Application of Segmented Transformer Feature Extraction in Near Infrared Spectral Data Classification

Table 1. Band range of different feature subsets
View table
Table 1. Band range of different feature subsets
Subset 1 2 3
Band range 1‒D/2 D/4‒3D/4 D/2‒D

Table 2. Distribution of sample origin
View table
Table 2. Distribution of sample origin
Province Guangdong Guizhou Hubei Yunnan
Number 45 214 188 1830

Table 3. Comparison of the impact of different preprocessing algorithms on classification accuracy
View table
Table 3. Comparison of the impact of different preprocessing algorithms on classification accuracy
Preprocess method R_accuracy /%
Original data 96.40
1st derivative 99.67
2nd derivative 99.38
SNV 97.28
MSC 99.09

Table 4. Comparison of performance of Embedding layers with different structures
View table
Table 4. Comparison of performance of Embedding layers with different structures
Module R_accuracy /%
FC Embedding 64.57
1D-Conv Embedding 99.09

Table 5. Performance comparison of different number of feature subsets
View table
Table 5. Performance comparison of different number of feature subsets
Number of parts 2 3 4 5
R_accuracy /% 91.26 99.09 98.85 97.50

Table 6. Performance comparison of different data fusion methods and regression head
View table
Table 6. Performance comparison of different data fusion methods and regression head
Module R_accuracy /%
Concatenate+FC 94.42
Stack+1D-Conv 99.09

Table 7. Comparison of different algorithms
View table
Table 7. Comparison of different algorithms
Algorithm R_accuracy /% R_precision /% R_recall /%
SVM-linear 27.56 7.00 25.00
SVM-rbf 92.62 93.50 92.25
Decision Tree 97.05 96.75 97.00
ResNet 91.74 92.46 91.74
Proposed algorithm 99.09 99.09 99.11

Tools

Get Citation

Copy Citation Text

Yongsheng Li, Xianwei Hao, Shu Xiang, Yidan Shi, Xiaorun Li. Application of Segmented Transformer Feature Extraction in Near Infrared Spectral Data Classification[J]. Laser & Optoelectronics Progress, 2024, 61(13): 1330002

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Spectroscopy

Received: Sep. 28, 2023

Accepted: Nov. 8, 2023

Published Online: Jul. 17, 2024

The Author Email: Yidan Shi (22134042@zju.edu.cn)

DOI:10.3788/LOP232211

CSTR:32186.14.LOP232211

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Band range of different feature subsets

Table 1. Band range of different feature subsets

Table 2. Distribution of sample origin

Table 2. Distribution of sample origin

Table 3. Comparison of the impact of different preprocessing algorithms on classification accuracy

Table 3. Comparison of the impact of different preprocessing algorithms on classification accuracy

Table 4. Comparison of performance of Embedding layers with different structures

Table 4. Comparison of performance of Embedding layers with different structures

Table 5. Performance comparison of different number of feature subsets

Table 5. Performance comparison of different number of feature subsets

Table 6. Performance comparison of different data fusion methods and regression head

Table 6. Performance comparison of different data fusion methods and regression head

Table 7. Comparison of different algorithms

Table 7. Comparison of different algorithms