化學式,作為一種直觀、易于理解的知識表示模型,在化學學科教育和學術交流研究中的作用至關重要。目前,將化學式輸入電子設備仍依賴于傳統(tǒng)的點擊-拖動單元組件的方式。然而,該方法既不方便,也缺乏效率。現(xiàn)在,隨著觸屏電子設備的大范圍普及,手寫輸入對人機交互產(chǎn)生了很大的影響,它已經(jīng)成了很多用戶的首選。因此,能夠直接在把化學式手寫輸入到電子設備中,引起了研究人員的廣泛興趣。手寫化學式識別是一項有挑戰(zhàn)的研究,由于它存在以下困難:1)是手寫體的隨意性,同一字符書寫多樣化,并且有同一字符筆跡中斷、不同字符筆跡相連、出現(xiàn)多余的點或線條等現(xiàn)象存在;2)化學式長度不一樣,比如單質(zhì)C只有1個元素,而堿式碳酸銅（Ca2（OH）2C03）卻有多達1 1個元素;3)化學式中存空間信息,如下標元素個數(shù)。成功識別手寫化學式,從理論上來說,可以促進手寫化學方程式的識別;另一方面,可以促進其在不同場景中的應用,如建立化學式知識庫,讓手寫化學式搜索更加方便快捷,又如充當教學輔助。根據(jù)輸入格式的不同,手寫化學式識別可以分為在線和離線兩個領域。在線識別中,輸入數(shù)據(jù)是一系列筆畫,而在離線領域中,輸入數(shù)據(jù)就是含有手寫化學式的圖像。本文將... 

【文章來源】：華中師范大學湖北省 211工程院校 教育部直屬院校

【文章頁數(shù)】：69 頁

【學位級別】：碩士

【文章目錄】：
Abstract
1 Introduction
    1.1 The problem of handwritten chemical formulae recognition
    1.2 Research motivations
    1.3 Objectives
    1.4 Contributions of the work
    1.5 Thesis organisation
2 Related works
    2.1 Literature review for handwritten chemical notations recognition
    2.2 Artificial neural network
        2.2.1 Convolutional neural network
        2.2.2 VGGNet
        2.2.3 Recurrent neural network
        2.2.4 Long short-term memory network
        2.2.5 Transfer learning
    2.3 Connectionist temporal classification technique
    2.4 Attention model
3 Handwritten chemical formulae recognition using CTC technique
    3.1 Feature sequence extractor
    3.2 Sequence dependency representation
    3.3 Transcription
    3.4 Objective function
4 Handwritten chemical formulae recognition using attention model
    4.1 Feature sequence extractor
    4.2 RNN encoder
    4.3 Decoder with attention mechanism
    4.4 Objective function
5 Experimentation
    5.1 Data set
        5.1.1 Chemical formulae selection
        5.1.2 Procedure of collecting the samples
        5.1.3 Data pre-process
    5.2 Experimentation on handwritten chemical formulae recognition using CTC-based method
        5.2.1 Experiment process
        5.2.2 Experiment results
        5.2.3 Analysis of prediction results
    5.3 Experimentation on handwritten chemical formulae recognition using attention-based method
        5.3.1 Training process
        5.3.2 Experiment results
        5.3.3 Visualization of attention on the dataset
    5.4 Comparative Evaluation
6 Conclusions and perspectives
    6.1 Conclusions
    6.2 Limitations
    6.3 Future work
References
Appendix A 摘要
Appendix B Acknowledgements



本文編號：3116399