A method for monitoring vegetation spectral diversity by integrating multi-temporal remote sensing information

Abstract

The present invention discloses a vegetation spectral diversity monitoring method integrating multi-temporal remote sensing information. The method calculates the spectral diversity index of each single-temporal window information of the original vegetation spectral image and the vegetation index image and the integrated multi-temporal window information by a moving window method, comprehensively considers the change of spectral information and the influence of time dimension, can more comprehensively and accurately reflect the spectral diversity of vegetation, and improves the accuracy and reliability of remote sensing monitoring of vegetation diversity.

Description

Technical Field

[0001] The invention relates to the field of ecological remote sensing technology, and in particular to a vegetation spectral diversity monitoring method integrating multi-temporal remote sensing information. Background Art

[0002] Plant diversity, as the complex ecological system formed by plants and their environment, as well as the sum of various ecological processes closely related to them, covers multiple levels such as species, functions and ecosystems.

[0003] Researchers have been committed to exploring the application of remote sensing technology in plant diversity assessment. Previously, some researchers have proposed the view that spectral heterogeneity is associated with vegetation diversity, namely the spectral variation hypothesis. Based on this hypothesis, a common research method has been developed to establish a direct connection between plant diversity indicators and spectral diversity indicators to assess plant diversity. Existing spectral diversity indicators mainly include indicators based on vegetation index or its changes and indicators based on spectral information theory. The former monitors biodiversity through the relationship between vegetation index or its changes (such as standard deviation, variance) and plant diversity, and the latter monitors biodiversity through the relationship between spectral information theory indicators and plant diversity. Among them, methods such as the coefficient of variation method based on all spectra and the spectral angle mapping method can fully capture the changes in spectral information and provide strong support for the assessment of plant diversity. At the same time, methods such as the convex hull volume method and the convex hull area method based on principal component analysis dimensionality reduction simplify the calculation process by reducing the dimension of the data while retaining key spectral features.

[0004] However, the current research hotspots in the field of plant spectral identification mainly include the fact that the monitoring ability of this method has room for development and that only single-phase remote sensing data is used, while only the static relationship between spectral diversity and plant diversity is explored, ignoring the impact of the time dimension on this relationship. In fact, the monitoring time and duration have a significant impact on the perception of biodiversity distribution, because the physiological and structural characteristics of plants change over time, which in turn affects their optical properties. Therefore, ignoring the time dimension will miss a lot of remote sensing information that can be used to accurately assess biodiversity. Summary of the invention

[0005] In view of the above problems, the present invention proposes a vegetation spectral diversity monitoring method that integrates multi-temporal remote sensing information, aiming to overcome the limitations of single-phase remote sensing data and improve the accuracy of plant diversity monitoring.

[0006] The technical solution of the present invention is as follows:

[0007] A vegetation spectral diversity monitoring method integrating multi-temporal remote sensing information comprises the following steps:

[0008] Acquire remote sensing images of different phases of the target area, create a mask image to remove non-vegetation pixels in the remote sensing image, and obtain multiple single-phase original spectral remote sensing images of vegetation;

[0009] Calculating the vegetation index of each single-phase in the original spectral remote sensing image of the vegetation, and fusing all the vegetation indices into the same image to obtain the vegetation index image of each single-phase;

[0010] A vegetation pixel number threshold is set, and the original vegetation spectral remote sensing image and the vegetation index image are traversed non-repeatedly based on a moving window method to obtain a plurality of single-phase window information, and multi-phase fusion window information is obtained by fusing the single-phase window information of different phases but the same spatial position;

[0011] According to the single-phase window information of the original vegetation spectral remote sensing image, the first spectral diversity index of each single phase and the second spectral diversity index of the fused multi-phase are calculated respectively; according to the multi-phase fusion window information of the vegetation index image, the third spectral diversity index of each single phase and the fourth spectral diversity index of the fused multi-phase are calculated respectively;

[0012] An inverse proportional weighting strategy is adopted to superimpose and fuse the first spectral diversity index and the third spectral diversity index, as well as the second spectral diversity index and the fourth spectral diversity index, respectively, to construct a first comprehensive spectral diversity index and a second comprehensive spectral diversity index based on the single-phase window information and the multi-phase fusion window information, respectively, which are used to characterize the spectral information and vegetation index information of plants in the target area, respectively.

[0013] In some embodiments, before the superposition and fusion, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, and the fourth spectral diversity index are normalized.

[0014] In some embodiments, during the traversal process of the moving window method, for areas in the vegetation original spectral remote sensing image and the vegetation index image that meet the requirements of the moving window method, the non-vegetation pixel areas in the vegetation original spectral remote sensing image and the vegetation index image are randomly interpolated based on the vegetation original spectral pixels and vegetation index pixels within the moving window to fill in the missing data in the non-vegetation pixel areas.

[0015] In some embodiments, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index all include four indicators: coefficient of variation, spectral angle mapping, envelope volume, and spectral α-diversity.

[0016] In some embodiments, a spectral diversity spatial distribution map of the target area is obtained, and based on the spectral diversity spatial distribution map, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index are respectively correlated and analyzed using the Pearson correlation coefficient to measure the feasibility of the diversity index.

[0017] The beneficial effects of the present invention are as follows: by calculating the spectral diversity index of each single-phase window information of the original spectral image of vegetation and the vegetation index image and the fusion of multi-phase window information through the moving window method, and comprehensively considering the changes in spectral information and the influence of the time dimension, the spectral diversity of vegetation can be more comprehensively and accurately reflected, thereby improving the accuracy and reliability of remote sensing monitoring of vegetation diversity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a vegetation spectral diversity monitoring method integrating multi-temporal remote sensing information disclosed in an embodiment of the present invention;

[0019] Figure 2 Schematic diagram of spatial distribution of the first spectral diversity index and the second spectral diversity index;

[0020] Figure 3 It is a schematic diagram of the spatial distribution of the third spectral diversity index and the fourth spectral diversity index;

[0021] Figure 4 It is a schematic diagram of the results of the first comprehensive spectral diversity index and the second comprehensive spectral diversity index;

[0022] Figure 5 It is a thermal diagram of the correlation between the first spectral diversity index and the second spectral diversity index at different time phases;

[0023] Figure 6 It is a heat diagram of the correlation between different indicators of the first spectral diversity index and the second spectral diversity index at the same time phase;

[0024] Figure 7 It is a heat diagram of the correlation between the first spectral diversity index and the second spectral diversity index and the third spectral diversity index and the fourth spectral diversity index at the same time phase and the same index;

[0025] Figure 8 This is a thermal diagram of the correlation between the same indicators of the third spectral diversity index and the fourth spectral diversity index at different phases;

[0026] Fig. 9This is a thermal diagram of the correlation between different indicators of the third spectral diversity index and the second spectral diversity index at the same time phase;

[0027] Fig.10 It is a thermal diagram of the correlation between the same indicators of the first comprehensive spectral diversity index and the second comprehensive spectral diversity index at different time phases;

[0028] Fig.11 This is a thermal diagram of the correlation between different indicators of the first comprehensive spectral diversity index and the second comprehensive spectral diversity index at the same phase. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the content of the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It is understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all the contents.

[0030] This embodiment proposes a vegetation spectral diversity monitoring method that integrates multi-temporal remote sensing information. Figure 1 As shown, the following steps S1-S5 are included:

[0031] S1, obtain remote sensing images of different phases of the target area, create a mask image to remove non-vegetation pixels in the remote sensing image, and obtain multiple single-phase vegetation original spectral remote sensing images.

[0032] The following remote sensing images of four different phases are selected from Shaoguan Danxia Lingxi River Forest Resort Park (113°47′29.603″-113°52′14.410″E, 24°49′11.465″-24°53′31.238″N) as examples. Based on Sentinel-2L2A-level multispectral remote sensing image data, the image has been atmospherically corrected, that is, the B10 cirrus cloud spectrum has been removed, leaving 12 spectra. The four phases are: 20220404, 20220723, 20220926 and 20221225, which are intended to reflect the changes in surface cover in the region under different seasons and climatic conditions. The spatial resolution of each spectral image is resampled to 10m using the nearest neighbor interpolation method, and then cropped according to demand to eliminate the errors caused by differences in spatial resolution between different bands. Finally, the 8km*8km original spectral remote sensing image of vegetation used in this specific embodiment is obtained.

[0033] Specifically, a mask image is created through the normalized vegetation index threshold results of multiple phases, non-vegetation in the remote sensing image is eliminated, and background information unrelated to vegetation is removed as much as possible, so as to obtain a vegetation original spectral remote sensing image with consistent spatial resolution, dominated by vegetation, and suitable for multi-phase data fusion. In one example, by calculating the normalized vegetation index (NDVI) of each phase remote sensing image, the NDVI threshold is set to 0.3 and pixels above this threshold are screened out, and then the intersection of pixels that meet the conditions in all phases is taken to create a mask image. This mask ensures that the analysis of all phases is targeted at vegetation coverage pixels with consistent distribution areas, and finally the original spectral remote sensing image of vegetation is obtained.

[0034] S2, calculate the vegetation index of each single-phase in the original spectral remote sensing image of vegetation, and fuse all the vegetation indices into the same image to obtain the vegetation index image of each single-phase.

[0035] In this embodiment, all vegetation indices calculated by S2 are fused into a vegetation index image in a similar spectral form. In order to avoid the influence of different vegetation indices due to different dimensions or units, the vegetation indices are normalized.

[0036] The vegetation index of each phase is calculated based on the original spectral image of vegetation, including the atmospheric impedance vegetation index ARVI, the enhanced vegetation index EVI, the chlorophyll vegetation index GCVI, the green normalized vegetation index GNDVI, the improved soil adjusted vegetation index MSAVI, the normalized vegetation index NDVI, the normalized red edge vegetation index NDRE, the optimized soil adjusted vegetation index OSAVI, the red edge chlorophyll vegetation index RECI, and the structure-insensitive pigment vegetation index SIPI. In order to prevent the influence of the range of vegetation index, each vegetation index needs to be normalized.

[0037] S3, setting the threshold of vegetation pixel number, traversing the original vegetation spectral remote sensing image and vegetation index image non-repetitively based on the moving window method, obtaining multiple single-phase window information, and obtaining multi-phase fused window information by fusing single-phase window information of different phases but the same spatial position.

[0038] In this embodiment, the vegetation pixel number threshold is set to 1 / 3 of the total number of window pixels (ie, 5*5=25), that is, if the proportion of vegetation pixels in the window is greater than 1 / 3, spectral diversity calculation is performed, otherwise the window is skipped.

[0039] The operation method of the multi-temporal fusion window information described in S3 above is as follows: first, the moving window pixels corresponding to the four single-temporal vegetation original spectral remote sensing images and vegetation index images are two-dimensionalized, that is, the three-dimensional array form is converted into a two-dimensional array form. After two-dimensionalization, the four two-dimensional arrays are spliced. Through this step, the spectral data of different phases can be integrated into a unified two-dimensional array to form a two-dimensional array of shape 100×12, which is the original spectral remote sensing image of vegetation and the vegetation index image pixel information contained in the fusion multi-temporal window.

[0040] Optionally, it also includes S301. During the traversal process of the moving window method, for areas in the original spectral remote sensing image of vegetation and the vegetation index image that meet the requirements of the moving window method, random interpolation is performed on the non-vegetation pixel areas in the original spectral remote sensing image of vegetation and the vegetation index image based on the original spectral pixels and vegetation index pixels in the moving window to fill in the missing data in the non-vegetation pixel areas, thereby maintaining the spatial continuity of the data.

[0041] S4, according to the single-phase window information of the original spectral remote sensing image of the vegetation, the first spectral diversity index of each single phase and the second spectral diversity index of the fused multi-phase are calculated respectively; according to the multi-phase fusion window information of the vegetation index image, the third spectral diversity index of each single phase and the fourth spectral diversity index of the fused multi-phase are calculated respectively.

[0042] Among them, the first spectral diversity index and the second spectral diversity index are diversity indices based on the original spectral remote sensing image of vegetation, while the third spectral diversity index and the fourth spectral diversity index are diversity indices based on the vegetation index image. In this embodiment, based on the original spectral remote sensing image of vegetation and the vegetation index image obtained by S1 and S2, a 5*5 (i.e., 50*50m) moving window is used to simultaneously traverse each phase to calculate the spectral diversity index (coefficient of variation CV, spectral angle mapping SAM, envelope volume CHV and spectral α-diversity SDα) of each single phase, and the threshold of the number of vegetation pixels is set based on the moving window method, and the window with significant influence of non-vegetation area is ignored to improve the accuracy and reliability of data fusion.

[0043] The first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index all include four indicators, namely, coefficient of variation, spectral angle mapping, envelope volume, and spectral α-diversity. Then, a spectral diversity spatial distribution map of the target area is obtained. On the basis of the spectral diversity spatial distribution map, the Pearson correlation coefficient is used to perform correlation analysis on the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index to measure the feasibility of the diversity index.

[0044] The calculation methods of the coefficient of variation (CV) based on single-phase window information and multi-phase fusion window information are:

[0045]

[0046]

[0047] In the formula, is the wavelength λ The spectral reflectance of the remote sensing image at S represents the spectrum number, and Respectively represent the wavelength of all pixels in the moving window λ The standard deviation and mean of the reflectivity at It is m vegetation index, n represents the total number of vegetation indices selected for the study, and Respectively represent all pixels in the window m The standard deviation and mean of vegetation indices. The coefficient of variation (CV) calculates the coefficient of variation of the spectral characteristic parameters within the window or the mean coefficient of variation of each spectrum of all pixels.

[0048] The calculation methods of spectral angle mapping (SAM) based on single-phase window information and multi-phase fusion window information are:

[0049]

[0050]

[0051] In the formula, arccos represents the arccosine of the inverse trigonometric function, “∙” represents the inner product of two vectors, and || ||2 represents the Euclidean distance of the vectors. Indicates L The average spectral vector of the moving windows, represents the average spectrum / vegetation index vector of the Lth single-temporal window, Indicates L The average spectrum / vegetation index vector of the fused multi-temporal windows, Indicates L Window K The spectral vector of pixels, J Indicates the number of pixels included in a single-phase window; Z indicates the number of phases. Spectral Angle Mapping (SAM) converts all spectra / vegetation indices of pixels in the window into spectral vectors of spectra / vegetation indices of the same dimension, and calculates the angle between each vector in the window and the average vector of the window as an indicator of the difference in spectral / vegetation index.

[0052] The calculation method of the CHV based on the single-phase window information and the multi-phase fusion window information includes:

[0053] The CHV method selects the first three principal components containing the main spectral / vegetation index in the moving window, and calculates the volume of the envelope occupied by the sample pixel in the spectral space constructed by the three principal components. It should be noted that the principal component analysis of single-phase remote sensing information is only based on a single phase, while the principal component analysis of multi-phase information fusion is based on multi-phase remote sensing images.

[0054] The calculation methods of spectral α-diversity based on single-phase window information and multi-phase fusion window information are:

[0055]

[0056]

[0057] In the formula, Indicates k Window i The pixel k Spectra / vegetation indices, Indicates k Window j Spectral / vegetation index averages, u represents the number of pixels in the window, v Indicates the phase number, p Represents the total number of spectral / vegetation indices.

[0058] S5, using an inverse proportional weighting strategy, superimposes and fuses the first spectral diversity index and the third spectral diversity index, as well as the second spectral diversity index and the fourth spectral diversity index, respectively, to construct the first comprehensive spectral diversity index and the second comprehensive spectral diversity index based on single-phase window information and multi-phase fusion window information, respectively, which are used to characterize the spectral information and vegetation index information of plants in the target area, respectively.

[0059] Before superposition and fusion, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index and the fourth spectral diversity index are normalized.

[0060] The spectral diversity index of each single-phase vegetation original spectral remote sensing image and vegetation index image is weighted by the inverse proportion of the kurtosis:

[0061]

[0062] In the formula, represents the kurtosis value of the spectral diversity index, It represents the average value of the spectral diversity index of the image. It represents the standard deviation of the spectral diversity index of the image. Indicates expected value. and They respectively represent the weights of the first spectral diversity index and the third spectral diversity index at this phase. and Indicates the first spectral diversity index and the third spectral diversity index in this phase. Represents the first comprehensive spectral diversity index. For the second spectral diversity index and the fourth spectral diversity index of multi-temporal fusion, a similar inverse proportional weighting method is used to obtain the second comprehensive spectral diversity index. Based on the above S1-S4, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index of the target area are obtained. Then, the Pearson correlation coefficient is used to perform correlation analysis on the results of different methods and different times, so as to measure the feasibility of the comprehensive spectral diversity index and the spectral diversity index of multi-temporal remote sensing information fusion.

[0063] Figure 2-4 represents the spatial distribution of the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index ( Figure 2 and Figure 3 The blank areas are non-vegetated areas; CHV and SDα are transformed by a certain ratio). By comparing different methods and different phases of visualization, we can deeply understand the spatial variation law and distribution characteristics of each spectral diversity index.

[0064] Figure 5The thermal diagram of the correlation between the same indicators of the first spectral diversity index and the second spectral diversity index at different time phases. The results show that the correlation of the spectral diversity index itself changes over time. This is because different plants are in different physiological and structural states at different times, and these states will be captured by the spectrum, which directly shows that this part of vegetation information will be ignored when only single-phase monitoring is used.

[0065] Figure 6 This is a heat diagram of the correlation between different indicators of the first spectral diversity index and the second spectral diversity index at the same time phase. The results show that there is a significant correlation between different spectral diversity indices at the same time phase, that is, in this practical case, different spectral diversity indices have similar monitoring capabilities at any time.

[0066] Figure 7 This is a heat diagram of the correlation between the first spectral diversity index and the second spectral diversity index and the third spectral diversity index and the fourth spectral diversity index in the same phase. The results show that the spectral diversity index calculated based on the single-phase window information and the spectral diversity index calculated based on the fusion of multi-phase information are significantly correlated and the correlation is strong, which means that the third spectral diversity index and the fourth spectral diversity index calculated based on the vegetation index image have similar monitoring capabilities as the first spectral diversity index and the second spectral diversity index calculated based on the original spectral remote sensing image of vegetation.

[0067] Figure 8 The third spectral diversity index and the fourth spectral diversity index are the same indicators in different phases. The results show that the monitoring capabilities of the third spectral diversity index and the fourth spectral diversity index will change dynamically over time, but their dynamic change characteristics are slightly different from those of the first spectral diversity index and the second spectral diversity index. The subsequent monitoring accuracy is worth exploring.

[0068] Fig. 9 The third and fourth spectral diversity indices are shown in Figure 1. The third and fourth spectral diversity indices are shown in Figure 2. The third and fourth spectral diversity indices are shown in Figure 3. The third and fourth spectral diversity indices are shown in Figure 4. The third and fourth spectral diversity indices are shown in Figure 5. The third and fourth spectral diversity indices are shown in Figure 6. The third and fourth spectral diversity indices are shown in Figure 7. The third and fourth spectral diversity indices are shown in Figure 8. The third and fourth spectral diversity indices are shown in Figure 9. The third and fourth spectral diversity indices are shown in Figure 10. The third and fourth spectral diversity indices are shown in Figure 11. The third and fourth spectral diversity indices are shown in Figure 12.

[0069] Fig.10The thermal diagram of the correlation between the same indicators of the first comprehensive spectral diversity index and the second comprehensive spectral diversity index in different phases. The results show that the inverse proportional weighting method can effectively improve the situation where CHV is affected by extreme values, with little effect on SDα, but a greater impact on the fusion multi-phase of CV in December and SAM.

[0070] Fig.11 This is a heat diagram of the correlation between different indicators in the same phase of the first comprehensive spectral diversity index and the second comprehensive spectral diversity index. The results show that under the single-phase window information, the inverse proportional weighting generally improves the correlation between different indicators. For the fusion of multi-phase information, the correlation between different indicators is generally reduced.

[0071] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A vegetation spectral diversity monitoring method integrating multi-temporal remote sensing information, characterized in that: The following steps are involved: Acquire remote sensing images of different phases of the target area, create a mask image to remove non-vegetation pixels in the remote sensing image, and obtain multiple single-phase original spectral remote sensing images of vegetation; Calculating the vegetation index of each single-phase in the original spectral remote sensing image of the vegetation, and fusing all the vegetation indexes into the same image to obtain the vegetation index image of each single-phase; A vegetation pixel number threshold is set, and the original vegetation spectral remote sensing image and the vegetation index image are traversed non-repeatedly based on a moving window method to obtain a plurality of single-phase window information, and multi-phase fusion window information is obtained by fusing the single-phase window information of different phases but the same spatial position; According to the single-phase window information of the original vegetation spectral remote sensing image, the first spectral diversity index of each single phase and the second spectral diversity index of the fused multi-phase are calculated respectively; according to the multi-phase fusion window information of the vegetation index image, the third spectral diversity index of each single phase and the fourth spectral diversity index of the fused multi-phase are calculated respectively; Adopting an inverse proportional weighting strategy, the first spectral diversity index and the third spectral diversity index, as well as the second spectral diversity index and the fourth spectral diversity index are superimposed and fused respectively, so as to construct a first comprehensive spectral diversity index and a second comprehensive spectral diversity index based on the single-phase window information and the multi-phase fusion window information, respectively, which are used to characterize the spectral information and vegetation index information of the plants in the target area; Before the superposition and fusion, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index and the fourth spectral diversity index are normalized; During the traversal process of the moving window method, for the areas in the original vegetation spectral remote sensing image and the vegetation index image that meet the requirements of the moving window method, the non-vegetation pixel areas in the original vegetation spectral remote sensing image and the vegetation index image are randomly interpolated based on the original vegetation spectral pixels and the vegetation index pixels in the moving window to fill in the missing data in the non-vegetation pixel areas; The first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index all include four indicators: coefficient of variation, spectral angle mapping, envelope volume, and spectral α-diversity; A spectral diversity spatial distribution map of the target area is obtained. Based on the spectral diversity spatial distribution map, the first spectral diversity index, the second spectral diversity index, the third spectral diversity index, the fourth spectral diversity index, the first comprehensive spectral diversity index and the second comprehensive spectral diversity index are respectively correlated using the Pearson correlation coefficient to measure the feasibility of the diversity index.

Images