Integration of case-based reasoning and object-based image classification to classify SPOT images: a case study of aquaculture land use mapping in coastal areas of Guangdong province, China

Figures & data

Figure 1. A map showing our study area. The parallelograms show footprints of the SPOT images used in this study. These images were acquired on (A) 3 December 2003, (B) 5 November 2004, (C) 10 October 2004, (D) 10 October 2004, (E) 16 October 2004, (F) 16 October 2004, (G) 15 October 2004, (H) 23 May 2004, (I) 11 December 2004, (J) 14 September 2004, (K) 1 January 2005, (L) 23 October 2003, (M) 23 October 2003, (N) 19 December 2006, (O) 9 October 2004.

Figure 2. A flow chart showing the procedures to segment images and classify image objects in this study.

Table 1. Attributes used to describe the segmented objects in this study

No.	Indicator	Components	Description
1	S 1	Spectral	Average of all digital numbers within a specific object on band 4
2	S 2	Spectral	Average of all digital numbers within a specific object on all 4 bands
3	G 1	Geometric	Shape index (4): smooth objects have a lower shape index
4	T 1	Textural	The grey level co-occurrence matrix (GLCM) homogeneity (all direction) of band 3
5	T 2	Textural	GLCM contrast (all direction) of band 1
6	T 3	Textural	GLCM entropy (45°) of band 1
7	T 4	Textural	GLCM dissimilarity (all direction) of band 4
8	D 1	Spatial	Nearest distance: perpendicular distance from the centroid of a land object to the costal line

Table 2. Past cases with their attributes in the case library

	Attributes
Case ID	S 1	S 2	G 1	T 1	T 2	T 3	T 4	D 1	Scale	LC*
1	67.11	93.31	1.688	0.2087	368.8	6.654	10.93	3664	40	1
2	54.66	86.71	1.600	0.3435	331.7	5.637	7.837	3367	40	1
3	60.00	83.98	1.263	0.2567	258.8	5.261	11.30	1666	40	2
4	41.84	64.35	1.624	0.4082	166.0	4.827	7.220	348.4	40	3
…	…	…	…	…	…	…	…	…	…	…
6976	19.72	44.05	1.520	0.5969	10.76	3.266	3.233	0	80	1
6977	68.58	105.4	1.922	0.170	183.6	7.603	8.172	1436	80	4
Note: *LC (land use category) with 1, 2, 3, and 4 denoting aquiculture, pond, reservoir, and salt field, respectively.

Table 3. Number and total area of old cases per land use type in the case library

		Land use types
Scale	Case	Aquaculture	Pond	Reservoir	Salt field	Total
80	Number	1191	112	132	29	1464
	Area (km^2)	546.45	16.43	21.96	8.21	593.05
40	Number	4656	460	274	123	5513
	Area (km^2)	657.48	22.98	22.27	14.47	717.20

Figure 3. A flow chart showing image classification procedures using CBR.

Table 4. Error matrixes of the 10-fold cross-validation experiment

		Reference
Area (km^2)		Aquaculture	Pond	Reservoir	Salt field	User's accuracy (%)
Test 1
Classification	Aquaculture	122.32	0.55	0.87	0.60	98.38
	Pond	0.72	2.50	0.17	0.00	73.76
	Reservoir	0.21	0.96	3.34	0.00	74.11
	Salt field	1.06	0.00	0.00	1.51	58.75
	Others	0.03	0.00	0.00	0.00
	Producer's accuracy (%)	98.40	62.40	76.33	71.55
	Overall accuracy (%)					96.17
Test 2
Classification	Aquaculture	118.17	1.26	0.45	2.01	96.95
	Pond	0.63	2.48	0.53	0.00	68.10
	Reservoir	0.28	0.04	3.26	0.00	91.06
	Salt field	0.50	0.00	0.00	1.09	68.43
	Others	0.41	0.00	0.00	0.00
	Producer's accuracy (%)	98.82	65.59	76.89	35.15
	Overall accuracy (%)					95.34
Test 3
Classification	Aquaculture	128.96	0.38	0.38	1.71	98.12
	Pond	1.07	2.09	0.26	0.00	61.13
	Reservoir	0.10	0.36	4.58	0.00	90.90
	Salt field	1.07	0.00	0.00	0.71	39.88
	Others	0.09	0.00	0.00	0.00
	Producer's accuracy (%)	98.29	73.88	87.78	29.26
	Overall accuracy (%)					96.18
Test 4
Classification	Aquaculture	122.01	0.70	0.70	0.55	98.43
	Pond	0.68	3.04	0.49	0.00	72.17
	Reservoir	0.07	0.11	3.26	0.00	94.77
	Salt field	0.36	0.00	0.00	1.07	75.02
	Others	0.35	0.00	0.00	0.00
	Producer's accuracy (%)	99.10	79.02	73.21	66.21
	Overall accuracy (%)					97.00
Test 5
Classification	Aquaculture	112.34	0.61	0.11	0.39	99.02
	Pond	0.31	2.89	0.32	0.00	82.13
	Reservoir	0.25	0.28	3.45	0.00	86.57
	Salt field	1.30	0.00	0.00	1.07	45.14
	Others	0.38	0.00	0.00	0.00
	Producer's accuracy (%)	98.37	76.36	88.84	73.26
	Overall accuracy (%)					96.80
		Ground truth
Test 6
Classification	Aquaculture	122.76	0.75	0.83	0.39	98.42
	Pond	0.87	2.77	0.31	0.00	70.13
	Reservoir	0.24	0.28	3.40	0.00	86.80
	Salt field	0.52	0.00	0.00	1.67	76.15
	Others	0.24	0.00	0.00	0.00
	Producer's accuracy (%)	98.68	73.02	74.93	80.92
	Overall accuracy (%)					96.72
Test 7
Classification	Aquaculture	128.59	0.68	0.85	1.04	98.04
	Pond	0.45	3.46	0.37	0.00	80.97
	Reservoir	0.77	0.55	2.64	0.00	66.63
	Salt field	0.84	0.00	0.00	1.42	62.79
	Others	0.05	0.00	0.00	0.00
	Producer's accuracy (%)	98.42	73.76	68.52	57.61
	Overall accuracy (%)					96.05
Test 8
Classification	Aquaculture	117.22	0.83	0.52	1.06	97.98
	Pond	0.36	2.49	0.06	0.00	85.29
	Reservoir	0.05	0.53	3.22	0.00	84.72
	Salt field	0.67	0.00	0.00	0.90	57.32
	Others	0.06	0.00	0.00	0.00
	Producer's accuracy (%)	99.08	64.73	84.61	45.78
	Overall accuracy (%)					96.75
Test 9
Classification	Aquaculture	118.41	1.54	0.25	0.48	98.12
	Pond	0.85	3.48	0.47	0.00	72.52
	Reservoir	0.48	0.28	4.40	0.00	85.28
	Salt field	0.68	0.00	0.00	2.44	78.16
	Others	0.10	0.00	0.00	0.00
	Producer's accuracy (%)	98.33	65.66	85.88	83.60
	Overall accuracy (%)					96.17
Test 10
Classification	Aquaculture	113.54	0.48	0.06	0.17	99.38
	Pond	0.60	2.64	0.26	0.00	75.45
	Reservoir	0.11	0.41	4.43	0.00	89.52
	Salt field	0.59	0.00	0.00	2.39	80.12
	Others	0.56	0.00	0.00	0.00
	Producer's accuracy (%)	98.86	74.78	93.41	93.50
	Overall accuracy (%)					97.44

Table 5. Confusion matrix of the CBR classification results of Jiangmen test area

		Reference
Area (km^2)		Aquaculture	Pond	Reservoir	Salt field	Others	User's accuracy (%)
Classification	Aquaculture	30.05	0.34	0.72	0	23.59	54.93
	Pond	0.19	0.02	0.09	0	0.55	2.77
	Reservoir	0.00	0.00	0.00	0	0.42	0.02
	Salt field	0.64	0.00	0.00	0	2.76	0.00
	Others	9.49	0.49	0.11	0	149.80	93.69
	Producer's accuracy (%)	74.43	2.77	0.01		84.58
	Overall accuracy (%)						82.03

Table 6. Confusion matrix of the CBR classification results of Shatian test area

		Reference
Area (km^2)		Aquaculture	Pond	Reservoir	Salt field	Others	User's accuracy (%)
Classification	Aquaculture	6.27	0.55	0.16	0	3.20	61.55
	Pond	0.00	0.17	0.00	0	0.17	50.28
	Reservoir	0.00	0.03	0.03	0	0.09	22.20
	Salt field	1.74	0.06	0.00	0	2.82	0.00
	Others	0.50	0.80	0.07	0	90.20	98.50
	Producer's accuracy (%)	73.69	10.81	12.21		93.49
	Overall accuracy (%)						90.46

Table 7. Confusion matrix of the MLC classification results of Jiangmen test area

		Reference
Area (km^2)		Aquaculture	Pond	Reservoir	Others	User's accuracy (%)
Classification	Aquaculture	23.48	0.19	0.03	26.80	46.50
	Pond	9.26	0.36	0.18	12.36	1.63
	Reservoir	1.52	0.16	0.63	0.65	21.13
	Others	6.11	0.15	0.09	137.30	95.58
	Producer's accuracy (%)	58.17	42.02	68.16	77.52
	Overall accuracy					73.78

Figure 4. CBR classification results of the two test areas: (A) Jiangmen and (B) Shatian.

Table 8. Confusion matrix of the MLC classification results of Shatian test area

		Reference
Area (km^2)		Aquaculture	Pond	Reservoir	Others	User's accuracy (%)
Classification	Aquaculture	5.31	0.34	0.00	0.73	83.20
	Pond	1.50	0.61	0.02	8.23	5.86
	Reservoir	1.33	0.31	0.15	0.54	6.57
	Others	0.37	0.36	0.09	86.99	99.07
	Producer's accuracy (%)	62.44	37.54	57.04	90.16
	Overall accuracy					87.08