A canopy coverage considering aggregation index inversion method and correction method

Abstract

The application discloses a canopy coverage considering aggregation index inversion method and correction method, relates to the technical field of remote sensing image processing and aggregation index inversion, collects remote sensing images of a vegetation canopy, extracts hot spot reflectivity and dark spot reflectivity respectively, calculates a normalized hot spot and dark spot index NDHD, calculates an initial value of an aggregation index CI according to a linear relationship between the NDHD and the CI, and corrects the initial value of the CI by using a canopy coverage Ccan, wherein an error between the initial value of the CI and a real value of the CI has a linear relationship with the Ccan. The application first establishes a CI correction model based on the Ccan, and solves the deviation problem of a traditional algorithm in a sample field with high or low Ccan. By improving a mainstream CI inversion algorithm based on the NDHD, the application significantly improves the precision of CI inversion, and provides more reliable technical support for forest ecological monitoring, carbon cycle research and remote sensing application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing image processing and clumping index inversion, and particularly to a clumping index inversion method and a correction method considering canopy coverage. BACKGROUND

[0002] Forests play an important role in carbon cycle, and their leaves can absorb carbon dioxide and release oxygen through photosynthesis, which plays a crucial role in maintaining ecological balance, protecting biodiversity, regulating climate and water resources, and protecting soil. Therefore, the research on forests is one of the hotspots of global carbon cycle research.

[0003] Leaf area index (LAI) is defined as half of the total leaf area per unit of ground surface area, and is a key parameter for characterizing structure and ecological state. Research on LAI is an important part of research on photosynthesis, transpiration, rain interception, and forest leaf litter, and it quantifies the leaf area per unit of ground surface area.

[0004] Clumping index (CI) is an important canopy structure parameter, which is used to describe the degree of random distribution of leaves in the canopy, and is an important indicator for describing the radiation transmission process in the canopy. It is a necessary parameter for estimating the true leaf area index in the canopy and distinguishing between "sunlit leaves" and "shaded leaves" in the canopy, and is of great significance in the study of global carbon and water cycles. CI is defined as the ratio of effective leaf area index (LAIe) to true leaf area index. When CI<1, it indicates that the leaves in the canopy are clumped, and the smaller the CI value, the higher the degree of leaf clumping in the canopy; when CI=1, it indicates that the leaves in the canopy are randomly distributed; when CI>1, it indicates that the leaves in the canopy are uniformly distributed, and the larger the CI, the more uniform the distribution of leaves in the canopy. In recent years, significant progress has been made in CI inversion algorithm research, among which the method based on normalized difference between hotspot and darkspot (NDHD) proposed by Chen et al. is the most widely used (current mainstream algorithm). However, the existing mainstream algorithm based on NDHD for CI inversion has two problems due to the assumption that tree crowns are randomly distributed: (1) random distribution of tree crowns is a relatively ideal distribution, which is not common in nature; (2) CI true value is independent of tree crown number, and further derivation shows that CI true value is not closely related to canopy cover (Ccan), while in actual situation CI true value is closely related to tree crown number and Ccan. SUMMARY

[0005] In order to overcome the defects in the prior art, the present application provides a canopy coverage considering aggregation index inversion method and correction method, and first establishes a CI correction model based on Ccan, solving the deviation problem of traditional algorithms in sample plots with high or low Ccan.

[0006] To achieve the above object, the present application adopts the following technical scheme, comprising:

[0007] A canopy coverage considering aggregation index inversion method,

[0008] Collecting remote sensing images of vegetation canopy, extracting hot spot reflectivity and dark spot reflectivity respectively;

[0009] Calculating normalized hot spot and dark spot index NDHD, and calculating CI initial value according to the linear relationship between NDHD and aggregation index CI;

[0010] Using canopy coverage Ccan to correct the CI initial value, wherein the error between the CI initial value and the true value of CI and Ccan is linearly related.

[0011] Preferably, the correction method is:

[0012] ;

[0013] Wherein, CI new CI correction value; CI old CI initial value; d Slope; F Intercept.

[0014] Preferably, the slope of the infrared band d Take -1.1, the slope of the near-infrared band d Take -1; the intercept of the infrared band F The intercept of the near-infrared band F 0.23 larger.

[0015] Preferably, ; wherein, e Constant, e Take 0.0004; l Sample plot scale; f Constant, infrared band f Take 0.73, near-infrared band f Take 0.5.

[0016] Preferably, the calculation method of CI initial value is:

[0017] ;

[0018] ;

[0019] wherein, h is the hotspot reflectance, d is the darkspot reflectance; a and b is a coefficient determined by linear regression; the infrared band takes a -1.2, b 1.28; the near-infrared band takes a -1.34, b 1.20.

[0020] Preferably, the canopy coverage Ccan is calculated based on the pixel two-part model:

[0021] ;

[0022] wherein, NDVI is the normalized vegetation index of the pixel, NDVI veg is the NDVI value of the pure vegetation pixel, NDVI soil is the NDVI value of the pure non-vegetation pixel.

[0023] The present application also provides a correction method for the aggregation index considering the canopy coverage,

[0024] ;

[0025] wherein, CI new is the corrected CI correction value, CI old is the CI initial value before correction; d is the slope; e is the constant; l is the sample plot scale; f is the constant.

[0026] Preferably, the CI initial value is specifically: the CI inversion value obtained based on the normalized hotspot and darkspot index method; when CI is inverted through the infrared band d -1.1 is taken, and when CI is inverted through the near-infrared band d -1 is taken; e 0.0004 is taken; when CI is inverted through the infrared band f 0.73 is taken, and when CI is inverted through the near-infrared band f 0.5 is taken.

[0027] The application further provides a computer program product comprising computer programs / instructions which, when executed by a processor, implement the canopy coverage considering aggregation index inversion method.

[0028] The application further provides a computer program product comprising computer programs / instructions which, when executed by a processor, implement the canopy coverage considering aggregation index correction method.

[0029] The application has the following advantages:

[0030] (1) The application first establishes a CI correction model based on Ccan, solving the deviation problem of traditional algorithms in sample plots with high or low Ccan.

[0031] (2) The application proposes a band-specific parameter optimization strategy, and optimal correction coefficients are designed for infrared and near-infrared bands respectively, realizing the collaborative inversion of multispectral data. The application introduces a dynamic scale adaptation mechanism, so that the model can match the needs of different spatial resolutions, effectively ensuring the consistency of inversion from local sample plots to regional scale.

[0032] (3) The CI correction model based on Ccan of the application provides a solution for accurate estimation of vegetation canopy structure parameters. By improving the mainstream NDHD-based CI inversion algorithm, the application significantly improves the accuracy of CI inversion, providing more reliable technical support for forest ecological monitoring, carbon cycle research and remote sensing applications.

[0033] (4) The technical achievements of the application have wide application prospects: in the field of ecological environment monitoring, the improved CI inversion model can provide more accurate data support for forest growth assessment and biodiversity protection; in global carbon cycle research, high-precision canopy structure parameters help to improve the simulation accuracy of vegetation productivity models; in the business application of remote sensing, the simple design and low input requirements of the model greatly reduce the technical implementation threshold, which is conducive to large-scale popularization and use. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 Error scatter plot of Ccan and near-infrared band CI inversion.

[0035] Figure 2 Error scatter plot of Ccan and infrared band CI inversion.

[0036] Figure 3 Comparison chart of real CI value and near-infrared band CI value at 500m scale.

[0037] Figure 4The contrast chart of CI value for 500m scale and infrared band inversion CI value.

[0038] Figure 5 The correction effect chart of CI inversion for three scales (100m, 500m, 1000m) of infrared band.

[0039] Figure 6 The correction effect chart of CI inversion for three scales (100m, 500m, 1000m) of near infrared band.

[0040] Figure 7 The flow chart of the aggregation index inversion method considering canopy coverage in the embodiment. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0042] Embodiment 1

[0043] The aggregation index inversion method considering canopy coverage of the present application comprises the following contents.

[0044] I. Construction of the aggregation index (CI) correction model based on canopy coverage (Ccan)

[0045] (1) Introduction of the current mainstream algorithm

[0046] The current mainstream method for remotely inverting CI is to invert CI based on the linear relationship between normalized hot spot and dark spot index (NDHD) and CI:

[0047] ;

[0048] ;

[0049] wherein, h hot spot reflectivity, d dark spot reflectivity; a and b are coefficients determined through linear regression; when CI is inverted through infrared band, take a -1.2, b 1.28; when CI is inverted through near infrared band, take a -1.34, b 1.20.

[0050] (2) Deficiency of the mainstream algorithm

[0051] First, the random distribution of tree crowns, which is considered by the existing mainstream algorithm, is an ideal distribution that is not common in nature. Second, even if the tree crowns in the real scene satisfy the random distribution, another deficiency of the algorithm can be proved by the following derivation, as follows.

[0052] Define CI as the effective leaf area index LAI e The ratio of the true leaf area index LAI

[0053] ;

[0054] LAI The total area of the plant leaves per unit ground area, LAI e The apparent total leaf area of the vegetation canopy per unit ground area, which is calculated by the optical gap rate under the assumption of random distribution of leaves.

[0055] Among them, LAI The true leaf area index can be expressed as:

[0056] ;

[0057] Among them, The leaf area of a single tree of the sample plot representative crown, n The number of crowns in the sample plot, S The area of the sample plot.

[0058] Under the assumption that the leaves in the continuous canopy are randomly distributed in space, the canopy gap rate can be calculated as follows:

[0059] ;

[0060] Among them, The observation zenith angle, Indicates the projection coefficient of the leaves in the vertical plane of the observation direction, which is related to the leaf inclination angle.

[0061] However, the assumption of random distribution of leaves is often incorrect in reality, especially for forest scenes, where leaves will gather in each tree crown, and the tree crowns will satisfy a certain spatial distribution in the forest. A correction parameter (the aggregation index CI) is used to describe the degree of non-random spatial distribution of leaf elements in the canopy. The non-random distribution of leaves in the canopy is as follows: ​​

[0062] ;

[0063] Combining the two equations above, we can get:

[0064] ;

[0065] The current mainstream (CI inversion) algorithm assumes that the crown distribution is Poisson distribution, and the canopy gap fraction in the zenith direction of the opaque crown can be expressed as:

[0066] ;

[0067] where, P represents the canopy gap fraction at the lowest point, t a is the projection area of the crown in the direction of the lowest point, n is the number of crowns in the sample plot, S is the area of the sample plot. It should be noted that the S ( θ ) of the horizontal ground will become [ S-cos ( θ )].

[0068] The canopy gap fraction in any direction can be expressed as the following equation:

[0069] ;

[0070] where, t a ( θ ) and S ( θ ) represent the projection area of the crown and the forest (i.e. the sample plot) in the direction θ , respectively.

[0071] When considering the internal gap of the crown, the equation can be expressed as:

[0072] ;

[0073] where, P gap ( θ ) represents the gap fraction of a single crown in the direction θ .

[0074] When the number of crowns n is large enough, the equation is transformed into:

[0075] ;

[0076] Substitute and into Simplify, finally can obtain:

[0077] ;

[0078] Through the above formula derivation, it can be known that the aggregation index in the formula is irrelevant to the number of tree crowns n However, in actual situation, the true value of CI is closely related to the number of tree crowns n and Ccan (crown coverage, which represents the proportion of the crown projection in the study area). Therefore, the mainstream algorithm based on NDHD to retrieve CI has limitations.

[0079] (3) The improved algorithm idea of the application

[0080] In order to study the relationship between the error of the mainstream algorithm in retrieving CI and the number of tree crowns and the tree crown density, the application introduces Ccan. It is found through research that the error of the mainstream algorithm in retrieving CI has a strong linear relationship with Ccan, which means that the error of the mainstream algorithm in retrieving CI is greatly affected by Ccan. Specifically, the retrieval result of CI of the mainstream algorithm is better when applied to the sample plot with moderate Ccan, but the retrieval result of CI is poorer when applied to the sample plot with higher and lower Ccan.

[0081] Figure 1 and Figure 2 respectively represent the scatter plots of the error of the mainstream algorithm in retrieving CI and Ccan under the near-infrared band and the infrared band. Figure 1 and Figure 2 In and, the vertical coordinate is the error between the CI retrieved by the mainstream algorithm and the true CI, and the horizontal coordinate is the value of Ccan. From Figure 1 and Figure 2 It can be seen that the determination coefficient (R²) in the near-infrared band even reaches more than 0.7, and the determination coefficient (R²) in the infrared band also reaches more than 0.55. It shows that the error of the mainstream algorithm in retrieving CI is greatly affected by Ccan, and higher and lower Ccan will cause the result of retrieving CI to deviate, and when Ccan is around 0.5, the result of retrieving CI will be close to the true CI. Figure 1 and Figure 2 In and, (a), (b), (c) respectively represent the sample plot scale of 100m, 500m and 1000m, and the linear relationship between Ccan and the error under the scale of 100m, 500m and 1000m is very strong, and the slope of the fitting curve in the infrared band and the near-infrared band tends to -1. From Figure 1 and Figure 2 It can be seen that, for the intercept, the intercept of the near-infrared band is about 0.23 smaller than that of the infrared band.

[0082] The plot scale determines the size of the analysis unit, and the plot scale represents the length of the side of a single analysis unit (plot) in remote sensing images or ground observations.

[0083] In summary, it is not difficult to find that the error of CI inversion based on mainstream algorithms has a strong linear relationship with Ccan, and the inversion error is greatly affected by Ccan.

[0084] Based on the results of Figure 1 and Figure 2 , the present application introduces Ccan as a key correction factor on the basis of mainstream (CI inversion) algorithms, and establishes a CI correction model based on Ccan:

[0085] ;

[0086] Among them, CI new is the corrected CI inversion value, CI old is the CI inversion value of the mainstream algorithm before correction; d is the slope of the fitting curve, and the infrared band d takes -1.1, and the near-infrared band d takes -1; e is a constant, and takes 0.0004; l is the plot scale; f is a constant, and the infrared band f takes 0.73, and the near-infrared band f takes 0.5.

[0087] The CI correction model based on Ccan of the present application fully considers the characteristic differences of different observation scales and spectral bands, and the parameter system in the model fully considers the multi-scale observation requirements and band-specific differences.

[0088] II. As shown in Figure 7 , the aggregation index (CI) inversion considering canopy coverage (Ccan)

[0089] In order to implement the improved model, two types of key data are needed: CI value inverted by mainstream algorithm, Ccan.

[0090] The acquisition of CI value inverted by mainstream algorithm mainly depends on remote sensing image analysis technology. First, high-quality multi-spectral or hyperspectral remote sensing data need to be obtained, and common data sources include satellite images such as MODIS, Landsat and Sentinel-2. These images need to be strictly corrected for radiation and atmosphere to eliminate the influence of atmospheric scattering and sensor noise. The corrected image can be used to extract hotspot reflectivity (BHR) and land surface temperature (LST) of the study area. h ) and dark spot reflectance ( d Then, the Normalized Hot Spot and Dark Spot Index (NDHD) is calculated. Depending on the band characteristics, different infrared bands are used ( a =-1.2, b =1.28) and near-infrared band ( a =-1.34, b The regression coefficient (=1.20) was obtained through a linear relationship. The CI value obtained by the mainstream algorithm inversion is calculated.

[0091] Ccan measurements can be performed using a canopy coverage inversion method based on a pixel-division model.

[0092] In this embodiment, the canopy coverage of the study area is retrieved based on the annual maximum NDVI using a pixel-based binary model. The calculation formula is as follows:

[0093] ;

[0094] NDVI is the pixel-normalized vegetation index of remote sensing images. veg NDVI is the NDVI value of a pure vegetation pixel. soil The NDVI value is for pixels with no vegetation.

[0095] After obtaining two types of key data, the plot scale parameter ( l Three scales were selected: 100m, 500m, and 1000m. The model correction parameters ( d , e , f The data is obtained from regression analysis of a large amount of measured data.

[0096] Substitute the above data into the CI correction model based on Ccan:

[0097] ;

[0098] in, CI new This is the corrected CI inversion value. CI old It is the CI inversion value of the mainstream algorithm before the correction; d It is the slope of the fitted curve, in the infrared band. d Take -1.1, near-infrared band d Take -1; e It is a constant, taken as 0.0004; l Represents the scale of the sample plot; f It is a constant, infrared band f Take 0.73, near-infrared bandf Take 0.5.

[0099] The corrected CI value is obtained CI new The model effectively solves the systematic error problem of the mainstream algorithm based on NDHD in the sample plot with high or low Ccan by introducing Ccan as a key correction factor. This systematic data acquisition and processing method can obtain more accurate vegetation canopy structure parameters, providing reliable data support for ecological monitoring and carbon cycle research.

[0100] Embodiment 2

[0101] A correction method for the clumping index considering the canopy coverage is as follows:

[0102]

[0103] Wherein, CI new CI is the corrected CI correction value, CI old CI is the CI initial value before correction; d is the slope; e is a constant; l represents the sample plot scale; f is a constant.

[0104] The CI initial value is the CI inversion value calculated by the CI inversion algorithm; the CI inversion algorithm has similar Ccan-dependent error as the normalized hot spot and dark spot index method in embodiment 1, which can be used after adjusting the parameters. It should be noted that the CI inversion algorithm needs to be verified first to determine whether the error is linearly related to Ccan.

[0105] Embodiment 3

[0106] This embodiment uses computer simulation technology to verify the model calculation results, and simulates the radiation transmission process under different canopy structure conditions by constructing a high-precision three-dimensional virtual forest scene. Based on the two advanced radiation transmission models LESS and DART, a simulation scene with actual structure height consistent with Saihan Dam forest farm is constructed at three typical scales of 100m, 500m and 1000m. Figure 3 is the true CI value and near-infrared band CI value under the 500m scale, Figure 4 is the true CI value and infrared band CI value under the 500m scale, Figure 3 and Figure 4 ​The comparison of the CI values of the real test and the inversion is shown. It is found that the mainstream NDHD inversion algorithm has uncertainty in two special sample plots: one is the sample plot with high spatial distribution of tree crowns and low overall Ccan, and the other is the sample plot with uniform spatial distribution of tree crowns and high overall Ccan.

[0107] Figure 3 In the figure, the bottom picture is the coverage information of the sample plots, each square represents a sample plot of 500 m, and the first row of numbers in each square is the CI of the real test, and the second row of numbers is the CI of the NDHD inversion based on the near-infrared band. Figure 3 In the figure, the area shown by the solid line box is the sample plot with high spatial distribution of tree crowns and low overall Ccan, and the area shown by the dashed line box is the sample plot with uniform spatial distribution of tree crowns and high overall Ccan.

[0108] Figure 4 In the figure, the bottom picture is the coverage information of the sample plots, each square represents a sample plot of 500 m, and the first row of numbers in each square is the CI of the real test, and the second row of numbers is the CI of the NDHD inversion based on the near-infrared band. Figure 4 In the figure, the area shown by the solid line box is the sample plot with high spatial distribution of tree crowns and low overall Ccan.

[0109] In-depth verification results show that in the sample plot with high spatial distribution of tree crowns and low overall Ccan, the distribution of crown layer gaps presents high non-uniformity, which leads to deviation of reflectance characteristics, and the inversion CI value is generally overestimated. In the sample plot with uniform spatial distribution of tree crowns and high overall Ccan, the reflectance difference between hot spots and dark spots is too large, and the inversion CI value is continuously underestimated. In view of these problems, the Ccan is introduced as a key correction factor to construct a CI correction model based on Ccan. The correction model can effectively correct the overestimated CI value in the sample plot with high spatial distribution of tree crowns and low overall Ccan, and accurately compensate for the underestimated CI value in the sample plot with uniform spatial distribution of tree crowns and high overall Ccan. The correction model can significantly reduce the inversion error of these two types of sample plots, and improve the applicability and reliability of the algorithm under various Ccan conditions. This verification work not only fully proves the superior performance of the correction model, but also provides important technical reference and method reference for future research on remote sensing inversion of vegetation parameters.

[0110] As Figure 5 (infrared), Figure 6 (near-infrared) is the error effect picture after the correction model is corrected. Figure 5 and Figure 6In the table, (a1), (b1), (c1) represent the errors of CI of the main stream algorithm before correction at the scales of 100m, 500m and 1000m respectively, and (a2), (b2), (c2) represent the errors of CI after correction at the scales of 100m, 500m and 1000m respectively.

[0111] From Figure 5 It can be seen that the errors after correction are significantly reduced. A large part of the errors of CI before correction in the infrared band exceeds 0.5, and most of the errors of CI after correction are below 0.3. The maximum error of CI before correction at the scale of 100m in the infrared band is 1.155, and the maximum error of CI after correction is 0.7223. The maximum error of CI before correction at the scale of 500m is 0.8968, and the maximum error of CI after correction is 0.4113. The maximum error of CI before correction at the scale of 1000m is 0.7041, and the maximum error of CI after correction is 0.1683. The average error at the scale of 100m is reduced from 0.3203 before correction to 0.1524 after correction, with a same-period error reduction of 52.42%. The average error at the scale of 500m is reduced from 0.4587 before correction to 0.1337 after correction, with a same-period error reduction of 70.85%. The average error at the scale of 1000m is reduced from 0.5292 before correction to 0.0806 after correction, with a same-period error reduction of 84.78%.

[0112] From Figure 6 It can be seen that the errors in the near-infrared band are slightly smaller than those in the infrared band. The maximum error of CI before correction at the scale of 100m in the near-infrared band is 0.9585, and the maximum error of CI after correction is 0.6164. The maximum error of CI before correction at the scale of 500m is 0.6136, and the maximum error of CI after correction is 0.3173. The maximum error of CI before correction at the scale of 1000m is 0.4539, and the maximum error of CI after correction is 0.0179. The average error at the scale of 100m is reduced from 0.2388 before correction to 0.1186 after correction, with a same-period error reduction of 50.34%. The average error at the scale of 500m is reduced from 0.2248 before correction to 0.0852 after correction, with a same-period error reduction of 62.01%. The average error at the scale of 1000m is reduced from 0.2598 before correction to 0.0733 after correction, with a same-period error reduction of 71.79%.

[0113] In summary, whether in the infrared band or in the near-infrared band, the accuracy of CI after correction by the correction model is greatly improved, and the error is significantly reduced, especially in the infrared band.

[0114] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, and any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for inverting the aggregation index taking into account canopy cover, characterized in that, Remote sensing images of the vegetation canopy were collected, and hotspot reflectance and dark spot reflectance were extracted respectively. Calculate the Normalized Hot Spot and Dark Spot Index (NDHD), and calculate the initial value of CI based on the linear relationship between NDHD and the Clustering Index (CI). The initial CI value is corrected using the canopy coverage Ccan, where the error between the initial CI value and the actual CI value is linearly related to Ccan; The initial value of CI is calculated as follows: in, h For hotspot reflectivity, d Reflectance of dark spots; a and b The coefficients were determined through linear regression; the infrared band was selected. a It is -1.

2. b It is 1.28; the near-infrared band is taken as... a It is -1.

34. b It is 1.20; Canopy coverage Ccan is calculated based on a pixel-based binary model: In the formula, NDVI is the normalized vegetation index of a pixel. veg NDVI is the NDVI value of a pure vegetation pixel. soil The NDVI value for pixels with no vegetation. The correction method is as follows: in, CI new This is the CI correction value; CI old Initial value for CI; d The slope; F This is the intercept.

2. The aggregation index inversion method considering canopy coverage according to claim 1, characterized in that, Infrared band slope d Take -1.1, near-infrared band slope d Set to -1; the intercept in the infrared band. F Compared to the near-infrared band intercept F Larger than 0.

23.

3. The aggregation index inversion method considering canopy coverage according to claim 1, characterized in that, ;in, e It is a constant. e Take 0.0004; l For plot size; f It is a constant, infrared band f Take 0.73, near-infrared band f Take 0.

5.

4. A computer program product, characterized in that, It includes a computer program / instruction that, when executed by a processor, implements a clustering index inversion method taking into account canopy coverage as described in any one of claims 1 to 3.

Images