CN104376204A - Method for inverting vegetation coverage by adopting improved pixel dichotomy - Google Patents

Abstract

The invention relates to a method for retrieving vegetation coverage by using an improved pixel dichotomy method, which comprises the following steps: S1 calculates the NDVI of the entire area image; S2 determines the effective remote sensing image area in the entire area; S3 passes through the effective remote sensing image area The calculated NDVI unit values are counted, and the upper and lower limits of the confidence interval are obtained as NDVI _veg and NDVI _soil ; S4 uses the pixel dichotomy method to calculate the vegetation coverage. The invention adopts the improved pixel dichotomy method, which is most widely used because of its simplicity, less required parameters and higher precision. The present invention can eliminate the influence of the image background value when determining the pure vegetation NDVI _veg and pure soil NDVI _soil in the pixel dichotomy method, so that the pure vegetation NDVI and pure soil NDVI determined by the method of giving the upper limit and lower limit of the confidence interval are more reliable , so that the accuracy of vegetation coverage obtained by pixel dichotomy inversion is higher.

Description

A Method of Inverting Vegetation Coverage Using Improved Pixel Dichotomy Method

technical field

The invention relates to the field of earth system science, in particular to a method for retrieving vegetation coverage by an improved pixel dichotomy method.

Background technique

Vegetation is the most basic part of terrestrial ecosystems, and all other organisms depend on vegetation. Changes in the type, quantity and quality of vegetation have a profound impact on terrestrial ecosystems. Their roots go deep into the soil, and their branches and leaves touch the air. The unique transpiration and photosynthesis make soil, atmosphere, water and other physical and geographical elements interconnect and interact with each other, realizing the terrestrial ecosystem. matter-energy exchange and biochemical cycles. Vegetation coverage is an important parameter to describe the vegetation coverage on the ground, and it is also a basic and objective indicator indicating the change of the ecological environment. It occupies an important position in the atmosphere, pedosphere, hydrosphere and biosphere on the earth's surface. Vegetation coverage is an important biophysical parameter in the soil-vegetation-atmosphere transport model to simulate the exchange between the surface and the atmospheric boundary layer, and it needs to be accurately estimated in the numerical simulation of surface processes, climate change, and weather forecasting. From the application level, vegetation coverage has a wide range of applications in agriculture, forestry, resource and environment management, land use, hydrology, disaster risk monitoring, drought monitoring and other fields.

Vegetation coverage measurement methods usually include surface measurement and remote sensing measurement. Due to the significant spatio-temporal differentiation of vegetation coverage, although the surface measurement method based on discrete points may have high accuracy in local small areas, it has great uncertainty when extended to a large area; the remote sensing monitoring method is based on Spatial continuous data have certain advantages in estimating vegetation coverage in large and medium-scale areas, and are currently attracting attention.

The methods of extracting vegetation coverage from remote sensing data mainly include empirical model method, vegetation index method and mixed pixel decomposition method. The empirical model method is to use a single band, band combination or calculated vegetation index and measured vegetation coverage to establish a regression model, and then calculate the vegetation coverage of a larger area. The regression model is suitable for remote sensing images with a relatively recent time period. For images of earlier years, such as image data from ten or decades ago, due to changes in surface vegetation coverage, it is usually impossible to obtain the measured vegetation coverage of the sample area in the corresponding year Data, the application of this method is limited in time; the regression model has high accuracy in local areas, but has limitations in spatial application, and is only applicable to specific areas and specific vegetation types, and has no universal significance. The mixed pixel decomposition method is derived from the linear spectral mixed model of quantitative remote sensing. The basic idea is to decompose the pixels of remote sensing images into two parts: vegetation information and non-vegetation information, and estimate the proportion of vegetation information, that is, vegetation coverage. Among them, the pixel dichotomy method is the most widely used in the mixed pixel decomposition method. It does not need to establish a regression model and is less dependent on the surface measured data. Compared with the regression model method, it has more general significance and can be extended to a large range after verification. area. These advantages make the pixel dichotomy method one of the most widely used methods in remote sensing retrieval of vegetation coverage.

The pixel dichotomy assumes that a pixel is composed of vegetation and soil. The standard difference vegetation index (NDVI) of this pixel is the weighting of pure vegetation NDVI (NDVI _veg ) and pure soil NDVI (NDVI _soil ) according to the area. and. Then the vegetation coverage (F _c ) is expressed as:

F _c ＝(NDVI-NDVI _soil )/(NDVI _veg -NDVI _soil )

Some studies believe that if there are pure vegetation pixels and pure soil in the image, the maximum value of NDVI in the image is NDVI _veg , and the minimum value of NDVI is NDVI _soil . However, some scholars believe that the influence of the atmosphere and image noise should be removed, and a confidence interval is given for the value of NDVI, and the NDVI at a certain upper limit of the confidence interval is taken as NDVI _veg , and the NDVI at a certain lower limit of the confidence interval is taken as NDVI _soil . This method of determining pure vegetation and pure soil NDVI through confidence intervals is more scientific and reasonable. However, when determining the upper and lower limits of the confidence interval, as shown in Figure 1, the current method has certain errors in the statistically obtained NDVI _veg and NDVI _soil due to the existence of background values outside the effective area of the image, which affects the inversion of vegetation coverage precision and reliability. There is an urgent need for a method that can overcome the influence of background values in the process of obtaining NDVI _veg and NDVI _soil .

Contents of the invention

(1) Technical problems to be solved

The purpose of the present invention is to provide a method for retrieving vegetation coverage by pixel dichotomy that can eliminate background value region growth, so that the accuracy of vegetation coverage obtained by pixel dichotomy is higher.

(2) Technical solution

The present invention is achieved through the following technical solutions:

A method for retrieving vegetation coverage using an improved pixel dichotomy method, comprising the following steps:

S1 calculates the NDVI pixel value of the entire area image;

S2 determines the effective remote sensing image area in the entire area;

S3 counts the calculated NDVI pixel values in the effective remote sensing image area, and obtains the upper and lower limits of the confidence interval as NDVI _veg and NDVI _soil ;

S4 uses the pixel dichotomy method to calculate the vegetation coverage.

Wherein, the step S2 includes the following specific steps:

Starting from the selected initial growth point, the neighborhood growth method is adopted: if the value of each band of a pixel in the neighborhood of the initial point is 0, mark it as a new growth point, and use the new growth point Perform eight-neighborhood regrowth for the center until no new pixels are marked, and the boundary formed by the termination points is the boundary of the effective remote sensing image area. The technical effect of the technical solution is that the boundary of the effective remote sensing image area can be quickly and comprehensively determined.

Further, the neighborhood includes pixel positions in eight directions of up, down, left, right, upper left, lower left, upper right, and lower right centered on the growing point, so as to more comprehensively determine the boundary of the effective remote sensing image area.

Wherein, the selection method of the initial growth point is as follows:

Select the four corner points and the boundary points of the four sides of the remote sensing image as the pre-selected growth starting point, and then judge the four corner points and the boundary points of the four sides. If the value of each band of the point is 0, then the The point is marked as the growth start point, if it does not meet the conditions, it is not marked. The beneficial effect of the technical solution is that the growth starting point can be determined quickly and accurately.

Wherein, in the step S4, the vegetation coverage is calculated using the pixel dichotomy method,

The calculation formula of the pixel dichotomy is:

F _c ＝(NDVI-NDVI _soil )/(NDVI _veg -NDVI _soil );

F _c is the vegetation coverage, NDVI _veg is the NDVI value of pure vegetation; NDVI _soil is the NDVI value of pure soil; NDVI is the weighted sum of NDVI _veg and NDVI _soil according to the area, so as to accurately calculate the vegetation coverage.

Further, in the step S3, the following specific steps are included:

S31 arranges the NDVI pixel values in sequence;

S32 calculates the cumulative pixel number at each NDVI pixel value after sorting and accounts for the cumulative pixel percentage of the total number of pixels in the effective range of the image;

S33 If the cumulative pixel percentage at a certain NDVI value is compared with the value obtained by subtracting the confidence parameter from 1, the difference is the smallest, then the NDVI value is NDVI _soil ;

An NDVI value is NDVI _veg if the difference between the cumulative percentage of cells and the confidence parameter is the smallest at that NDVI value. The technical effect of the technical solution is that the values of NDVI _soil and NDVI _veg can be accurately and quickly determined.

Preferably, in the step S33, the value range of the confidence parameter is 1%-5%, so as to effectively avoid the NDVI being too low or too high due to the influence of noise.

Among them, the calculation method of the NDVI pixel value of the whole area image is:

NDVI=(p(nir)-p(red))/(p(nir)+p(red))

Where p(nir) represents the spectral reflectance or gray value of the near-infrared band, and p(red) represents the spectral reflectance or gray value of the visible red band. The technical effect of the technical solution is that the image NDVI pixel value of the entire area can be accurately measured through the calculation method of the NDVI pixel value.

(3) Beneficial effects

Compared with prior art and product, the present invention has following advantage:

The invention adopts the improved pixel dichotomy method, which is most widely used because of its simplicity, less required parameters and higher precision. The present invention can eliminate the influence of the image background value when determining the pure vegetation NDVI _veg and pure soil NDVI _soil in the pixel dichotomy method, so that the pure vegetation NDVI and pure soil NDVI determined by the method of giving the upper limit and lower limit of the confidence interval are more reliable , so that the accuracy of vegetation coverage obtained by pixel dichotomy inversion is higher.

Description of drawings

Fig. 1 is the NDVI image figure that considers background value in the background technology;

Fig. 2 does not consider the NDVI image figure of background value provided by the present invention;

Fig. 3 is a schematic diagram of the steps of the region growing algorithm provided by the present invention;

Fig. 4 is the calculation method step schematic diagram of pure vegetation NDVI _veg and pure soil NDVI _soil in the pixel dichotomy method provided by the present invention;

Fig. 5 is a schematic diagram of the steps of the method for retrieving vegetation coverage by using the improved pixel dichotomy method provided by the present invention.

Detailed ways

In order to make it easier for those skilled in the art to understand and implement the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

As shown in Figure 5, the present embodiment provides a method for retrieving vegetation coverage using an improved pixel dichotomy method, including the following steps:

S1 calculates the NDVI of the entire image (including invalid areas) through the NDVI calculation formula. The NDVI is used to determine the NDVI sorting in step S3, determine NDVI _soil and NDVI _veg , and the NDVI calculation formula is as follows:

NDVI=(p(nir)-p(red))/(p(nir)+p(red))

Where p(nir) represents the spectral reflectance or gray value of the near-infrared band, and p(red) represents the spectral reflectance or gray value of the visible red band.

S2 determines the effective remote sensing image area in the whole area.

The specific method for determining the effective remote sensing affected area is shown in Figure 3. Select the four corner points and the boundary points of the four sides of the remote sensing image as the pre-selected growth starting point, and then distinguish the four corner points and the boundary points of the four sides , if the value of each band of this point is 0, then mark this point as the growth starting point, if it does not meet the conditions, it will not be marked, so that the boundary of the effective remote sensing image area can be quickly and comprehensively determined.

Take the selected growth starting point as the starting point to perform eight-neighborhood growth (eight-neighborhood refers to the pixel positions in the eight directions of up, down, left, right, left, down, right, and down right centered on the pixel). If the starting point is eight If the value of each band of a certain pixel in the neighborhood is 0, it is marked as a new growth point, and the eight-neighborhood re-growth is performed with this point as the center until no new pixels are marked. The boundary formed by the termination points is the effective remote sensing image boundary with practical significance. The process of growing the whole region is shown in Figure 3.

S3 obtains the NDVI value of the effective remote sensing image area through calculation and statistics, and obtains the upper and lower limits of the confidence interval as NDVI _veg and NDVI _soil .

In step S3, the specific steps are as shown in Figure 4:

S31 arranges the NDVI unit values in sequence;

S32 calculates the cumulative pixel percentage of the cumulative pixel number at each NDVI value after sorting to the total number of pixels in the effective range of the image;

S33 If the difference between the cumulative pixel percentage at a certain NDVI value and (1-confidence parameter) is the smallest, then the NDVI value is NDVI _soil ;

An NDVI value is NDVI _veg if the difference between the cumulative percentage of cells and the confidence parameter is the smallest at that NDVI value.

S4 uses the pixel dichotomy method to calculate the vegetation coverage. Through the specific method of the above step S3, the values of NDVI _soil and NDVI _veg can be determined accurately and quickly.

Wherein, the calculation formula of the pixel dichotomy is:

F _o ＝(NDVI-NDVI _soil )/(NDVI _veg -NDVI _soil );

F _c is vegetation coverage, NDVI _veg is NDVI of pure vegetation; NDVI _soil is NDVI of pure soil; NDVI is the weighted sum of NDVI _veg of pure vegetation and NDVI _soil of pure soil according to the area.

Vegetation coverage is a basic and objective indicator to indicate changes in the ecological environment. It is an important biophysical parameter in the soil-vegetation-atmosphere transfer model to simulate the exchange of the surface and atmospheric boundary layers. It is also used in agriculture, forestry, resource and environment management, land use, Hydrology, disaster risk monitoring, drought monitoring and other fields have a wide range of applications. As one of the remote sensing methods for retrieving vegetation coverage, the pixel dichotomy method is the most widely used because of its simplicity, fewer parameters required, and higher accuracy.

As shown in Figure 2, the method of inverting vegetation coverage by using the improved pixel dichotomy method in this embodiment can eliminate the influence of the image background value when determining the pure vegetation NDVI _veg and pure soil NDVI _soil in the pixel dichotomy method, It makes the pure vegetation NDVI _veg and pure soil NDVI _soil determined by the method of giving the upper and lower limits of the confidence interval more reliable, so that the accuracy of the vegetation coverage obtained by the pixel dichotomy inversion is higher.

According to the size of a test area and the resolution of the image, in order to avoid too low or too high a value of NDVI caused by noise, it is preferable to establish a confidence parameter value ranging from 1% to 5%. Considering the NDVI image of the background value as shown in Figure 1, the NDVI _veg and NDVI _soil selected according to the traditional method in Table 1 can be obtained. As shown in Figure 2, the NDVI image without considering the background value can be obtained from the NDVI _veg and NDVI _soil selected according to the region growing algorithm in Table 2. There is still a clear gap between the two, which has an obvious impact on the inversion of vegetation coverage.

Table 1 NDVI _veg and NDVI _soil selected according to the traditional method

Table 2 shows the NDVI _veg and NDVI _soil selected according to the region growing algorithm

The above example is only one implementation of the present invention, and its description is more specific and detailed, but it should not be construed as limiting the patent scope of the present invention. Its specific structure and size can be adjusted accordingly according to actual needs. It should be pointed out that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all belong to the protection scope of the present invention.

Claims (8)

1. a kind of method adopting improved pixel dichotomy inversion vegetation coverage, is characterized in that, comprises the following steps: S1 calculates the NDVI pixel value of the entire area image; S2 determines the effective remote sensing image area in the entire area; S3 counts the calculated NDVI pixel values in the effective remote sensing image area, and obtains the upper and lower limits of the confidence interval as NDVI _veg and NDVI _soil ; S4 uses the pixel dichotomy method to calculate the vegetation coverage. 2. the method for adopting improved pixel dichotomy to invert vegetation coverage according to claim 1, is characterized in that, described step S2 comprises the following concrete steps: Starting from the selected initial growth point, the neighborhood growth method is adopted: if the value of each band of a pixel in the neighborhood of the initial point is 0, mark it as a new growth point, and use the new growth point Perform eight-neighborhood regrowth for the center until no new pixels are marked, and the boundary formed by the termination points is the boundary of the effective remote sensing image area. 3. the method according to claim 2 adopting improved pixel dichotomy inversion vegetation coverage, it is characterized in that, described neighborhood comprises growth point as the center up and down, left and right, upper left, lower left, upper right, lower right eight pixel position in one direction. 4. the method according to claim 2 adopting improved pixel dichotomy inversion vegetation coverage, it is characterized in that, the selection method of described initial growth point is as follows: Select the four corner points and the boundary points of the four sides of the remote sensing image as the pre-selected growth starting point, and then judge the four corner points and the boundary points of the four sides. If the value of each band of the point is 0, then the The point is marked as the growth start point, if it does not meet the conditions, it is not marked. 5. the method for adopting the improved pixel dichotomy method to invert the vegetation coverage according to claim 1, is characterized in that, in the step S4, utilizes the pixel dichotomy to calculate the vegetation coverage, The calculation formula of the pixel dichotomy is: F _c ＝(NDVI-NDVI _soil )/(NDVI _veg -NDVI _soil ); F _c is the vegetation coverage, NDVI _veg is the NDVI value of pure vegetation; NDVI _soil is the NDVI value of pure soil; NDVI is the weighted sum of NDVI _veg and NDVI _soil according to the area. 6. the method for adopting improved pixel dichotomy method inversion of vegetation coverage according to claim 1, is characterized in that, in described step S3, comprises the following concrete steps: S31 arranges the NDVI pixel values in sequence; S32 calculates the cumulative pixel number at each NDVI pixel value after sorting and accounts for the cumulative pixel percentage of the total number of pixels in the effective range of the image; S33 If the cumulative pixel percentage at a certain NDVI value is compared with the value obtained by subtracting the confidence parameter from 1, the difference is the smallest, then the NDVI value is NDVI _soil ; An NDVI value is NDVI _veg if the difference between the cumulative percentage of cells and the confidence parameter is the smallest at that NDVI value. 7. The method for retrieving vegetation coverage by the improved pixel dichotomy method according to claim 6, characterized in that, in the step S33, the value range of the confidence parameter is 1%-5%. 8. according to the method described in any one of claim 1～7 adopting improved pixel dichotomy to invert vegetation coverage, it is characterized in that, the computing method of the NDVI pixel value of whole area image is: NDVI=(p(nir)-p(red))/(p(nir)+p(red)) Where p(nir) represents the spectral reflectance or gray value of the near-infrared band, and p(red) represents the spectral reflectance or gray value of the visible red band.