Wood Recognition and Quality Imaging Inspection Systems

Table 1

Comparison of various works focused on wood classification methods, based on various classifiers.

Reference	Year	Methods	Features	Classifier	Categories	Dataset (train./valid./test)	Dataset expansion	Accuracy (%)
[8]	2007	—	GLCM	MLP	5	250 (-/-/-)	—	72.00
[9]	2008	—	GLCM	MLP	20	2100 (93/7/-)	—	95.00
[51]	2008	Img. Segm.+SVM/KNN/NN/LDA	Img. Segm.	NN	7	101 (99/1/-)	—	80.00
[50]	2008	GLCM/DGLC+NN/KNN	GLCM	MLP	5	500 (50/50/-)	—	72.80
[16]	2009	Gabor/GLCM+KNN	Gabor	KNN	6	12 (50/50/-)	aug.	85.00
[12]	2009	GLCM+thresholding	GLCM	Thresh.	6	510 (88/12/-)	—	80.00
[21]	2010	—	LBP	NN	37	3700 (80/20/-)	—	96.60
[22]	2010	GLCM/Gabor	GLCM+Gabor	MLP	30	3000 (90/10/-)	—	90.30
[47]	2010	GLCM/Gabor	GLCM	NN	20	2010 (90/10/-)	—	91.00
[11]	2010	Color/GLCM+MLP	Color+GLCM	MLP	22	1270 (32/18/50)	—	80.80
[18]	2013	GLB/SPPD/BGLAM+LDA	GLCM+SPPD+BGLAM	LDA	52	5200 (70/30/-)	—	98.69
[19]	2013	GLCM/Gabor	GLCM+Gabor	MLP	25	500 (80/10/10)	—	92.60
[7]	2013	LBP/GLCM+KNN/LDA/SVM	LBP	LDA	112	2240 (40/20/40)	—	80.70
[10]	2014	Color/LBP/LPQ/CLBP/Gabor	CLBP, color, LBP	SVM	41	2942 (35/15/50)	sub.	97.64
[52]	2014	—	GLCM	Corr.	10	1050 (95/5/-)	—	95.00
[48]	2014	—	—	CNN	41	3M (5f-cross-valid.)	sub.	97.32
[20]	2016	BGLAM+SVM	BGLAM	SVM	52	5200 (90/10/-)	—	99.84
[13]	2018	—	—	CNN	3	1115 (70/15/15)	sub.	99.00
[14]	2018	—	—	CNN	23	1006 (5f-cross-valid.)	sub.	97.81
[15]	2018	DW+LBP+SVM	LBP	SVM	3	4320 (66/34/-)	sub., aug.	85.00

(A) NN: (Artificial) Neural Network; BGLAM: Basic Gray Level Aura Matrix; CLBP: Completed Local Binary Patterns; CNN: convolutional neural network; GLCM: gray level covariance matrix; KNN: -Nearest Neighbor; LBP: Local Binary Patterns; LDA: Linear Discriminant Analysis; LPQ: Local Phase Quantization; MLP: multilayer perceptron; NN: neural network; SVM: support vector machine; aug.: augmentation; sub.: subimages; : an indication that dataset is available online for free.