Method and apparatus for extracting mountain landscape buildings based on high-resolution remote sensing images

Abstract

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.

Claims

1. A method for extracting mountain landscapes buildings based on high-resolution remote sensing images, comprising the following steps: S1: segmenting a remote sensing image into patches of a first scale by using a first scale parameter A1, calculating a normalized difference vegetation index NDVI of each patch, and extracting the segmented patches with NDVI greater than a first threshold T1 as vegetation areas, while other patches as non-vegetation areas; S2: segmenting the non-vegetation areas by using a second scale parameter A2 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 T2 as building areas, while regarding other patches as non-building areas; S3: segmenting the non-vegetation areas by using a third scale parameter A3 to obtain patches of a third scale, and calculating a normalized difference build shadow index NSBI of each patch; S4: calculating a normalized difference build shadow index separator S.sub.x(NSBI) of each patch according to the NSBI, and extracting the areas with a separator greater than a third threshold T3 as landscape buildings, wherein the normalized difference build shadow index separator Sx (NSBI) is calculated with the following formula: $S_x\left(\mathrm{NSBI}\right)=\frac{\underset{x_i\in n\left(x\right),m_{x_i}\left(l\right)<m_x\left(l\right)}{.Math.}B\left(x,x_i\right)\left(m_x\left(\mathrm{NSBI}\right)-m_{x_i}\left(\mathrm{NSBI}\right)\right)}{\underset{x_i\in n\left(x\right)}{.Math.}B\left(x,x_i\right)}$ wherein, x represents the current calculated patch, n(x) represents a set of all patches adjacent to the current calculated patch, B(x, x.sub.i) is the length of a common side of the current calculated patch x and the adjacent patch x.sub.i, and m.sub.x(NSBI) represents the NSBI value of the calculated patch x; wherein the first scale parameter A1>the second scale parameter A2>the third scale parameter A3.

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 S1 is calculated with the following formula: $\mathrm{NDVI}=\frac{\mathrm{NIR}-R}{\mathrm{NIR}+R}$ 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 S2 is calculated with the following formula: $\mathrm{NDBI}=\frac{\mathrm{NIR}+R}{\mathrm{NIR}-R}$ 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 S3: $\mathrm{NSBI}=\mathrm{NDBI}*\mathrm{NIR}=\frac{\left(\mathrm{NIR}+R\right)*\mathrm{NIR}}{\mathrm{NIR}-R}$ 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 T1, the second threshold T2 and the third threshold T3 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: S1: segmenting a remote sensing image into patches of a first scale by using a first scale parameter A1, calculating a normalized difference vegetation index NDVI of each patch, and extracting the segmented patches with NDVI greater than a first threshold T1 as vegetation areas, while other patches as non-vegetation areas; S2: segmenting the non-vegetation areas by using a second scale parameter A2 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 T2 as building areas, while regarding other patches as non-building areas; S3: segmenting the non-vegetation areas by using a third scale parameter A3 to obtain patches of a third scale, and calculating a normalized difference build shadow index NSBI of each patch; S4: calculating a normalized difference build shadow index separator S.sub.x(NSBI) of each patch according to the NSBI, and extracting the areas with a separator greater than a third threshold T3 as landscape buildings, wherein the normalized difference build shadow index separator Sx (NSBI) is calculated with the following formula: $S_x\left(\mathrm{NSBI}\right)=\frac{\underset{x_i\in n\left(x\right),m_{x_i}\left(l\right)<m_x\left(l\right)}{.Math.}B\left(x,x_i\right)\left(m_x\left(\mathrm{NSBI}\right)-m_{x_i}\left(\mathrm{NSBI}\right)\right)}{\underset{x_i\in n\left(x\right)}{.Math.}B\left(x,x_i\right)}$ wherein, x represents the current calculated patch, n(x) represents a set of all patches adjacent to the current calculated patch, B(x, x.sub.i) is the length of a common side of the current calculated patch x and the adjacent patch x.sub.i, and m.sub.x(NSBI) represents the NSBI value of the calculated patch x; wherein the first scale parameter A1>the second scale parameter A2>the third scale parameter A3.

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 S1 is calculated with the following formula: $\mathrm{NDVI}=\frac{\mathrm{NIR}-R}{\mathrm{NIR}+R}$ 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 S2 is calculated with the following formula: $\mathrm{NDBI}=\frac{\mathrm{NIR}+R}{\mathrm{NIR}-R}$ 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 S3: $\mathrm{NSBI}=\mathrm{NDBI}*\mathrm{NIR}=\frac{\left(\mathrm{NIR}+R\right)*\mathrm{NIR}}{\mathrm{NIR}-R}$ 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 T1, the second threshold T2 and the third threshold T3 are adaptively calculated with the OTSU method.

Description

DESCRIPTION OF DRAWINGS

(1) FIG. 1 is a flow chart of the method for extracting mountain landscape buildings based on high-resolution remote sensing images according to the present invention;

(2) FIG. 2 is a remote sensing image (partial) of Jizu Mountain Scenic Area at a segmentation scale of 100 according to an embodiment of the present invention;

(3) FIG. 3 is a NDVI analysis view of Jizu Mountain Scenic Area and surrounding areas according to an embodiment of the present invention;

(4) FIG. 4 is an analysis view of extraction with an OTSU algorithm with a first threshold according to an embodiment of the present invention;

(5) FIG. 5 is a NDBI analysis view of Jizu Mountain Scenic Area and surrounding areas according to an embodiment of the present invention;

(6) FIG. 6 is a remote sensing image (partial) of the Scenic Area at a segmentation scale of 20 according to an embodiment of the present invention;

(7) FIG. 7 is a remote sensing image (partial) of the Scenic Area at a segmentation scale of 10 according to an embodiment of the present invention;

(8) FIG. 8 is a NSBI analysis view of the landscape buildings and their surroundings according to an embodiment of the present invention;

(9) FIG. 9 is a NSBI characteristic value analysis view of the landscape buildings and their surroundings according to an embodiment of the present invention;

(10) FIG. 10 shows the final result view of extraction of buildings in Jizu Mountain Scenic Area and surrounding areas according to an embodiment of the present invention.

DETAILED DESCRIPTION

(11) Hereunder the technical solution of the present invention will be further detailed with reference to the drawings. It should be understood that the following embodiments are only for the purpose of disclosing the present invention in detail and completely, and fully conveying the technical concept of the present invention to the person skilled in the art. The present invention may also be implemented in many different forms and is not limited to the embodiments described here. Terms used in the exemplary embodiments shown in the drawings are not intended to limit the present invention.

(12) In view of the problems that most of the buildings in scenic areas are located in mountainous areas and woodlands, the spectral characteristic values of the sunny side and shady side of mountainous areas are greatly different owing to shadow difference, and the landscape buildings are small in size and may be shaded easily, and the existing technical means have high cost, low efficiency, poor performance, cumbersome process and low accuracy, the present invention provides a method for extracting mountain landscape buildings based on high-resolution remote sensing images, which utilizes the chlorophyll concentration difference between landscape buildings and surrounding vegetation in a scenic area and the brightness difference of building shadows, combines a normalized difference buildup index (NDBI) related with chlorophyll in remote sensing interpretation, a normalized difference shadow index (NDSI) related with brightness, a segmentation-based image interpretation idea and a human vision saliency mechanism, decreases the influence of large shadow areas in mountainous shady areas on the extraction result, puts forth a normalized difference build shadow index (NSBI) separator, obtains a threshold of the NSBI separator with reference to the NSBI, and thereby extracts landscape buildings in the scenic area by means of threshold segmentation of this feature.

(13) With reference to FIG. 1, the method for extracting mountain landscape buildings based on high-resolution remote sensing images comprises the following steps:

(14) step S1: segmenting a remote sensing image into first scale patches by using a first scale parameter A1, calculating a normalized difference vegetation index (NDVI) of each patch, and extracting non-vegetation areas according to the NDVI;

(15) In an embodiment, the remote sensing images of Jizu Mountain Scenic Area and the surrounding areas in Dali, Yunnan, China are processed. The area is a typical scenic area in scattered distribution. The raw high-resolution remote sensing images obtained are in a resolution of 3.2 meters and image size of 1,419 rows x 823 columns, and contain four bands, namely a blue band (0.45 to 0.52 μm), a green band (0.52 to 0.59 μm), a red band (0.63 to 0.69 μm) and a near infrared band (0.77 to 0.89 μm). The objects to be extracted are landscape buildings.

(16) The image segmentation algorithm may use algorithms such as Multiresolution Segmentation algorithm or simple linear iterative clustering (SLIC) algorithm. The first scale parameter A1 employs a larger scale, and the value range is preferably within 80 to 150. In this embodiment, the Multiresolution Segmentation algorithm is used, and the scale parameter A1 is set to 100, as shown in FIG. 2.

(17) After image segmentation, a normalized difference vegetation index NDVI of each patch is calculated with the following formula:

(18) $\mathrm{NDVI}=\frac{\mathrm{NIR}-R}{\mathrm{NIR}+R}$

(19) 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. −1<=NDVI<=1, wherein a negative value indicates that the ground is covered by cloud, water, snow, etc., and is highly reflective to visible light; value 0 indicates that there are rocks or bare soil, etc., and NIR and R are approximately equal to each other; a positive value indicates that there is vegetation coverage, which increases with the increasing of coverage. Through threshold segmentation based on the near infrared band and the NDVI, vegetation areas and non-vegetation areas can be separated effectively. The obtained NDVI analysis view is shown in FIG. 3.

(20) Segmented patches with NDVI greater than a first threshold T1 are extracted as vegetation areas, while other patches as non-vegetation areas. The setting criterion for the first threshold T1 is that the vegetation areas and the non-vegetation areas can be distinguished in the current image, and the threshold is obtained by using a maximum interclass variance method (OTSU method) determined with an adaptive threshold. The calculated threshold is shown in FIG. 4, in which the black areas are vegetation areas, the white areas are non-vegetation areas, and the boundary lines between them are determined by the threshold. The threshold T1=0.1021 is obtained, and areas with NDVI greater than 0.1021 are set as vegetation areas, while areas with NDVI smaller than 0.1021 are set as non-vegetation areas, as shown in FIG. 5.

(21) step S2: segmenting the non-vegetation areas by using a second scale parameter A2 to obtain patches of a second scale, calculating a normalized difference buildup index NDBI of each patch, and extracting building areas according to the NDBI;

(22) Specifically, with reference to the image segmentation method in the step S1, the non-vegetation areas are segmented again by using a smaller scale parameter A2. The setting criterion for the second scale parameter A2 is that the segmented patches can be as large as possible without confusing the landscape building areas and the vegetation areas. The second scale parameter A2 is similar to the first scale parameter A1; it is possible to further divide A1 on the basis of A1 to divide the internal elements of A1 more clearly with a smaller scale. The segmentation threshold is interpreted for general areas with distinctive features in this case based on a human vision saliency mechanism. For example, the site is segmented and then observed, the areas with most segmentation elements in the site, i.e., the areas with all basic elements in the site in a patch, are taken as general areas, and a segmentation threshold range that has good segmentation separability for plants, buildings and water bodies is extracted, and the segmentation scale A2 in a range of 20 to 50, which has high adaptability to this step, is obtained. In this embodiment, the second scale parameter A2 is set to 20, as shown in FIG. 6.

(23) For the patches after the second segmentation, the normalized difference buildup index NDBI of each patch in the non-vegetation areas is calculated. The calculation formula is as follows:

(24) $\mathrm{NDBI}=\frac{\mathrm{NIR}+R}{\mathrm{NIR}-R}$
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. By calculating the NDBI, the approximate locations of landscape buildings in the scenic area can be delineated. The obtained NDBI analysis map is shown in FIG. 5.

(25) Patches of the second scale with NDBI greater than a second threshold T2 are identified as building areas, while other patches as non-building areas. The second threshold T2 is similar to the first threshold T1. Compared with the vegetation areas and non-vegetation areas distinguished by T1, the value of T2 focuses on the threshold of separator between buildings and non-buildings in the non-vegetation areas. The value of separator between buildings and non-buildings is divided on a smaller scale with human vision saliency mechanism and OTSU method (see the step of T1 for the details). In this embodiment, T2=−0.4021 is obtained.

(26) step S3: segmenting the non-vegetation areas again by using a third scale parameter A3 to obtain patches of a third scale, and calculating a normalized difference build shadow index NSBI of each patch;

(27) Non-vegetation areas refer to the mixed areas of buildings and surrounding vegetations and shadows extracted in the step 2. Specifically, with reference to the image segmentation method in the step S1, in conjunction with the human vision saliency mechanism, the segmentation threshold is interpreted for general areas with distinguishing features in this case, and a segmentation scale A3 in a range of 5 to 10, which has high adaptability to this step, is obtained. The non-vegetation areas are segmented again by using a smaller scale parameter A3. The setting criterion for the third scale parameter A3 is that the segmented patches can be as large as possible without confusing the landscape building areas and the shadow areas. In this embodiment, A3 is set to 10, as shown in FIG. 7.

(28) The normalized difference build shadow index NSBI of each patch in the non-vegetation areas is calculated. For the non-vegetation areas, the influence of building areas and shadows and other interference items in the non-vegetation areas can be reduced. The calculation formula is as follows:

(29) $\mathrm{NSBI}=\mathrm{NDBI}*\mathrm{NIR}=\frac{\left(\mathrm{NIR}+R\right)*\mathrm{NIR}}{\mathrm{NIR}-R}$
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. 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. Based the normalized difference buildup index NDBI, NSBI introduces a near infrared spectral band (NIR band) that has a significant response to shadow changes, so as to improve the difference in NDBI between shadows and buildings in non-vegetation areas. The obtained NSBI analysis view is shown in FIG. 8.

(30) step S4: calculating a NSBI separator S.sub.x(NSBI) of each patch according to the NSBI, and judging and extracting landscape buildings on the basis of the separator.

(31) The NSBI separator S.sub.x(NSBI) of each patch in the non-vegetation areas is calculated with the following formula:

(32) $S_x\left(\mathrm{NSBI}\right)=\frac{\underset{x_i\in n\left(x\right),m_{x_i}\left(l\right)<m_x\left(l\right)}{.Math.}B\left(x,x_i\right)\left(m_x\left(\mathrm{NSBI}\right)-m_{x_i}\left(\mathrm{NSBI}\right)\right)}{\underset{x_i\in n\left(x\right)}{.Math.}B\left(x,x_i\right)}$
where, x represents the current calculated patch, n(x) represents a set of all patches adjacent to the current calculated patch, B(x, x.sub.i) is the length of a common side of the current calculated patch x and the adjacent patch x.sub.i, and m.sub.x(NSBI) represents the NSBI value of the calculated patch x.

(33) The S.sub.x(NSBI) separator enhances the NSBI relative difference between landscape patches and surrounding patches after segmentation, enhances the spectral relative difference between landscape buildings in bright areas, landscape buildings in shadow areas and shadow areas, and decreases the probability of spectral confusion, so that landscape buildings and adjacent patches can be distinguished significantly. FIG. 9 shows an analysis view of minimum extraction based on mixed areas of buildings and surrounding shadow areas.

(34) Then, landscape building areas are extracted on the basis of the calculated NSBI separator S.sub.x(NSBI). Specifically, with reference to the image threshold interpretation method in the step S1, in conjunction with the human vision saliency mechanism, the extraction threshold is interpreted for general areas with distinguishing features in this case, an extraction threshold T3 which has high adaptability to this step is obtained. The scale parameter is used to obtain the smallest unit of analysis (as shown in FIGS. 2, 6 and 7), and the threshold here is used to extract boundary values of the NSBI separator S.sub.x(NSBI) analyzed in the smallest unit. In this embodiment, a separating threshold value T3=4.5732 between the landscape buildings and their adjacent shadows areas and the vegetation areas is obtained. The final extraction result is shown in FIG. 10. It can be seen that the present invention has strong extraction ability for landscape building settlements in scattered distribution.

(35) Based on the same technical concept as the method embodiment, according to another embodiment of the present invention, a computer device is provided, which comprises: 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 the one or more processors, and when the program is executed by the processors, the steps in the method embodiment of the present invention is implemented.

(36) The person skilled in the art should understand that the embodiments of the present application may be provided as a method, a system, or a computer program product. Therefore, the present application may be in the form of a pure hardware embodiment, a pure software embodiment, or a embodiment that combines the aspects of software and hardware. Furthermore, the present application may be in the form of a computer program product implemented on one or more computer-readable storage media that contain computer-readable program codes (including but not limited to disk storage, CD-ROM, optical storage, etc.).

(37) The present application is described with reference to flow charts and/or block charts of the method, device (system) and computer program product according to the embodiments of the present application. It should be understood that each flow and/or block in the flow charts and/or block charts and combinations of the flows and/or blocks in the flow charts and/or block charts can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processing device or any other programmable data processing device to produce a machine, so that the instructions, which are executed by the processor of the computer or other programmable data processing device, produce a device for implementing the functions specified by one or more flows in the flow charts and/or one or more blocks in the block charts.

(38) These computer program instructions may also be stored in a computer-readable memory unit that can direct the computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory unit produce a product including an instruction device which implements the functions specified by one or more flows in the flow charts and/or one or more blocks in the block charts.

(39) These computer program instructions may also be loaded onto a computer or other programmable data processing device, so that a series of operational steps are performed on the computer or other programmable data processing device to produce a computer-implemented process, thus the instructions executed on the computer or other programmable data processing device provide steps for implementing the functions specified by one or more flows in the flow charts and/or one or more blocks in the block charts.

(40) Finally, it should be noted that the above embodiments are only provided to explain the technical solution of the present invention, but do not constitute any limitation to the present invention. Although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that various modifications or equivalent replacements may be made to the embodiments of the present invention without departing from the spirit and scope of the present invention, and all of such modifications or equivalent replacements shall be deemed as falling in the protection scope as defined by the claims of the present invention.