-pen calligraphy images.The data set contains 3794 Chinese characters,each of which has more than 400 samples for use.(2)In response to the existing problems of complex feature extraction and poor retrieval results when solving calligraphy image retrieval problems,a calligraphy character image retrieval algorithm based on Siamese network and manual features is proposed.Firstly,a calligraphy character image retrieval model is constructed based on the Siamese neural network framework and the data set made in(1),which can screen out the candidate sets of calligraphy character images in massive data.Then,a calligraphy character image feature sorting algorithm is proposed based on manually defined structural features,which is used to further optimize retrieval results by sorting calligraphy character images in candidate sets obtained by the retrieval model.(3)In response to the problem of low accuracy of existing intelligent evaluation algorithms for calligraphy characters,a hard-pen calligraphy character intelligent evaluation algorithm is proposed based on structural and stroke features,referring to the primary level exam standards for calligraphy and the opinions of professional calligraphy teachers,which can obtain relatively accurate evaluation results.(4)An intelligent evaluation system for hard-pen calligraphy characters has been developed.After users upload their hard-pen calligraphy works,the system first extracts each calligraphy character in the work using the algorithm proposed in(1),then finds the corresponding original character using the algorithm proposed in(2),and finally displays the intelligent evaluation result of each calligraphy character on the user interface based on the algorithm presented in(3).