X-Ray Fluorescence Spectroscopy Combined with MLP-DNN for Rapid Classification of Cosmetic Paper Packaging Boxes

Xingyu Ma; Xiaoguang Hu; Hong Jiang; Baofa Hu; Shijian Wang; Ji Man

doi:10.3788/LOP242272

Laser & Optoelectronics Progress, Volume. 62, Issue 13, 1330001(2025)

X-Ray Fluorescence Spectroscopy Combined with MLP-DNN for Rapid Classification of Cosmetic Paper Packaging Boxes

Xingyu Ma¹, Xiaoguang Hu^1、*, Hong Jiang^2、**, Baofa Hu¹, Shijian Wang¹, and Ji Man³

Author Affiliations

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

²Center of Forensic Science Beijing Hui Zheng Zhuo Yue Technology Co.,Ltd., Beijing 102446, China

³Beijing Huayi Honrizon Technology Co., Ltd., Beijing 100123, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Manual classification results of cosmetic paper packaging boxes

Download full size

Fig. 2. Variation in clustering coefficient with categorical number

Download full size

Fig. 3. Structure of MLP-DNN model

Download full size

Fig. 4. Comparison of accuracies and macro average indicators for MLP-DNN and random forest

Download full size

Fig. 5. Comparison of accuracies and macro average indicators for MLP-DNN, MLP, and DNN

Download full size

Table 1. Informations of experimental samples

View table

Table 1. Informations of experimental samples

Serial number	Brand	Source	Case colour	Serial number	Brand	Source	Case colour
1	Forever Key	Shanghai	White	32	HBN	Zhongshan	Pink
2	Kato-Kato	Suzhou	White	33	HBN	Zhongshan	Pink
3	Run Pei	Huzhou	White	34	Fog Island Forest	Shanghai	Pink
4	Black Tea	Guangzhou	White	35	Estee Lauder	France	Blue
5	Xi Mu Yuan	Zhejiang	White	36	L'Oréal	Suzhou	Blue
6	Menow	Shantou	White	37	Pu Qian Na	Guangzhou	Blue
7	La Roche-Posay	France	White	38	Florasis	Zhongshan	Blue
8	Ukiss	Shanghai	White	39	Rockingzoo	Shanghai	Green
9	La Roche-Posay	France	White	40	Watercome	Guangzhou	Green
10	Menow	Shantou	White	41	Everbab	Shanghai	Brown
11	Xun Hui Ji	Hainan	White	42	Chanel	Parries	Brown
12	Sha Qi Nuo	Shantou	Pink	43	Black Tea	Guangzhou	Yellow
13	AKF	Guangdong	Pink	44	An Ye Wu	Guangzhou	Brown
14	Veecci	Zhejiang	Pink	45	Marie Beauty	Shanghai	Brown
15	Sha Qi Nuo	Shantou	Pink	46	Jian Nan	Guangzhou	Brown
16	Jillleen	Jiangyin	Pink	47	Perrier	Hangzhou	Silver
17	Sha Qi Nuo	Shantou	Pink	48	Jing Bai Ku Shuang	Guangzhou	Silver
18	Ju Duo	Shanghai	Pink	49	Mo Fa Shi Jia	Guangzhou	Silver
19	HBN	Guangzhou	Pink	50	Cosmax	Shanghai	Silver
20	Fresh	America	Pink	51	Perrier	Hangzhou	Silver
21	Zuo Guang	Guangzhou	White	52	Pramy	Zhuhai	Silver
22	Curel	Tokyo	White	53	AKF	Zhongshan	Silver
23	Gu Yu	Guangzhou	White	54	Hai Chang	Danyang	Silver
24	Mei Ke	Jiangsu	White	55	Kiko	Shanghai	Silver
25	Perrier	Huzhou	White	56	Mei Ke	Shanghai	Silver
26	AFU	Guangzhou	White	57	La Mei La	Zhejiang	Black
27	Shu Dai Zi	Shanghai	White	58	La Mei La	Yiwu	Brown
28	RNW	Guangzhou	White	59	Shaqinuo	Shantou	White
29	La Roche-Posay	Suzhou	White	60	Zero	Shanghai	Pink
30	Fontti Fratti	Guangzhou	White	61	Hapsode	Hangzhou	Blue
31	RNW	Shanghai	White

Table 2. Results of reproducible experiments

View table

Table 2. Results of reproducible experiments

Element	Mean ± standard deviation /（μg·g^-1）	Kurtosis /（μg·g^-1）	Skewness /（μg·g^-1）	Relative standard deviation /%
Cl	203.600±8.181	-2.074	0.525	4.018
Ca	1293.200±31.930	0.270	0.267	2.469
Ti	849.300±20.451	0.895	1.186	2.408
Fe	1317.200±16.212	-0.856	0.108	1.231
Sr	43.600±1.506	-0.671	0.615	3.453
Ba	0±0	—	—	—

Table 3. Mass fractions of 6 elements in 10 samples
View table
Table 3. Mass fractions of 6 elements in 10 samples
Serial number Cl Ti Ba Ca Fe Sr
1 0 27 79 28444 793 48
2 0 50 0 42745 1090 56
3 0 0 0 50185 218 53
4 0 24 0 35693 815 44
5 0 89 0 37677 1124 56
6 0 122 85 33103 1010 49
7 0 199 0 32759 1051 48
8 0 798 0 1584 1241 43
9 0 254 0 12485 1023 50
10 0 678 0 14182 1154 46

Table 4. Training indicators of MLP-DNN model
View table
Table 4. Training indicators of MLP-DNN model
Indicator Value
Accuracy 0.89
Macro average Precision is 0.94， recall is 0.84， F1-score is 0.87
Weighted average Precision is 0.92， recall is 0.89， F1-score is 0.89

Table 5. Training indicators of random forest model
View table
Table 5. Training indicators of random forest model
Indicator Value
Accuracy 0.85
Macro average Precision is 0.68， recall is 0.75， F1-score is 0.71
Weighted average Precision is 0.74， recall is 0.85， F1-score is 0.78

Tools

Get Citation

Copy Citation Text

Xingyu Ma, Xiaoguang Hu, Hong Jiang, Baofa Hu, Shijian Wang, Ji Man. X-Ray Fluorescence Spectroscopy Combined with MLP-DNN for Rapid Classification of Cosmetic Paper Packaging Boxes[J]. Laser & Optoelectronics Progress, 2025, 62(13): 1330001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites