Pigment Classification Method of Mural Multi-Spectral Image Based on Multi-Scale Superpixel Segmentation

Yamin Chen; Ke Wang; Zhan Wang; Huiqin Wang; Yuan Li; Gang Zhen

doi:10.3788/LOP240671

Laser & Optoelectronics Progress, Volume. 61, Issue 18, 1837017(2024)

Pigment Classification Method of Mural Multi-Spectral Image Based on Multi-Scale Superpixel Segmentation

Yamin Chen¹, Ke Wang¹, Zhan Wang², Huiqin Wang^1、*, Yuan Li³, and Gang Zhen²

¹College of Information and Control Engineering, Xi'an University of Architecture and Technology, Xi'an 710055, Shaanxi , China

²Shaanxi Institute for the Preservation of Cultural Heritage, Xi'an 710075, Shaanxi , China

³Xi'an Museum, Xi'an 710074, Shaanxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. Schematic diagram of multi-scale superpixel segmentation pigment classification method

Download full size

Fig. 2. Superpixel segmentation effect on mural pseudo color images at different scales. (a) K=50; (b) K=100; (c) K=200; (d) K=500

Download full size

Fig. 3. Simulated mural multispectral imaging data. (a) Data cube; (b) multispectral imaging data for each channel

Download full size

Fig. 4. Simulated mural image. (a) Pseudo color chart; (b) true value chart

Download full size

Fig. 5. Comparison diagram of pseudo color effects. (a) OIF; (b) MI; (c) PCA; (d) ABS

Download full size

Fig. 6. Simulated mural superpixel segmentation results. (a) SLIC; (b) SLIC0; (c) SNIC; (d) NMI_SLIC; (e) LSC; (f) proposed

Download full size

Fig. 7. Impact of different scale numbers on classification accuracy of proposed method

Download full size

Fig. 8. Classification results comparison of different algorithms on simulated murals. (a) MDC; (b) SID; (c) SAM; (d) SVM;(e) SuperPCA; (f) SLIC_SVM; (g) MSP_SSA; (h) proposed

Download full size

Fig. 9. Real mural data. (a) The 13th his holiness true color image; (b) multi spectral data of target area; (c) training set distribution

Download full size

Fig. 10. Classification results of different comparison algorithms on real murals. (a) MDC; (b) SID; (c) SAM; (d) SVM; (e) SuperPCA; (f) SLIC_SVM; (g) MSP_SSA; (h) proposed

Download full size

Table 1. Multispectral image band numbers and corresponding index table

View table

Table 1. Multispectral image band numbers and corresponding index table

No.	Band /nm	Index	No.	Band /nm	Index
Band 1	400	17.9	Band 9	660	30.8
Band 2	420	25.8	Band 10	680	30.9
Band 3	460	45.3	Band 11	700	33.0
Band 4	520	30.2	Band 12	740	28.3
Band 5	540	31.8	Band 13	780	43.7
Band 6	580	46.7	Band 14	820	32.3
Band 7	620	35.7	Band 15	860	27.8
Band 8	640	39.2	Band 16	900	24.8

Table 2. Performance evaluation of band selection methods
View table
Table 2. Performance evaluation of band selection methods
Index Method
OIF MI PCA ABS
AIE 14.1193 14.1359 13.9826 14.1121
ARE 0.8888 0.4395 0.0398 0.8955
ACC 0.0133 0.1200 0.9883 0.0127

Table 3. Classificiation performance of proposed method at different superpixel segmentation scales

View table

Table 3. Classificiation performance of proposed method at different superpixel segmentation scales

Category	Segmentation scale /%
Category	-4	-3	-2	-1	0	1	2	3	4
Mercuric sulfide	89.94	78.92	92.38	94.11	95.94	66.25	54.11	69.00	52.03
Lazurite	96.26	96.26	96.26	96.26	96.26	96.26	93.96	93.65	92.98
Minium	94.50	94.50	89.33	93.75	78.45	0.00	0.00	0.00	0.00
Mineral green	73.07	84.25	83.23	78.26	79.93	75.27	86.90	71.66	86.22
Chrome yellow	93.60	97.78	97.78	97.78	98.00	97.45	97.31	97.58	96.87
Graphite	79.07	93.55	94.27	95.84	95.35	95.30	95.16	92.05	89.55
Background	94.58	99.88	98.25	99.15	99.15	97.91	97.52	95.23	95.48
A_OA /%	95.23	97.14	95.91	97.94	98.14	97.80	95.76	95.59	93.61
A_AA /%	92.79	93.53	95.76	95.67	96.93	95.19	91.49	92.20	87.27
Kappa	0.923	0.955	0.949	0.949	0.965	0.945	0.925	0.923	0.912

Table 4. Comparison of classification accuracy of different algorithms

View table

Table 4. Comparison of classification accuracy of different algorithms

Category	Classification accuracy /%
Category	MDC	SID	SAM	SVM	SuperPCA	SLIC_SVM	SSA_SVM	Proposed
Mercuric sulfide	92.37	93.56	92.16	96.88	95.36	95.98	96.80	99.17
Lazurite	82.33	92.23	89.52	91.46	94.83	94.80	92.96	98.74
Minium	20.69	85.72	91.17	87.50	84.59	75.52	89.68	95.43
Mineral green	50.26	86.26	87.33	91.28	92.35	92.56	87.49	98.22
Chrome yellow	93.56	96.33	93.87	98.46	98.85	98.55	98.67	99.41
Graphite	80.52	91.36	93.97	92.49	94.60	94.45	93.63	98.75
Background	89.22	93.48	95.58	94.11	99.12	99.54	99.57	99.81
A_OA /%	69.46	93.23	89.78	90.36	93.55	96.46	90.49	98.84
A_AA /%	72.70	91.28	91.94	93.17	94.24	93.06	94.12	98.50
Kappa	58.6	89.2	89.2	91.3	95.8	95.8	88.3	98.1

Table 5. Comparison of scales of subjective image evaluation

View table

Table 5. Comparison of scales of subjective image evaluation

Level	Absolute measurement scale	Relative measurement scale
1	Very good	The best in the group
2	Better	Better than the average in the group
3	Generally	Average in the group
4	Poor	Worse than the average in the group
5	Very bad	Worst in the group

Table 6. Evaluation results of dual stimulation injury classification method
View table
Table 6. Evaluation results of dual stimulation injury classification method
Algorithm Result
MDC Very bad
SID Poor
SAM Poor
SVM Generally
SuperPCA Better
SLIC_SVM Better
MSP_SSA Better
Proposed Very good

Tools

Get Citation

Copy Citation Text

Yamin Chen, Ke Wang, Zhan Wang, Huiqin Wang, Yuan Li, Gang Zhen. Pigment Classification Method of Mural Multi-Spectral Image Based on Multi-Scale Superpixel Segmentation[J]. Laser & Optoelectronics Progress, 2024, 61(18): 1837017

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites