Similar Wood Species Classification Within Pterocarpus Genus Using Feature Fusion

Cheng-kun WANG; Peng ZHAO; Xiang-hua LI

doi:10.3964/j.issn.1000-0593(2022)07-2247-08

Journals >Spectroscopy and Spectral Analysis >Volume 42 >Issue 7 >Page 2247 > Article

Spectroscopy and Spectral Analysis
Vol. 42, Issue 7, 2247 (2022)

Cheng-kun WANG^2、*, Peng ZHAO^{1、1; 2; *;}, and Xiang-hua LI^2、2;

Author Affiliations

¹1. College of Computer Science and Electronics, Guangxi University of Science and Technology, Liuzhou 545006, China

²2. College of Information and Computer Engineering, Northeast Forestry University, Harbin 150040, China

show less

DOI: 10.3964/j.issn.1000-0593(2022)07-2247-08 Cite this Article

Cheng-kun WANG, Peng ZHAO, Xiang-hua LI. Similar Wood Species Classification Within Pterocarpus Genus Using Feature Fusion[J]. Spectroscopy and Spectral Analysis, 2022, 42(7): 2247 Copy Citation Text

show less

Fig. 1. Wood feature acquisition platforms
(a): Spectrum acquisition; (b): RGB image acquisition

Download full size | View in the Article

Images of Pterocarpus section(a): Pterocarpus macrocarpus; (b): Pterocarpus erinaceus; (c): Pterocarpus antunesii;(d): Pterocarpus soyauxii; (e): Pterocarpus tinctorius

Fig. 2. Images of Pterocarpus section
(a): Pterocarpus macrocarpus; (b): Pterocarpus erinaceus; (c): Pterocarpus antunesii;(d): Pterocarpus soyauxii; (e): Pterocarpus tinctorius

Download full size | View in the Article

Fig. 3. Original spectra and SNV corrected spectra
(a): Original spectra; (b): SNV corrected spectra

Download full size | View in the Article

Fig. 4. Feature dimension and accuracy

Download full size | View in the Article

Fig. 5. Identification of test set samples

Download full size | View in the Article

Fig. 6. Average spectral curves of cross sections of 30 wood species
(a): The first 15 tree species in Table 7;(b): The last 15 tree species in Table 7

Download full size | View in the Article

Fig. 7. Transverse section of 30 wood species (the number of each illustrations corresponding to Table 7)

Download full size | View in the Article

序号	中文	拉丁文
1	大果紫檀	Pterocarpus macrocarpus
2	刺猬紫檀	Pterocarpus erinaceus
3	安氏紫檀	Pterocarpus antunesii
4	非洲紫檀	Pterocarpus soyauxii
5	赞比亚紫檀	Pterocarpus tinctorius

Table 1. Sample data

View in the Article

切面	横切面		弦切面		径切面
方法	维度	正确率/%	维度	正确率/%	维度	正确率/%
PCA	28	94.40	26	90.40	30	93.20
KPCA	76	89.20	56	86.40	60	88.40
Laplacian	32	82.80	28	78.40	36	82.80
SPA	选择波段/nm						正确率/%
横切面	376.64, 378.97, 386.31, 495.03, 630.43, 695.46, 806.52, 1 010.28, 1 017.95, 1 025.29, 1 026.29						92.80
弦切面	378.30, 408.65, 529.04, 600.41, 713.14, 849.208, 1 019.29, 1 021.96, 1 025.29, 1 025.96						91.60
径切面	377.64, 379.97, 586.74, 745.15, 937.91, 995.28, 1 016.62, 1 025.29, 1 025.96, 1 026.29						93.60

Table 2. The highest accuracies under different dimension reduction methods

View in the Article

方法	横切面	弦切面	径切面
GLCM	67.60	63.20	64.80
LBP	80.00	77.60	74.00
I-BGLAM	75.60	72.40	75.60
MFS	62.00	68.00	63.20

Table 3. Accuracies of wood species using textures features (%)

View in the Article

融合策略	concat				sum
方法	PCA	KPCA	Laplacian	SPA	PCA	KPCA	Laplacian	SPA
	横切面
GLCM	93.60	82.80	82.40	97.60	92.40	82.80	81.60	96.00
I-BGLAM	99.20	88.40	83.20	98.40	98.80	90.00	82.00	97.20
MFS	98.00	86.80	86.40	99.20	92.80	83.20	86.00	96.40
LBP	96.80	87.60	88.00	98.00	93.60	89.20	90.00	95.20
	弦切面
GLCM	96.00	92.80	92.40	98.80	94.80	92.40	90.40	97.20
I-BGLAM	99.20	91.20	84.00	98.80	99.20	88.80	82.00	98.40
MFS	90.80	80.80	78.80	98.40	89.20	78.40	78.40	92.00
LBP	93.60	89.60	83.20	98.40	92.00	88.80	84.80	95.60
	径切面
GLCM	97.60	91.20	90.80	98.40	98.00	92.80	90.80	98.40
I-BGLAM	98.80	89.60	88.00	99.20	98.80	89.20	86.80	99.20
MFS	92.40	82.40	80.80	96.40	90.40	81.20	82.40	92.80
LBP	99.20	88.00	86.80	98.80	96.00	86.40	85.20	97.20

Table 4. Accuracies of “concat” and “sum” fusion schemes (%)

View in the Article

方法	正确率/%
CNN	80.00
颜色矩	74.40
SPPD+I-BGLAM	77.60
Fuzzy+SPPD+I-BGLAM	73.60
GA	56.00
GA+KDA	57.60
本方法	99.20

Table 5. Comparison of accuracies between other wood recognition methods and the method presented in this paper

View in the Article

	方法	时间/s
纹理	GLCM	0.017
	I-BGLAM	0.032
	MFS	1.32
	LBP	0.033
光谱	PCA	0.002 5
	KPCA	0.000 14
	Laplacian	0.000 71
	SPA	0.72
融合	CCA	0.002 2

Table 6. Extraction time of single sample features

View in the Article

序号	名称	属	拉丁语
1	海棠木	红厚壳属	Calophyllum inophyllum
2	香樟木	樟木属	Cinnamomum camphora
3	大非洲楝	非洲楝属	Entandrophragma candoLaplaciani
4	美洲白蜡木	白蜡树属	Fraxinus chinensis
5	水曲柳	白蜡树属	Fraxinus mandshurica
6	古夷苏木	古夷苏木属	Guibourtia demeusei
7	双柱苏木	古夷苏木属	Guibourtia ehie
8	帕利印茄	印茄属	Intsia bijuga
9	黑核桃	核桃树	Juglans nigra
10	落叶松	落叶松属	Larix gmelinii
11	黑芯木莲	木莲属	Magnolia fordiana
12	非洲崖豆木	崖豆藤属	MiLaplacianttia laurentii
13	云杉	云杉属	Picea asperata
14	辐射松	松属	Pinups radiata
15	红松	松属	Pinus koraiensis
16	马尾松	松属	Pinus massoniana
17	樟子松	松属	Pinus sylvestris
18	番龙眼	番龙眼属	Pometia pinnata
19	花旗松木	黄杉属	Pseudotsuga menziesii
20	柞木	麻栎属	Quercus mongolica
21	麻栎	麻栎属	Quercus acutissima
22	刺槐	刺槐	Robinia pseudoacacia
23	无齿婆罗双	婆罗双属	Shorea contorta
24	平滑娑罗双	婆罗双属	Shorea laevis
25	槐树	槐树属	Sophora japonica
26	桃花芯	桃花心木属	Swietenia mahagoni
27	缅甸柚木	柚木属	Tectona grandis
28	榄仁木	榄仁树属	Terminalia cattapa
29	榆树	榆树属	Ulmus glabra
30	油桐	油桐属	Vernicia fordii

Table 7. Details of 30 wood species samples

View in the Article

	方法	正确率/%
纹理特征	I-BGLAM LBP	67.14 65.89
光谱特征	PCA SPA	91.89 93.09
融合特征	PCA+I-BGLAM PCA+LBP SPA+I-BGLAM SPA+LBP	97.77 95.20 96.57 98.29

Table 8. Classification accuracy of 35 tree species data

Cheng-kun WANG, Peng ZHAO, Xiang-hua LI. Similar Wood Species Classification Within Pterocarpus Genus Using Feature Fusion[J]. Spectroscopy and Spectral Analysis, 2022, 42(7): 2247

Download Citation

Tools

Save the article for my favorites

Paper Information