Mountain forest biomass identification method and system based on high-resolution remote sensing data

Abstract

The application relates to the technical field of remote sensing biomass detection, in particular to a mountain forest biomass recognition method and system based on high-resolution remote sensing data.The method comprises the following steps: based on the slope characteristic value of each pixel, all clustering clusters are divided into high-slope shady slope clusters and low-slope sunny slope clusters; based on the pixel value of all pixels in each low-slope sunny slope cluster, feature pixels are screened out from each low-slope sunny slope cluster; based on the fusion result of the pixel value mean value and the texture characteristic value of all pixels in each high-slope shady slope cluster, and in combination with the confusion degree of the coordinates of all feature pixels in each low-slope sunny slope cluster, the biological complexity is determined, and the biomass of a to-be-recognized mountain forest region is recognized. The application solves the problem that the biomass in the mountain forest is wrongly estimated due to the interference of terrain factors, and improves the accuracy of the recognition of the biomass condition of the mountain forest.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing biomass detection, in particular to a mountain forest biomass recognition method and system based on high-resolution remote sensing data. BACKGROUND

[0002] As an important part of the terrestrial ecosystem, the evaluation of forest aboveground biomass is a key indicator to measure the structure, function and carbon storage of forest ecosystems, which has important significance in the research of carbon cycle, climate change and sustainable utilization of forest resources. Traditional biomass survey has problems such as high research cost, long cycle, and difficulty in covering large areas, especially poor adaptability in complex terrain or remote forest areas, while high-resolution remote sensing technology can detect and evaluate forest biomass in a large area in a non-destructive way, supporting multi-scale biomass estimation from sample plots to regions.

[0003] Remote sensing feature extraction of forest vegetation is the core step of recognizing mountain forest biomass through high-resolution remote sensing data. Slope affects the flow of water and soil nutrients in the vertical direction, and the steepness of the slope has a significant impact on soil thickness. However, traditional high-resolution remote sensing data evaluation of biomass ignores the impact of slope and aspect on biomass in mountain terrain factors, which cannot accurately determine the mountain terrain, and thus cannot accurately capture the complex characteristics between terrain factors and biomass, resulting in an increased risk of biomass being overestimated due to terrain factors interference, and the inability to accurately identify the mountain forest biomass status. SUMMARY

[0004] In order to solve the above technical problems, the purpose of the present application is to provide a mountain forest biomass recognition method and system based on high-resolution remote sensing data, and the technical solution adopted is as follows:

[0005] In the first aspect, the present application provides a mountain forest biomass recognition method based on high-resolution remote sensing data, which comprises the following steps:

[0006] Obtain the optical remote sensing image of the mountain forest area to be identified, and the spectral images of the mountain forest area in the preset first wave band and the preset second wave band respectively, and record them as the first spectral image and the second spectral image respectively;

[0007] determining a scattering characteristic value of each pixel in the first spectral image based on the mean of reflectivity of all pixels in the neighborhood of each pixel and the difference of reflectivity between each pixel and all pixels in the neighborhood of each pixel; determining a reflectance dispersion of each pixel in the second spectral image based on the distribution of all pixels in the neighborhood of each pixel and the dispersion degree of reflectivity gradient value of each pixel in all directions in the neighborhood of each pixel, and determining a slope characteristic value of each pixel in combination with the scattering characteristic value; clustering all pixels in the optical remote sensing image, and dividing all clusters into high-slope shady slope clusters and low-slope sunny slope clusters based on the average distribution of slope characteristic values of all pixels in each cluster;

[0008] determining a texture characteristic value of each high-slope shady slope cluster based on the autocorrelation degree of pixel values of all pixels in each high-slope shady slope cluster; selecting feature pixels from each low-slope sunny slope cluster based on the pixel values of all pixels in each low-slope sunny slope cluster; determining the biological complexity of the mountain forest region to be identified based on the fusion result of the mean of pixel values of all pixels in each high-slope shady slope cluster and the texture characteristic value, and in combination with the confusion degree of coordinates of all feature pixels in each low-slope sunny slope cluster;

[0009] identifying the biomass of the mountain forest region to be identified based on the biological complexity.

[0010] Preferably, the method for determining the scattering characteristic value of each pixel is as follows:

[0011] calculating the cumulative sum of the difference of reflectivity between each pixel and all pixels in the neighborhood of each pixel in the first spectral image, and taking the fusion result of the mean of reflectivity of all pixels in the neighborhood of each pixel in the first spectral image and the cumulative sum as the scattering characteristic value of each pixel.

[0012] Preferably, the expression of the reflectance dispersion of each pixel is as follows: ; wherein, represents the reflectance dispersion of pixel i; represents the dispersion degree of reflectivity gradient value of pixel i in all directions in the neighborhood of pixel i in the second spectral image; represents the mean of reflectivity of all pixels in the neighborhood of pixel i in the second spectral image; represents a preset constant greater than 0.

[0013] Preferably, the slope characteristic value of each pixel is the result of taking the normalized value of the ratio of the scattering characteristic value and the reflectance dispersion of each pixel.

[0014] Preferably, the dividing of all clusters into high-slope shady slope clusters and low-slope sunny slope clusters comprises:

[0015] The mean value of the slope feature value of all pixels in each cluster is calculated and recorded as a feature mean value, and the mean values of all clusters are taken as the input of a threshold segmentation algorithm, and a segmentation threshold is outputted, and all clusters with a feature mean value greater than or equal to the segmentation threshold are recorded as high-slope shady-slope clusters, and all the remaining clusters are recorded as low-slope sunny-slope clusters.

[0016] Preferably, the texture feature value of each high-slope shady-slope cluster is the dispersion degree of all elements in the autocorrelation sequence of the pixel values of all pixels in the high-slope shady-slope cluster.

[0017] Preferably, the feature pixels are selected from each low-slope sunny-slope cluster, including:

[0018] The pixel values of all pixels in each low-slope sunny-slope cluster are taken as the input of a threshold segmentation algorithm, and a segmentation threshold is outputted and recorded as a pixel threshold, and all pixels with a pixel value greater than or equal to the pixel threshold are recorded as feature pixels.

[0019] Preferably, the expression of the biological complexity of the mountain forest region to be identified is: ; in the formula, represents the biological complexity of the mountain forest region to be identified; represents the reciprocal of the product of the mean value of the pixel values of all pixels in the jth high-slope shady-slope cluster and the texture feature value; represents the Shannon entropy of the coordinates of all pixels in the kth low-slope sunny-slope cluster; 、 respectively represent the number of all high-slope shady-slope clusters and the number of all low-slope sunny-slope clusters of the mountain forest region to be identified; and norm[ ] represents a normalization function.

[0020] Preferably, the biological complexity of the mountain forest region to be identified is identified, including:

[0021] If the biological complexity of the mountain forest region to be identified is greater than a preset threshold, then the biological quantity of the mountain forest region to be identified is abundant, and otherwise, the biological quantity of the mountain forest region to be identified is deficient.

[0022] In a second aspect, the embodiments of the present application also provide a mountain forest biological quantity identification system based on high-resolution remote sensing data, including a memory, a processor, and a computer program stored in the memory and running on the processor, and the processor implements the steps of the mountain forest biological quantity identification method based on high-resolution remote sensing data in any of the above aspects when executing the computer program.

[0023] The present application has at least the following beneficial effects:

[0024] The application analyzes the reflectivity mean value, difference and gradient dispersion of the pixels in the spectral image, constructs the scattering characteristic value and the reflection dispersion, and calculates the slope characteristic value, so as to effectively distinguish the high-slope shady slope and the low-slope sunny slope region, combine the terrain slope and slope direction information in the high-resolution optical image which is difficult to be directly quantified with the spectral information, and overcome the problem that the single spectrum or texture feature cannot accurately reflect the influence of the complex terrain in the mountainous area. Further, the application screens the pixels with high pixel values in the optical remote sensing image of the to-be-identified mountainous forest region, that is, the feature pixels, to represent the region corresponding to the exposed rock in the mountainous forest, can analyze the influence of these key ground objects on the overall biomass recognition in the mountainous forest, instead of being averaged by the dense vegetation pixels, and thus improves the accuracy of the biomass condition recognition in the mountainous forest region. Further, the application analyzes the texture uniformity of the high-slope shady slope region and the dispersion of the rock distribution in the low-slope sunny slope region, constructs the biological complexity index, which ingeniously combines the influence of the terrain and light on the vegetation growth, can effectively reflect the distribution condition and overall complexity of the biomass in the mountainous forest region, and improves the accuracy of the biomass condition recognition in the mountainous forest region. BRIEF DESCRIPTION OF DRAWINGS

[0025] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort.

[0026] Figure 1 The step flow chart of the mountainous forest biomass recognition method based on high-resolution remote sensing data provided by one embodiment of the present application is shown in the figure.

[0027] Figure 2 The cluster division schematic diagram provided by one embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0028] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the mountainous forest biomass recognition method and system based on high-resolution remote sensing data according to the present application, the specific implementation, structure, features and effects thereof are described in detail as follows by combining the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0030] The specific scheme of the mountainous forest biomass recognition method and system based on high-resolution remote sensing data provided by the present application will be specifically described below in combination with the drawings.

[0031] Please refer to Figure 1 which shows the step flowchart of the mountainous forest biomass recognition method based on high-resolution remote sensing data provided by an embodiment of the present application, which comprises the following steps:

[0032] Step S1: Obtain the optical remote sensing image of the mountainous forest region to be recognized, and the spectral images of the mountainous forest region in the preset first wave band and the preset second wave band, respectively, and mark them as the first spectral image and the second spectral image.

[0033] The optical remote sensing image of the mountainous forest region to be recognized, and the spectral images of the mountainous forest region in the preset first wave band and the preset second wave band are obtained by Jilin-1KF01A_0.75m optical high-resolution remote sensing satellite, and are marked as the first spectral image and the second spectral image, respectively. The size of the preset first wave band is 450-510nm, and the size of the preset second wave band is 510-580nm. 450-510nm is mainly blue light distribution wave band, containing a small amount of green light, and 510-580nm is mainly green light and part of yellow light. All the images obtained above are corrected by FLASH model for atmospheric correction, to correct the spatial heterogeneity interference caused by aerosol optical depth in the mountainous forest steep slope region. RPC orthographic correction model is used to geometrically correct the images obtained above, to eliminate the influence distortion caused by the terrain undulation.

[0034] In order to prevent the optical remote sensing image of the mountainous forest region obtained and the spectral images of the two wave bands from being seriously interfered by noise and affecting the subsequent analysis, the wavelet transform algorithm is used in this embodiment to denoise the optical remote sensing image obtained and the spectral images of the four wave bands.

[0035] Among them, FLASH model, RPC orthographic correction model and wavelet transform algorithm are all known technologies, and the specific principles are not described again.

[0036] In this embodiment, in order to facilitate analysis, the optical remote sensing image obtained above is subjected to gray scale processing, therefore, the pixel value of the pixel in the optical remote sensing image mentioned in the following content is the gray value of the pixel.

[0037] Step S2: determining a scattering feature value of each pixel based on the reflectivity mean value of all pixels in the neighborhood of each pixel in the first spectral image and the difference between the reflectivity of each pixel and all pixels in its neighborhood; determining a reflection dispersion of each pixel based on the distribution of all pixels in the neighborhood of each pixel in the second spectral image and the dispersion degree of the reflectivity gradient value of each pixel in all directions in its neighborhood, and combining the scattering feature value to determine the slope feature value of each pixel; clustering all pixels in the optical remote sensing image, and dividing all clusters into high-slope shady slope clusters and low-slope sunny slope clusters based on the average distribution of the slope feature values of all pixels in each cluster.

[0038] In the mountain forest region to be identified, light is one of the key factors for plant growth. In the region with sufficient light conditions, the growth of vegetation is healthier. In the steep slope region, the shadow area of the vegetation topography is significant, and the sun radiation is less. In the gentle slope, the vegetation receives more sun radiation. At the same time, the slope direction has a significant impact on the richness of the mountain forest plants. The vegetation in the sunny slope region can enjoy more sufficient light conditions than that in the shady slope region.

[0039] Specifically, the 450-510 nm band corresponds to the green wave band, that is, the vegetation chlorophyll absorption valley and the blue light scattering area. The increase of the slope reduces the angle of incidence of the sun, and the vegetation in the region is exposed as the slope direction is more towards the light shade, which will cause the reflectivity in the spectral image region to increase. The 510-580 nm band corresponds to the strong absorption area of vegetation chlorophyll. The decrease of the slope and the vegetation in the region can obtain sufficient light as the slope direction is more towards the light sunny side. The vegetation in the region is dense, the photosynthesis of chlorophyll is more sufficient, the low reflectivity condition in the region is less, and the gradient direction of the reflectivity in the neighborhood range is more regular under the influence of the plant growth light.

[0040] Based on the above analysis, in the embodiment, the scattering feature value of each pixel is determined based on the reflectivity mean value of all pixels in the neighborhood of each pixel in the first spectral image and the difference between the reflectivity of each pixel and all pixels in its neighborhood. The reflection dispersion of each pixel is determined based on the distribution of all pixels in the neighborhood of each pixel in the second spectral image and the dispersion degree of the reflectivity gradient value of each pixel in all directions in its neighborhood, and the slope feature value of each pixel is determined in combination with the scattering feature value, which is used to represent the topographic high slope and the shady slope direction of each region in the mountain forest spectral image. Specifically,

[0041] As an implementation manner, in the embodiment, first, the cumulative sum of the difference between the reflectivity of each pixel and all pixels in its neighborhood in the first spectral image is calculated. The result of the positive fusion of the reflectivity mean value of all pixels in the neighborhood of each pixel in the first spectral image and the cumulative sum is taken as the scattering feature value of each pixel.

[0042] It should be noted that the neighborhood construction process of each pixel is to construct a neighborhood with each pixel as the center, and the neighborhood size is In the embodiment, the value of n is 7, and in actual application, the implementer can also set it by himself according to the specific situation, which is not specially limited in the embodiment.

[0043] It should be understood that the forward fusion refers to combining two or more indexes together by adding or multiplying or other ways, so as to obtain a comprehensive index, so as to more comprehensively and accurately evaluate a phenomenon or problem. The fusion method is not limited to simple arithmetic operation, but also can include more complex statistical model and analysis method, and the implementer can select it according to the specific situation, which is not specially limited in the embodiment.

[0044] Preferably, as an implementation manner, the expression of the scattering characteristic value of each pixel is: ; in the formula, represents the scattering characteristic value of the pixel i; represents the mean value of the reflectivity of all pixels in the neighborhood of the pixel i in the first spectral map; represents the cumulative sum of the difference of the reflectivity between the pixel i and all pixels in the neighborhood of the pixel i in the first spectral map; exp() represents the exponential function with the natural constant as the base.

[0045] It should be noted that there are many methods to measure the difference between data, and in the embodiment, the absolute value of the difference of the reflectivity between each pixel in the first spectral map and all pixels in the neighborhood of the pixel is used as the difference of the reflectivity between the pixel i and all pixels in the neighborhood of the pixel in the first spectral map, and in actual application, as other implementation manners, the implementer can also use the square or ratio of the difference value or other methods to measure the difference between data, and the selection of the method to measure the difference between data is not specially limited in the embodiment.

[0046] According to the scattering characteristic value of each pixel, it can be understood that if the reflectivity mean value of all pixels in the neighborhood of pixel i in the first spectrum is larger, it means that the vegetation surface in the corresponding mountain forest area of ​​pixel i and its neighborhood reflects more blue-green light. This indicates that the slope of the area is higher and the angle of incidence of the sun is reduced, making it easier for light to be scattered to the sensor by the rough ground or exposed vegetation surface instead of being absorbed by chlorophyll. In addition, the area is steep and may also be facing the shade, because the relative coefficient of vegetation facing the shade reduces the absorption of light and increases the scattering of light. Therefore, , the larger the corresponding scattering characteristic value; at the same time, if the cumulative sum of the differences in reflectance between pixel i and all pixels in its neighborhood in the first spectrum is larger, it means that the spectral reflectance of different positions in the neighborhood of pixel i is very different. Combined with the influence of the large mean reflectance of all pixels in the neighborhood of pixel i in the first spectrum, it means that due to the geographical conditions of steep slope and shady slope, the reflectance difference between pixel i and different pixels in its neighborhood is larger. Therefore, the larger the corresponding scattering characteristic value, the more likely the current pixel is to correspond to a high-slope shady slope area.

[0047] On the contrary, if the mean reflectivity of all pixels in the neighborhood of pixel i in the first spectrum is smaller, it means that the vegetation surface in the corresponding mountain forest area of ​​pixel i and its neighborhood reflects less blue light, which means that the area is unlikely to have a high slope and a small solar incidence angle, and the light is not easily scattered by rough ground or exposed surfaces, or the vegetation itself has a strong absorption of blue light. Moreover, it is unlikely that the area is both steep and shady, because scattering will usually increase under such conditions. Therefore, the smaller the corresponding scattering characteristic value is. At the same time, if the cumulative sum of the differences in reflectance between pixel i and all pixels in its neighborhood in the first spectral map is smaller, it means that the spectral reflectances at different positions in the neighborhood of pixel i are not much different, and the overall spectral characteristics are relatively uniform. This indicates that not only is the blue light reflectance in the area where pixel i is located relatively low overall, but this low state is also distributed relatively consistently in the neighborhood without drastic changes. Therefore, the smaller the corresponding scattering characteristic value is, the more likely the current pixel is not in a high-slope shady area, and its terrain and vegetation conditions tend to be more gentle and sunny, or have uniform and dense vegetation coverage.

[0048] Furthermore, this embodiment determines the reflectance dispersion of each pixel based on the distribution of all pixels in the neighborhood of each pixel in the second spectrum and the dispersion of the reflectance gradient values ​​of each pixel in all directions in its neighborhood, specifically:

[0049] In this embodiment, the reflection discreteness of pixel i is The expression is: Where, Indicates the discrete degree of reflectance gradient value of pixel i in all directions within its neighborhood in the second spectrum image; represents the reflectivity mean value of all the pixels in the neighborhood of pixel i in the second spectral image; represents a preset constant greater than 0, used to prevent the denominator from being 0, The value of a is artificially set in the embodiment The value of b is 0.01, which is set by the implementer according to the specific situation without special limitation in the embodiment.

[0050] It should be noted that there are many methods to measure the dispersion degree of a group of data. In the embodiment, the variance of the reflectivity gradient values of pixel i in all directions in its neighborhood in the second spectral image is used as the dispersion degree of the reflectivity gradient values of pixel i in all directions in its neighborhood in the second spectral image. In actual application, as an alternative, the implementer can use other methods to measure the dispersion degree of data, such as standard deviation or dispersion coefficient. The selection of the method to measure the dispersion degree of data is not specially limited in the embodiment.

[0051] It should be further noted that the calculation method of variance is used in the embodiment to measure the dispersion degree of data.

[0052] In addition, it should be noted that the Laplacian operator is used to calculate the gradient value in the embodiment to calculate the reflectivity gradient values of pixel i in all directions in its neighborhood. In actual application, as an alternative, the implementer can use the Sobel operator to calculate the specific gradient value according to the specific situation. The selection of the method to calculate the gradient value is not specially limited in the embodiment.

[0053] The Laplacian operator is a known technology, and the specific process of calculating the gradient value using the Laplacian operator is not repeated here.

[0054] According to the reflection dispersion of each pixel, it can be understood that the greater the dispersion degree of the reflectance gradient value of pixel i in all directions in its neighborhood in the second spectral diagram, the greater the corresponding reflection dispersion, which indicates that the reflectance in different directions in the neighborhood of pixel i varies greatly, that is, the direction trend of the reflectance gradient is irregular, which may be caused by the complex structure of the vegetation canopy in the mountain forest region corresponding to the neighborhood of the pixel, resulting in a sharp spatial variation of the reflectance, and the vegetation canopy structure in the low-slope sunny slope region is more complex and the vegetation species are more abundant, so the corresponding reflection dispersion is greater. At the same time, if the average reflectance of all pixels in the neighborhood of pixel i in the second spectral diagram is smaller, it indicates that the vegetation in the neighborhood of pixel i corresponding to the mountain forest region is more dense and complex, which indicates that the neighborhood of pixel i corresponding to the mountain forest region is more likely to be a low-slope sunny slope region because of the strong photosynthesis and dense vegetation in the low-slope sunny slope region.

[0055] On the contrary, if the dispersion degree of the reflectance gradient value of pixel i in all directions in its neighborhood in the second spectral diagram is smaller, the corresponding reflection dispersion is smaller, which indicates that the reflectance in different directions in the neighborhood of pixel i varies little, that is, the direction trend of the reflectance gradient is regular and consistent, which may be caused by the relatively simple or uniform structure of the vegetation canopy in the mountain forest region corresponding to the neighborhood of the pixel, resulting in a gentle spatial variation of the reflectance, and the vegetation canopy structure in the high-slope shady slope region may be relatively sparse or affected by the terrain, or the terrain is steep but the shadow pattern is single due to consistent illumination conditions, so the corresponding reflection dispersion is smaller. At the same time, if the average reflectance of all pixels in the neighborhood of pixel i in the second spectral diagram is greater, it indicates that the vegetation in the neighborhood of pixel i corresponding to the mountain forest region is relatively sparse, or the bare ground part is more, and the absorbed light is relatively less, which indicates that the neighborhood of pixel i corresponding to the mountain forest region is more likely to be a high-slope shady slope region because of the poor illumination condition and limited vegetation growth in the high-slope shady slope region.

[0056] Further, the embodiment determines the slope feature value of each pixel based on the reflection dispersion of each pixel and in combination with the scattering feature value, specifically:

[0057] In the embodiment, the ratio of the scattering feature value of each pixel to the reflection dispersion is taken as the result of the normalized value as the slope feature value of each pixel.

[0058] According to the slope characteristic value of each pixel, it can be understood that the greater the scattering characteristic value of pixel i, the more likely pixel i corresponds to a region with relatively exposed vegetation on a steep and shady slope, and thus the greater the corresponding slope characteristic value. At the same time, the smaller the reflection dispersion of pixel i, the more likely pixel i corresponds to a high-slope shady region in a mountainous forest area, because the light condition is poor in a high-slope shady region, and the vegetation growth is limited, and thus the greater the corresponding slope characteristic value. Conversely, the smaller the scattering characteristic value of pixel i, the more likely pixel i corresponds to a region with relatively dense vegetation on a low-slope sunny slope, and thus the smaller the corresponding slope characteristic value. At the same time, the greater the reflection dispersion of pixel i, the more likely pixel i corresponds to a low-slope sunny region in a mountainous forest area, because the vegetation canopy structure is complex and the species are rich in a low-slope sunny region, resulting in a sharp change in reflectivity in space, and thus the smaller the corresponding slope characteristic value.

[0059] Further, the embodiment clusters all the pixels in the optical remote sensing image, and divides all the clusters into high-slope shady clusters and low-slope sunny clusters based on the average distribution of the slope characteristic values of all the pixels in each cluster.

[0060] In the embodiment, the k-means clustering algorithm is used to cluster all the pixels in the optical remote sensing image, wherein the distance metric is set as the absolute value of the difference between the slope characteristic values of the pixels, the number of the clusters is obtained by the elbow method, and finally all the clusters are output.

[0061] The k-means clustering algorithm and the elbow method are both known technologies, and their specific principles will not be repeated.

[0062] Further, the embodiment calculates the mean value of the slope characteristic values of all the pixels in each cluster, denoted as a feature mean value, uses the feature mean values of all the clusters as the input of the threshold segmentation algorithm, outputs a segmentation threshold, and records all the clusters with a feature mean value greater than or equal to the segmentation threshold as high-slope shady clusters, and records all the remaining clusters as low-slope sunny clusters.

[0063] It should be noted that there are many commonly used threshold segmentation algorithms, and the maximum inter-cluster variance algorithm is used to classify the clusters in the embodiment. In actual application, as other implementation manners, the implementer can also use other threshold segmentation algorithms according to the specific circumstances, and the embodiment does not make special limitations.

[0064] It should be further noted that the maximum inter-cluster contrast algorithm is used in the embodiment.

[0065] Preferably, the cluster division schematic diagram provided by the embodiment is as shown in Figure 2 .

[0066] So far, by analyzing the reflectance mean value, difference and gradient dispersion of the pixels in the spectral image, the scattering characteristic value and the reflection dispersion are constructed, and the slope characteristic value is calculated, which effectively distinguishes the high-slope shady slope and the low-slope sunny slope region. This method takes advantage of the significant influence of terrain and light on the spectral characteristics of vegetation, providing a powerful quantitative basis for the fine management of mountain forest regions, and further improving the accuracy of mountain forest biomass condition recognition.

[0067] Step S3: determining the texture characteristic value of each high-slope shady slope cluster based on the autocorrelation degree of the pixel values of all pixels in each high-slope shady slope cluster; screening out feature pixels from each low-slope sunny slope cluster based on the pixel values of all pixels in each low-slope sunny slope cluster; determining the biological complexity of the mountain forest region to be identified based on the fusion result of the mean value of the pixel values of all pixels in each high-slope shady slope cluster and the texture characteristic value, and combining the confusion degree of the coordinates of all feature pixels in each low-slope sunny slope cluster.

[0068] In the process of identifying mountain forest biomass by using high-resolution remote sensing images, the high-slope shady slope region has weak light evaporation and high soil humidity, which creates more favorable conditions for the growth of trees and other vegetation; while the vegetation in the low-slope sunny slope region can enjoy sufficient photosynthesis, but due to the strong light evaporation, it is not conducive to the growth of non-drought-tolerant vegetation, and it may also cause vegetation to be exposed and have low biomass density. Simple texture feature extraction cannot accurately capture the high biomass characteristics of mountain forest regions.

[0069] Specifically, in the mountain forest optical remote sensing gray image, when the vegetation biomass of the high-slope shady slope region is more dense, the continuous canopy of suitable growing plants such as trees will produce a uniform canopy layer, and the texture details in the optical remote sensing gray image will be more blurred. At the same time, due to the lush canopy, the internal shadow coverage of the mountain forest is wide, and the average pixel value is dark; while the more vegetation biomass of the low-slope sunny slope region, due to the high vegetation coverage, the plants are arranged closely and the plants usually have phototaxis, so the area with high pixel value, i.e. the rock or vegetation exposed area, accounts for a smaller proportion and is more dispersed.

[0070] Therefore, based on the above analysis, the embodiment determines the texture characteristic value of each high-slope shady slope cluster based on the autocorrelation degree of the pixel values of all pixels in each high-slope shady slope cluster; screens out feature pixels from each low-slope sunny slope cluster based on the pixel values of all pixels in each low-slope sunny slope cluster; determines the biological complexity of the mountain forest region to be identified based on the fusion result of the mean value of the pixel values of all pixels in each high-slope shady slope cluster and the texture characteristic value, and combines the confusion degree of the coordinates of all feature pixels in each low-slope sunny slope cluster, specifically:

[0071] As an implementation, in the embodiment, the dispersion degree of all elements in the autocorrelation sequence of all pixel values of each high-slope shady slope cluster is taken as the texture feature value of each high-slope shady slope cluster, where the dispersion degree is calculated by using variance.

[0072] Wherein, the method for obtaining the autocorrelation sequence is a known technology, and the specific obtaining process will not be repeated.

[0073] It is additionally explained that in the embodiment, all element values in the autocorrelation sequence are used to measure the autocorrelation degree of all pixel values of each high-slope shady slope cluster.

[0074] Further, in the embodiment, the pixel values of all pixels in each low-slope sunny slope cluster are taken as the input of the threshold segmentation algorithm, the segmentation threshold is output and recorded as the pixel threshold, and all pixels with pixel values greater than or equal to the pixel threshold are recorded as feature pixels, which are used to represent the exposed rock region.

[0075] Further, in the embodiment, the biological complexity of the to-be-identified mountain forest region is determined based on the fusion result of the mean value of the pixel values of all pixels in each high-slope shady slope cluster and the texture feature value, and the confusion degree of the coordinates of all feature pixels in each low-slope sunny slope cluster, and specifically:

[0076] In the embodiment, the biological complexity of the to-be-identified mountain forest region is expressed as: ; in the formula, represents the inverse of the multiplication result of the mean value of the pixel values of all pixels in the jth high-slope shady slope cluster and the texture feature value; represents the Shannon entropy of the coordinates of all pixels in the kth low-slope sunny slope cluster; , respectively represent the number of all high-slope shady slope clusters of the to-be-identified mountain forest region and the number of all low-slope sunny slope clusters; norm[ ] represents a normalization function.

[0077] Wherein, the method for calculating the Shannon entropy is a known technology, and the specific calculation process will not be repeated.

[0078] According to the biological complexity of the mountain forest region to be identified, it can be understood that the greater the reciprocal of the product of the mean value of the pixel values of all pixels in the high-slope shady slope cluster and the texture feature value, the more blurred the texture details in the optical remote sensing gray image of the corresponding high-slope shady slope region, and the lower the overall gray value, which usually reflects that the vegetation in the region grows densely and forms a continuous canopy layer, resulting in image texture homogenization. Therefore, the greater the reciprocal of the product, the smaller the corresponding biological complexity. At the same time, the greater the Shannon entropy of the coordinates of all pixels in the low-slope sunny slope cluster, the more dispersed and uniform the two-dimensional coordinate distribution of the feature pixels in the corresponding low-slope sunny slope region, and there is no obvious concentrated area, which usually means that the overall influence of rock exposure on vegetation growth is relatively small and is widely distributed. Therefore, the greater the Shannon entropy, the higher the corresponding biological complexity, which reflects that the rock exposure in the low-slope sunny slope region is widely and dispersedly distributed, and has a relatively small local inhibitory effect on vegetation growth, which is overall conducive to the maintenance of biomass or indicates that the biomass itself is relatively uniform.

[0079] On the contrary, the smaller the reciprocal of the product of the mean value of the pixel values of all pixels in the high-slope shady slope cluster and the texture feature value, the more clear the texture details in the optical remote sensing gray image of the corresponding high-slope shady slope region, and the higher the overall gray value, which usually reflects that the vegetation in the region grows sparsely and the canopy coverage is discontinuous, resulting in image texture complication. Therefore, the smaller the reciprocal of the product, the greater the corresponding biological complexity. At the same time, the smaller the Shannon entropy of the coordinates of all pixels in the low-slope sunny slope cluster, the more concentrated and non-uniform the two-dimensional coordinate distribution of the feature pixels in the corresponding low-slope sunny slope region, and there is an obvious concentrated area, which usually means that the overall influence of rock exposure on vegetation growth is relatively large and is not widely distributed. Therefore, the smaller the Shannon entropy, the lower the corresponding biological complexity, which reflects that the rock exposure in the low-slope sunny slope region is concentratedly distributed, and has a relatively large local inhibitory effect on vegetation growth, which is overall not conducive to the maintenance of biomass or indicates that the biomass itself is relatively non-uniform.

[0080] So far, by analyzing the texture homogeneity of the high-slope shady slope region and the dispersion of rock distribution in the low-slope sunny slope region, the biological complexity index is constructed, which ingeniously combines the influence of terrain and light on vegetation growth, and can effectively reflect the distribution and overall complexity of the biomass in the mountain forest region, providing a quantitative basis for fine ecological evaluation.

[0081] Step S4: identifying the biomass of the mountain forest region to be identified based on the biological complexity.

[0082] Based on the biological complexity obtained in step S3, the embodiment further identifies the biomass of the mountain forest region to be identified based on the biological complexity, specifically:

[0083] A preset threshold U is set. If the biological complexity of the mountain forest region to be identified is greater than or equal to the preset threshold U, it is indicated that the vegetation in the high-slope shady area of the current mountain forest region to be identified is dense, the tree canopy coverage range is relatively wide, the rock exposure in the low-slope sunny area is distributed discretely, the influence on the plant life suitable for growing in the low-slope sunny area is relatively slight, and the biomass of the mountain forest region to be identified is relatively rich. Conversely, if the biological complexity of the mountain forest region to be identified is less than the preset threshold U, it is indicated that the vegetation in the high-slope shady area of the current mountain forest region to be identified is sparse, the tree canopy coverage range is relatively small, the rock exposure in the low-slope sunny area is distributed concentratedly, the influence on the plant life suitable for growing in the low-slope sunny area is relatively serious, and the biomass of the mountain forest region to be identified is relatively deficient.

[0084] It should be noted that the value of the preset threshold U in this embodiment is 0.6. In actual application, as another implementation manner, the implementer can also set it by himself according to the specific situation, and this embodiment does not have special limitation.

[0085] So far, by analyzing the remote sensing image features of different slope and aspect areas of the mountain forest, the biological complexity index is constructed, and it is associated with the biomass, so that the effective identification of the biomass is realized. This method comprehensively considers the information such as terrain, illumination and vegetation structure, overcomes the limitation of single texture feature extraction, and improves the accuracy of the mountain forest biomass condition identification.

[0086] Based on the same inventive concept as the above method, the embodiment of the present application also provides a mountain forest biomass identification system based on high-resolution remote sensing data, which comprises a memory, a processor and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of any one of the above mountain forest biomass identification methods based on high-resolution remote sensing data are implemented.

[0087] It should be noted that the above sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above describes the specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0088] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment mainly describes the differences from other embodiments.

[0089] The above description is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for identifying forest biomass in mountainous areas based on high-resolution remote sensing data, characterized by, The method comprises the following steps: Obtaining an optical remote sensing image of a mountain forest region to be identified, and spectral images of the mountain forest region in a preset first wave band and a preset second wave band respectively, and denoting the first spectral image and the second spectral image as a first spectral image and a second spectral image respectively; Determining a scattering characteristic value of each pixel in the first spectral image based on a mean value of reflectivity of all pixels in a neighborhood of each pixel and a difference in reflectivity between each pixel and all pixels in the neighborhood thereof; determining a reflection dispersion degree of each pixel in the second spectral image based on a distribution of all pixels in a neighborhood of each pixel and a dispersion degree of reflectivity gradient values of each pixel in all directions in the neighborhood thereof; and determining a slope characteristic value of each pixel in combination with the scattering characteristic value; clustering all pixels in the optical remote sensing image, and dividing all clusters into high-slope shady-slope clusters and low-slope sunny-slope clusters based on an average distribution of slope characteristic values of all pixels in each cluster. Determining a texture characteristic value of each high-slope shady-slope cluster based on a self-correlation degree of pixel values of all pixels in each high-slope shady-slope cluster; selecting feature pixels from each low-slope sunny-slope cluster based on pixel values of all pixels in each low-slope sunny-slope cluster; and determining a biological complexity of the mountain forest region to be identified based on a result of fusion of a mean value of pixel values of all pixels in each high-slope shady-slope cluster and the texture characteristic value and in combination with a confusion degree of coordinates of all feature pixels in each low-slope sunny-slope cluster. Identifying a biomass of the mountain forest region to be identified based on the biological complexity. The expression of the reflection dispersion of each pixel is: ; in which, represents the reflection dispersion of the pixel i; represents the dispersion degree of the reflectivity gradient value of the pixel i in all directions within its neighborhood in the second spectral map; represents the average reflectivity of all pixels within the neighborhood of the pixel i in the second spectral map; represents a preset constant greater than 0; The expression of the biological complexity of the mountain forest region to be identified is: ; in the formula, represents the biological complexity of the mountain forest region to be identified; represents the reciprocal of the multiplication result of the pixel value mean and the texture feature value of all pixels in the jth high-slope shady slope cluster; represents the Shannon entropy of the coordinates of all pixels in the kth low-slope sunny slope cluster; , respectively represent the number of all high-slope shady slope clusters and the number of all low-slope sunny slope clusters of the mountain forest region to be identified; and norm[ ] represents a normalization function.

2. The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, characterized in that, The method for determining the scattering characteristic value of each pixel is as follows: Calculating a cumulative sum of the difference in reflectivity between each pixel in the first spectral image and all pixels in the neighborhood thereof, and taking a result of positive fusion of the mean value of reflectivity of all pixels in the neighborhood of each pixel in the first spectral image and the cumulative sum as the scattering characteristic value of each pixel.

3. The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, characterized in that, The slope characteristic value of each pixel is a result of taking a normalized value of a ratio of the scattering characteristic value of each pixel to the reflection dispersion degree. 4.The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, wherein, The dividing of all clusters into the high-slope shady-slope clusters and the low-slope sunny-slope clusters comprises the following steps: Calculating a mean value of slope characteristic values of all pixels in each cluster, denoting the mean value as a feature mean value, taking the mean value of all clusters as an input of a threshold segmentation algorithm, outputting a segmentation threshold, and denoting all clusters with a feature mean value greater than or equal to the segmentation threshold as high-slope shady-slope clusters and denoting all remaining clusters as low-slope sunny-slope clusters. 5.The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, wherein, The texture characteristic value of each high-slope shady-slope cluster is a dispersion degree of all elements in a self-correlation sequence of pixel values of all pixels in each high-slope shady-slope cluster. 6.The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, wherein, The selecting of the feature pixels from each low-slope sunny-slope cluster comprises the following steps: Taking the pixel values of all pixels in each low-slope sunny-slope cluster as an input of a threshold segmentation algorithm, outputting a segmentation threshold and denoting the segmentation threshold as a pixel threshold, and denoting all pixels with a pixel value greater than or equal to the pixel threshold as feature pixels.

7. The mountainous forest biomass recognition method based on high-resolution remote sensing data according to claim 1, characterized in that, The identifying of the biomass of the mountain forest region to be identified comprises the following steps: If the biological complexity of the mountain forest region to be identified is greater than a preset threshold, then the biomass of the mountain forest region to be identified is abundant, otherwise, the biomass of the mountain forest region to be identified is deficient.

8. A mountainous forest biomass recognition system based on high-resolution remote sensing data, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, The processor implements the steps of the mountainous forest biomass identification method based on high-resolution remote sensing data according to any one of claims 1-7 when executing the computer program.

Images