Multifeature Automatic Spectral Classification of Plastic Steel Window Based on Machine Learning Model at Molecular Level

Zhen Zhang; Jifen Wang; Pengwu Lu; Zhaokui Fu

doi:10.3788/LOP202259.1630005

Laser & Optoelectronics Progress, Volume. 59, Issue 16, 1630005(2022)

Multifeature Automatic Spectral Classification of Plastic Steel Window Based on Machine Learning Model at Molecular Level

Zhen Zhang¹, Jifen Wang^1、*, Pengwu Lu², and Zhaokui Fu¹

Author Affiliations

¹School of Investigation, People’s Public Security University of China, Beijing 100038, China

²School of Public Security Administration, People’s Public Security University of China, Beijing 100038, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(11)

Fig. 1. Schematic diagram of support vector machine^[11]

Download full size

Fig. 2. Infrared spectrum diagram of 126 plastic steel window samples

Download full size

Fig. 3. Overall classification accuracy of Fisher discrimination analysis of plastic steel window samples

Download full size

Fig. 4. Classification accuracy of Fisher discrimination analysis in full spectrum, functional cluster area, and fingerprint area

Download full size

Fig. 5. Influence of penalty factor and gamma value on overall classification accuracy of plastic steel window samples

Download full size

Fig. 6. Classification accuracy of different batches of conch brand plastic steel window

Download full size

Table 1. Basic information of 126 samples

View table

Table 1. Basic information of 126 samples

Brand	Production place	Quantity
Jinpeng	Binhai，Tianjin	27
Hailuo	Tangshan，Hebei Province	30
Vica plastic	Songjiang，Shanghai	21
Shuangfu	Nan’an，Fujian Province	9
Huaihai profile	Huai’an，Jiangsu Province	9
Ruihao	Suzhou，Jiangsu Province	30

Table 2. Parameters of Fourier infrared spectrometer
View table
Table 2. Parameters of Fourier infrared spectrometer
Facility Nicolet 5700 Fourier transform infrared spectrometer
Parameter Number of scans 64
Spectral resolution /cm^-1 2
Measuring range /cm^-1 4000-400
Dynamic adjustment /s^-1 130000
Noise-signal ratio 50000∶1

Table 3. Principal component analysis scores of Fourier transform infrared spectral data

View table

Table 3. Principal component analysis scores of Fourier transform infrared spectral data

Full range			Fingerprint region			Functional area
Component	Eigenvalue	Cumulative variance contribution rate /%	Component	Eigenvalue	Cumulative variance contribution rate /%	Component	Eigenvalue	Cumulative variance contribution rate /%
1	844.772	90.447	1	202.71	86.628	1	660.351	94.336
2	52.312	96.048	2	25.529	97.538	2	30.376	98.675
3	24.765	98.699	3	1.769	98.294	3	4.594	99.332
4	3.805	99.106	4	1.501	98.936	4	1.986	99.615
5	2.551	99.379	5	1.079	99.397
6	1.526	99.543
7	1.144	99.665

Table 4. Influence of penalty factor and gamma value on overall classification accuracy of plastic steel window samples

View table

Table 4. Influence of penalty factor and gamma value on overall classification accuracy of plastic steel window samples

Penalty factor	RBF σ
Penalty factor	0.2	0.5	0.8	1	1.2	1.5	2	2.5	3
1	87.5	87.5	87.5	90.62	90.62	90.62	90.62	93.75	96.88
5	90.62	90.62	93.75	93.75	96.88	96.88	96.88	96.88	100
10	90.62	93.75	96.88	96.88	96.88	96.88	96.88	100	100
50	96.88	100	100	100	100	100	100	100	100
100	100	100	100	100	100	100	100	100	100

Table 5. SVM model classification results of 6 brand samples

View table

Table 5. SVM model classification results of 6 brand samples

Sample set	Brand	Jinpeng	Hailuo	Vica plastic	Shuangfu	Huaihai profile	Ruihao	Accuracy
Training sample	Jinpeng	24						100
	Hailuo		21					100
	Vica plastic			18				100
	Shuangfu				3			100
	Huaihai profile					9		100
	Ruihao						21	100
Test sample	Jinpeng	3						100
	Hailuo		9					100
	Vica plastic			3				100
	Shuangfu				3			100
	Huaihai profile					3		100
	Ruihao						9	100

Tools

Get Citation

Copy Citation Text

Zhen Zhang, Jifen Wang, Pengwu Lu, Zhaokui Fu. Multifeature Automatic Spectral Classification of Plastic Steel Window Based on Machine Learning Model at Molecular Level[J]. Laser & Optoelectronics Progress, 2022, 59(16): 1630005

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites