Improved Thermal Infrared Image Downscaling Model and Its Application

Date	Data name	Aera	Remark
September 4, 2014	LC08_L1TP_123032_20140904_20170420_01_T1LC08_L1TP_123033_20140904_20170420_01_T1	Beijing	Visible-near infrared (30 m)Thermal infrared (100 m)
September 18, 2014	HJ1A-CCD1-4-68-20140918-L20001197797HJ1A-CCD1-4-64-20140918-L20001197817FY3C_MERSI_GBAL_L1_20140918_0235_0250M_MSFY3C_MERSI_GBAL_L1_20140918_0235_1000M_MS	Beijing	CCD (30 m) FY3/MERSI (250 m)FY3/MERSI (1000 m)

Table 2. Surface feature factors

View table

Table 2. Surface feature factors

Index	Definition	Meaning
NDVI	N_NDVI= $\frac{N_{NIR} - R_{RED}}{N_{NIR} + R_{RED}}$	N_NIR: near infrared bandR_RED: red band
UI	U_UI= $\frac{S_{SWIR 2} - N_{NIR}}{S_{SWIR 2} + N_{NIR}}$	S_SWIR2: mid-infrared bandN_NIR: near infrared band
NDBI	N_NDBI= $\frac{S_{SWIR 1} - N_{NIR}}{S_{SWIR 1} + N_{NIR}}$	S_SWIR1: mid-infrared bandN_NIR: near infrared band
BCI	B_BCI= $\frac{(T_{TC 1} + T_{TC 2}) / 2 - T_{TC 3}}{(T_{TC 1} + T_{TC 2}) / 2 + T_{TC 3}}$	T_TC1: normalized value of the first tasselled cap transformation components T_TC2: normalized value of the second tasselled cap transformation componentsT_TC3: normalized value of the third tasselled cap transformation components

Table 3. [in Chinese]
View table
Table 3. [in Chinese]
Index Building Vegetation
Verified data (270 m)
Verified data (90 m)
NDVI
BCI
NDBI
UI

Table 4. Quantitative evaluation indices

View table

Table 4. Quantitative evaluation indices

Evaluation index	Definition	Meaning
RMSE	$R_{RMSE} = \sqrt[]{\frac{\overset{n}{\sum_{i = 1}} (y_{i}^{ob} - y_{i}^{pre})^{2}}{n}}$	The smaller, the better
MAE	$M_{MAE} = \frac{1}{n} \overset{n}{\sum_{i = 1}} (y_{i}^{ob} - y_{i}^{pre})$	The smaller, the better
CC	C_CC= $\frac{\overset{n}{\sum_{i = 1}} (y_{i}^{ob} - {\bar{μ}}_{ob}) (y_{i}^{pre} - {\bar{μ}}_{pre})}{\sqrt[]{\overset{n}{\sum_{i = 1}} (y_{i}^{ob} - {\bar{μ}}_{ob})^{2} \sqrt[]{\overset{n}{\sum_{i = 1}} (y_{i}^{pre} - {\bar{μ}}_{pre})^{2}}}}$	The bigger, the better

Table 5. Quantitative evaluation for experimental results (single factor)

View table

Table 5. Quantitative evaluation for experimental results (single factor)

Index	Type	RMSE /K	MAE /K	CC
UI	Global	0.834	0.007	0.995
	Vegetation	0.752	0.015	0.995
	Building	0.924	-0.065	0.994
BCI	Global	0.899	-0.001	0.994
	Vegetation	0.956	-0.005	0.996
	Building	1.145	0.079	0.990
NDBI	Global	0.882	0.002	0.995
	Vegetation	0.854	0.021	0.997
	Building	0.994	-0.065	0.990
NDVI	Global	0.813	0.005	0.996
	Vegetation	0.653	0.026	0.998
	Building	1.224	-0.135	0.996

Table 6. Quantitative evaluation for experimental results (multiple factors)
View table
Table 6. Quantitative evaluation for experimental results (multiple factors)
Index RMSE /K MAE /K CC
N_NDVI+U_UI 0.763 0.006 0.997
N_NDVI +B_BCI 0.828 0.005 0.995
N_NDVI+N_NDBI 0.788 0.002 0.996

Table 7. Quantitative evaluation for urban heat island

View table

Table 7. Quantitative evaluation for urban heat island

Index	Type	MAE /K	RMSE /K	Thermal centroid	Thermal centroid change rate /%
Real data	HFII	0	0	78144	0
Real data	HI	0	0	97835	0
NDVI	HFII	-0.129	1.005	85254	9.098
NDVI	HI	-0.020	0.963	97652	-0.187
BCI	HFII	-0.130	1.033	83803	7.241
BCI	HI	0.192	0.925	99907	2.130
UI	HFII	-0.136	1.108	84393	7.996
UI	HI	0.168	0.998	97995	0.164
NDBI	HFII	-0.151	1.128	84878	8.617
NDBI	HI	0.161	1.058	98125	0.296
N_NDVI+B_BCI	HFII	-0.112	0.852	83696	7.105
N_NDVI+B_BCI	HI	0.004	0.804	99898	2.110
N_NDVI+U_UI	HFII	-0.132	0.751	84383	7.983
N_NDVI+U_UI	HI	0.004	0.751	97981	0.149
N_NDVI+N_NDBI	HFII	-0.051	0.886	84878	8.617
N_NDVI+N_NDBI	HI	0.024	0.850	98140	0.312

Table 8. Parameters of urban heat island at different time
View table
Table 8. Parameters of urban heat island at different time
Evaluation index September 4 September 18
HFII HI HFII HI
Thermal centroid 78144 97835 123241 142931
Thermal centroid change rate /% 13.89 17.39 21.91 25.41

Tools

Get Citation

Copy Citation Text

Wenqi Zhang, Cailan Gong, Yong Hu, Wentao Song, Dingbo Kuang. Improved Thermal Infrared Image Downscaling Model and Its Application[J]. Acta Optica Sinica, 2019, 39(9): 0928001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites