Spectroscopy and Spectral Analysis, Volume. 42, Issue 11, 3361(2022)
Identification Method of Pollen Typhae Processed Products Based on Convolutional Neural Network and Voting Mechanism
Figures & Tables(10)
Fig. 1. Identification model of processed products of Pollen Typhae based on CNN and voting mechanism
Fig. 3. Partial convolution kernel of one-dimensional convolution pool
Fig. 4. Partial convolution kernel of two-dimensional convolution pool
Fig. 5. CNN eigenvectors of four pre-processing methods
(a): Unchanged preprocesses CNN eigenvectors; (b): SNV preprocesses CNN eigenvector;(c): First-order difference preprocesses CNN eigenvectors; (d): Min_max preprocesses CNN eigenvectors
Fig. 7. Cross entropy loss of CNN based on four pre-processing methods
Fig. 8. Comparison of the proposed method with CNN, LDA and SNV-LDA in terms of test accuracy
Table 1. Weight distribution of different preprocessing methods
View tableView in ArticleTable 1. Weight distribution of different preprocessing methods
预处理方法 数据集 CNN预测准确率a 权重w 保持不变 Str1 a1=72% w1=0.3 SNV Str2 a2=88% w2=0.5 一阶差分 Str3 a3=92% w3=0.5 Min_max Str4 a4=76% w4=0.4
Tools
Get Citation
Copy Citation Text
Cheng-wu CHEN, Tian-shu WANG, Kong-fa HU, Bei-hua BAO, Hui YAN, Xi-chen YANG. Identification Method of Pollen Typhae Processed Products Based on Convolutional Neural Network and Voting Mechanism[J]. Spectroscopy and Spectral Analysis, 2022, 42(11): 3361
Paper Information
Category: Research Articles
Received: Oct. 12, 2020
Accepted: --
Published Online: Nov. 23, 2022
The Author Email: CHEN Cheng-wu (chenchengwu@njucm.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20