Method and apparatus for extracting mountain landscape buildings based on high-resolution remote sensing images
The present invention discloses a method and an apparatus for extracting mountain landscape buildings based on high-resolution remote sensing images. The method comprises: segmenting a remote sensing image, and extracting non-vegetation areas from the remote sensing image by using NDVI; segmenting the non-vegetation areas, and extracting building areas by using NDBI; segmenting the building areas again, and calculating a normalized difference build shadow index NSBI of each patch; calculating NSBI separator of each patch in the non-vegetation areas and setting a separator threshold, and extracting landscape building areas based on the threshold. In the present invention, by introducing a near infrared band in the remote sensing image spectrum, in which there is a significant difference between shadows and non-shadows, the influence of large shadow areas in mountainous shady areas in the remote sensing image on the result of extraction is reduced.
1. A method for extracting mountain landscapes buildings based on high-resolution remote sensing images, comprising the following steps:
S 1 : segmenting a remote sensing image into patches of a first scale by using a first scale parameter A 1 , calculating a normalized difference vegetation index NDVI of each patch, and extracting the segmented patches with NDVI greater than a first threshold T 1 as vegetation areas, while other patches as non-vegetation areas;
S 2 : segmenting the non-vegetation areas by using a second scale parameter A 2 to obtain patches of a second scale, calculating a normalized difference buildup index NDBI of each patch, and judging the patches of the second scale with NDBI greater than a second threshold T 2 as building areas, while regarding other patches as non-building areas;
S 3 : segmenting the non-vegetation areas by using a third scale parameter A 3 to obtain patches of a third scale, and calculating a normalized difference build shadow index NSBI of each patch;
S 4 : calculating a normalized difference build shadow index separator S x (NSBI) of each patch according to the NSBI, and extracting the areas with a separator greater than a third threshold T 3 as landscape buildings,
wherein the normalized difference build shadow index separator Sx (NSBI) is calculated with the following formula:
S
x
(
NSBI
)
=
∑
x
i
∈
n
(
x
)
,
m
x
i
(
l
)
<
m
x
(
l
)
B
(
x
,
x
i
)
(
m
x
(
NSBI
)
-
m
x
i
(
NSBI
)
)
∑
x
i
∈
n
(
x
)
B
(
x
,
x
i
)
wherein, x represents the current calculated patch, n(x) represents a set of all patches adjacent to the current calculated patch, B(x, x i ) is the length of a common side of the current calculated patch x and the adjacent patch x i , and m x (NSBI) represents the NSBI value of the calculated patch x;
wherein the first scale parameter A 1 >the second scale parameter A 2 >the third scale parameter A 3 .
2. The method for extracting mountain landscape buildings based on high-resolution remote sensing image according to claim 1 , wherein the method uses the Multiresolution Segmentation algorithm or Hyper-pixel Segmentation algorithm to segment the image.
3. The method for extracting mountain landscape buildings based on high-resolution remote sensing image according to claim 1 , wherein the normalized difference vegetation index NDVI in the step S 1 is calculated with the following formula:
NDVI
=
NIR
-
R
NIR
+
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
4. The method for extracting mountain landscape buildings based on high-resolution remote sensing image according to claim 1 , wherein the normalized difference build shadow index NSBI in the step S 2 is calculated with the following formula:
NDBI
=
NIR
+
R
NIR
-
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
5. The method for extracting mountain landscape buildings based on high-resolution remote sensing image according to claim 1 , wherein the normalized difference build shadow index NSBI of each patch is calculated with the following formula in the step S 3 :
NSBI
=
NDBI
*
NIR
=
(
NIR
+
R
)
*
NIR
NIR
-
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
6. The method for extracting mountain landscape buildings based on high-resolution remote sensing image according to claim 1 , wherein the first threshold T 1 , the second threshold T 2 and the third threshold T 3 are adaptively calculated with the OTSU method.
7. A computer device, comprising:
one or more processors;
a memory unit; and
one or more programs, which are stored in the memory unit and configured to be executed by said one or more processors, and, when the program is configured to execute a method for extracting mountain landscapes buildings based on high-resolution remote sensing images, the method comprising the following steps:
S 1 : segmenting a remote sensing image into patches of a first scale by using a first scale parameter A 1 , calculating a normalized difference vegetation index NDVI of each patch, and extracting the segmented patches with NDVI greater than a first threshold T 1 as vegetation areas, while other patches as non-vegetation areas;
S 2 : segmenting the non-vegetation areas by using a second scale parameter A 2 to obtain patches of a second scale, calculating a normalized difference buildup index NDBI of each patch, and judging the patches of the second scale with NDBI greater than a second threshold T 2 as building areas, while regarding other patches as non-building areas;
S 3 : segmenting the non-vegetation areas by using a third scale parameter A 3 to obtain patches of a third scale, and calculating a normalized difference build shadow index NSBI of each patch;
S 4 : calculating a normalized difference build shadow index separator S x (NSBI) of each patch according to the NSBI, and extracting the areas with a separator greater than a third threshold T 3 as landscape buildings,
wherein the normalized difference build shadow index separator Sx (NSBI) is calculated with the following formula:
S
x
(
NSBI
)
=
∑
x
i
∈
n
(
x
)
,
m
x
i
(
l
)
<
m
x
(
l
)
B
(
x
,
x
i
)
(
m
x
(
NSBI
)
-
m
x
i
(
NSBI
)
)
∑
x
i
∈
n
(
x
)
B
(
x
,
x
i
)
wherein, x represents the current calculated patch, n(x) represents a set of all patches adjacent to the current calculated patch, B(x, x i ) is the length of a common side of the current calculated patch x and the adjacent patch x i , and m x (NSBI) represents the NSBI value of the calculated patch x;
wherein the first scale parameter A 1 >the second scale parameter A 2 >the third scale parameter A 3 .
8. The device according to claim 7 , wherein the method uses the Multiresolution Segmentation algorithm or Hyper-pixel Segmentation algorithm to segment the image.
9. The device according to claim 7 , wherein the normalized difference vegetation index NDVI in the step S 1 is calculated with the following formula:
NDVI
=
NIR
-
R
NIR
+
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
10. The device according to claim 7 , wherein the normalized difference build shadow index NSBI in the step S 2 is calculated with the following formula:
NDBI
=
NIR
+
R
NIR
-
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
11. The device according to claim 7 , wherein the normalized difference build shadow index NSBI of each patch is calculated with the following formula in the step S 3 :
NSBI
=
NDBI
*
NIR
=
(
NIR
+
R
)
*
NIR
NIR
-
R
wherein, NIR is the mean valve of the near infrared band of the current calculated patch, and R is the mean valve of the red band of the current calculated patch.
12. The device according to claim 7 , wherein the first threshold T 1 , the second threshold T 2 and the third threshold T 3 are adaptively calculated with the OTSU method.