Coherence coefficient and model fusion-based grassland height inversion method and device
By adopting the method of fusion of coherence coefficients and model in grass height in inversion, including constructing a priori data model and k-mean clustering, the problem that SAR coherence coefficients and grass height are difficult to directly correlate, and the accuracy and efficiency of grass height inversion are improved.
Patent Information
- Application Number
- CN202411770507.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-04
- Publication Date
- 2025-03-14
AI Technical Summary
In the prior art, the relationship between the SAR coherence coefficient and grassland height is difficult to directly establish, and the existing models have high computational complexity and insufficient generalization ability when processing nonlinear data, resulting in low accuracy and efficiency of grassland height inversion.
The grassland height inversion method based on the fusion of coherence coefficient and model is adopted. By obtaining SAR coherence coefficient data and grassland height survey data, a prior data model is constructed, and the data is k-means clustered, and the grassland height inversion and correction are combined with interpolation and empirical models.
The accuracy and efficiency of grass height inversion are improved, and the problem that coherence coefficients and grass height are difficult to directly correlate, and large-area inversion can be effectively used to utilize a small amount of grass height field survey data.
Smart Images
Figure CN119625536A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of vegetation parameter inversion, and in particular relates to a grassland height inversion method and device based on coherence coefficient and model fusion. Background Art
[0002] In current research, grassland height inversion methods based on remote sensing images can be roughly divided into three categories: empirical model method, machine learning method and multi-source data fusion method. Among them, the empirical model method mainly relies on establishing an empirical relationship between remote sensing image features (such as coherence coefficient) and grassland height. Although this method is simple to implement, the applicability and inversion accuracy of the model are often limited when facing different geographical environments or vegetation types. The machine learning model method learns the nonlinear relationship between image features and grassland height through a large amount of training data. Although this method can improve the inversion accuracy, it has high requirements on the quality and quantity of training data. In addition, the method based on multi-source data fusion attempts to combine multiple data sources to improve the accuracy and stability of grassland height inversion. However, this type of method has high computational complexity and the implementation process is relatively cumbersome.
[0003] The main problems faced in inverting grassland height through coherence coefficient include: first, the coherence coefficient is affected by many complex factors, such as surface roughness, vegetation density, sensor noise, etc., which makes it difficult to directly establish the relationship between it and grassland height; second, the existing model for inverting grass height using coherence coefficient has the problems of high computational complexity and insufficient generalization ability when processing nonlinear data. Summary of the invention
[0004] In view of the above problems in the prior art, the purpose of the embodiments of the present invention is to provide a grass height inversion method and device based on coherence coefficient and model fusion. The grass height inversion method improves the accuracy of grass height inversion.
[0005] The technical solution adopted in the embodiment of the present invention is:
[0006] The grassland height inversion method based on coherence coefficient and model fusion includes the following processes:
[0007] Obtain SAR coherence coefficient data and grassland height survey data of the area to be detected;
[0008] Constructing a priori data model based on the grassland height survey data to obtain a grassland height data set of the priori model;
[0009] Performing k-means clustering on the SAR coherence coefficient data so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data, and associating the grass height survey data in the grass height dataset with the label of the cluster in the two-dimensional image;
[0010] The representative values of grass heights contained in all clusters in the k-means clustering are obtained, and the representative values of grass heights corresponding to all clusters are interpolated to obtain a first grass height of the area to be detected.
[0011] In some embodiments, the grass height inversion method further includes:
[0012] constructing a first empirical model based on the effective coherence coefficient in the SAR coherence coefficient data and obtaining a second grass height using the first empirical model;
[0013] The first grass height is corrected based on the second grass height to obtain a third grass height.
[0014] In some embodiments, the process of constructing the first empirical model includes:
[0015] The effective coherence coefficient data is subjected to a logarithmic square root transformation, and the first empirical model is constructed through a linear transformation relationship.
[0016] In some embodiments, correcting the first grass height based on the second grass height includes:
[0017] The first grass height and the second grass height are weighted averaged to obtain a third grass height.
[0018] In some embodiments, constructing a priori data model based on the grass height survey data to obtain a grass height dataset of the priori data model includes:
[0019] Sampling the acquired grassland height survey data;
[0020] Add noise to the sampled grassland height survey data;
[0021] The grass height survey data with added noise is spliced with the acquired original grass height survey data to form a grass height data set of the prior data model.
[0022] In some embodiments, it also includes:
[0023] The index expression of the grass height dataset of the prior data model is established by using the original grass height survey data:
[0024] m i =argmin j c d -h i |
[0025] m iis the index of the grass height dataset corresponding to the prior data model of the i-th original grass height survey data; c d is the value of the jth data point in the grass height dataset of the prior data model; |c d -h i | is the absolute difference between the jth prior model data point and the ith original grassland height survey data; argmin j Indicates that the expression |c is found d -h i |The smallest value of j.
[0026] In some embodiments, performing k-means clustering on the SAR coherence coefficient data so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data includes:
[0027] Extracting effective coherence coefficients from the SAR coherence coefficient data, and performing k-means clustering on the effective coherence coefficients;
[0028] Get the average value of the effective coherence coefficient within each cluster in k-means clustering;
[0029] Create labels for each cluster based on the average effective coherence coefficient within each cluster.
[0030] In some embodiments, it is characterized in that
[0031] The first grass height of the area to be detected is obtained by interpolation method, including:
[0032] Generate a mask for each cluster in the k-means clustering and obtain the coordinates of the pixels of each cluster in the two-dimensional image of the SAR coherence coefficient data;
[0033] The pixels in each cluster are interpolated based on the natural field interpolation method to obtain the grass height value of each pixel, so as to obtain the first grass height of the area to be detected.
[0034] In some embodiments, obtaining representative values of grass heights in all clusters in k-means clustering includes:
[0035] For a first cluster including grass height survey data, obtaining an average value of all grass height survey data in the first cluster as a representative value of grass height of the first cluster;
[0036] For the second cluster that does not contain grass height survey data, the grass height representative value c of the second cluster is obtained based on the second empirical model. m_h ;
[0037] The second empirical model is:
[0038]
[0039] a and b are empirical constants. Through experimental fitting, the value of a is 1.1, the value of b is 2.5, and median(γ i-c ) represents the median of the coherence coefficient in the cluster.
[0040] The grassland height inversion device based on coherence coefficient and model fusion includes:
[0041] An acquisition module is used to acquire SAR coherence coefficient data and grass height survey data of the area to be detected;
[0042] A construction module, used to construct a priori data model based on the grass height survey data to obtain a grass height data set of the priori model;
[0043] A clustering module, configured to perform k-means clustering on the SAR coherence coefficient data, so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data, and associate the grass height survey data in the grass height data set with the label of the cluster in the two-dimensional image;
[0044] The interpolation module is used to obtain the representative values of grass height contained in all clusters in the k-means clustering, and interpolate the representative values of grass height corresponding to all clusters to obtain the first grass height of the area to be detected.
[0045] Compared with the prior art, the beneficial effects of the embodiments of the present invention are:
[0046] The grassland height inversion method of the present application proposes a method based on coherence coefficient and model fusion, which solves the problem that the SAR coherence coefficient is difficult to directly correlate with the grassland height. At the same time, it improves the efficiency of inverting grass height over a large area using a small amount of field survey data on grass height, and to a certain extent improves the accuracy of grass height inversion.
[0047] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention.
[0048] This summary of various implementations or examples of the technology described in this disclosure is not a comprehensive disclosure of the entire scope or all features of the disclosed technology. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] In the drawings, which are not necessarily drawn to scale, the same reference numerals may describe similar parts in different views. The same reference numerals with letter suffixes or different letter suffixes may represent different instances of similar parts. The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the specification and claims, are used to illustrate the embodiments of the invention. When appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.
[0050] Figure 1 This is a flow chart of a method for estimating the height of a grassland vegetation canopy based on satellite multi-angle observation data according to an embodiment of the present invention;
[0051] Figure 2 This is a comparison diagram of the original grassland height survey data before and after data enhancement in the prior model of the embodiment of the present invention;
[0052] Figure 3 This is an analysis diagram of the number of valid data in each cluster after K-Means clustering is performed on the effective coherence coefficient in the SAR coherence coefficient data according to an embodiment of the present invention;
[0053] Figure 4 This is a visualization diagram after all valid data are mapped back to the original image position after clustering in an embodiment of the present invention;
[0054] Figure 5 A relationship diagram of the grassland height survey data and the coherence coefficient data of the corresponding coordinates according to an embodiment of the present invention;
[0055] Figure 6 It is a fitting diagram for error analysis of the final inverted grass height data obtained by the fusion model of the embodiment of the present invention and the field survey data;
[0056] Figure 7 It is a structural schematic diagram of a grassland height inversion device based on coherence coefficient and model fusion according to an embodiment of the present invention. DETAILED DESCRIPTION
[0057] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0058] In order to keep the following description of the embodiments of the present invention clear and concise, detailed descriptions of well-known functions and well-known components are omitted.
[0059] The embodiment of the present invention provides a grassland height inversion method based on coherence coefficient and model fusion, and the grassland height inversion method includes the following process:
[0060] Step S100: Acquire SAR coherence coefficient data and grass height survey data of the area to be detected;
[0061] In step 100, an interference pattern may be generated using two-view SAR data to obtain SAR coherence coefficient data.
[0062] Among them, the SAR coherence coefficient calculation formula is as follows:
[0063]
[0064] Among them, A i and B i are the complex values of the two SAR images, B i * It is B i The complex conjugate of .
[0065] Step S200: constructing a priori data model based on the grass height survey data to obtain a grass height data set of the priori model;
[0066] In step 200, constructing a priori data model based on the grass height survey data includes:
[0067] Step S201: Sampling the obtained grassland height survey data. Specifically, Bootstrap sampling can be used to generate multiple enhanced samples s h , which is then used for subsequent data enhancement and model building. Bootstrap sampling can ensure the adequacy of data enhancement.
[0068] Among them, the Bootstrap sampling formula is as follows:
[0069] s h ={h i1 ,h i2 ,...,h i1000}
[0070] Among them, h i To obtain grass height data on site, h i1 ,h i2 ,...,h i1000 are the indices obtained by random sampling.
[0071] Step S202: adding noise to the sampled grassland height survey data. Specifically, adding normal distribution (mean value is 0, standard deviation is 0.01) random noise to the sampled grassland height survey data to simulate measurement error and data uncertainty.
[0072] Among them, the formula for the enhanced data sample carrying noise is as follows:
[0073] n h =s h +Ν(0,σ n 2 )
[0074] n h is the enhanced sample data with noise, Ν(0,σ n 2 ) has a mean of 0 and a standard deviation of σ n Normally distributed noise.
[0075] Step S203: splicing the grass height survey data with added noise with the acquired original grass height survey data to form a data set of the priori data model.
[0076] The concatenation expression is as follows:
[0077] h a ={h i}∪{n h}
[0078] Step S204: Establishing an index expression of the original grass height survey data obtained in the grass height data set of the prior data model:
[0079] Specifically, for each data point in the acquired original grassland height survey data, the closest prior model data point is found to achieve the best data matching.
[0080] The index expression of the original grass height survey data obtained in the grass height data set of the prior data model is:
[0081] m i =argmin j c d -h i
[0082] m i is the index of the grass height dataset corresponding to the prior data model of the i-th original grass height survey data; c d is the value of the jth data point in the grass height dataset of the prior data model; |c d -h i| is the absolute difference between the data point in the jth prior model and the ith original grassland height survey data; argmin j Indicates that the expression |c is found d -h i |The smallest value of j.
[0083] Step S300: performing k-means clustering on the SAR coherence coefficient data, so that the label of each cluster in the k-means clustering is assigned to the two-dimensional image of the SAR coherence coefficient data, and associating the grass height survey data in the grass height data set with the label of the cluster in the two-dimensional image;
[0084] In some embodiments, performing k-means clustering on the SAR coherence coefficient data so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data includes:
[0085] Step 301: extracting effective coherence coefficients from the SAR coherence coefficient data, and performing k-means clustering on the effective coherence coefficients;
[0086] Specifically, the effective coherence coefficients in the SAR coherence coefficient data are extracted, and invalid (NaN) values are eliminated.
[0087] Convert the effective coherence coefficients from a two-dimensional format to a one-dimensional array suitable for K-Means clustering.
[0088] Initialize K-Means clustering and select the appropriate number of clusters. Through multiple experiments, the number of clusters can be set to 45, that is, 45 clusters.
[0089] Finally, the effective coherence coefficient is input into the K-Means model to perform K-Means clustering. The clustering expression is as follows:
[0090]
[0091] μ i is the centroid of the ith cluster, C i is the set of data points in the cluster, γ j is the data point within the cluster, which is the coherence coefficient of clustering, and J is the objective function, which is used to measure the compactness of the data points within the cluster.
[0092] Step 302: Obtain the average value of the effective coherence coefficient in each cluster in the k-means clustering;
[0093] Specifically, each cluster in the K-Means model is traversed, and the average value of the effective coherence coefficients of all data points in each cluster is calculated.
[0094] Preferably, the average expression of the effective coherence coefficient in each cluster in the K-Means model is as follows:
[0095]
[0096] C i represents the set of all data points belonging to the i-th cluster, γ i Represents the coherence coefficient data in the i-th cluster.
[0097] Step 303: Create a label for each cluster based on the average effective coherence coefficient within each cluster.
[0098] Specifically, the average effective coherence coefficient of each cluster in the K-Means model is used to sort the clusters from low to high for reallocating cluster labels.
[0099] A new label is created for each cluster using the previous sorting results. The labels of the new clusters are assigned according to the sorting results of the average effective coherence coefficient. The smaller the label value, the lower the average effective coherence coefficient.
[0100] After the effective coherence coefficient completes K-Means clustering, the reallocated cluster labels are mapped back to the two-dimensional image of the SAR coherence coefficient data.
[0101] In addition, associating the grass height survey data in the grass height dataset with the labels of the clusters in the two-dimensional image may specifically include:
[0102] Step 304: Using the mask previously extracted from the effective coherence coefficient data, each pixel point of the obtained two-dimensional image will have a valid cluster label value and an invalid background value.
[0103] The longitude and latitude of the grassland height survey data in the grassland height dataset are converted into image indexes, and then the grassland height survey data and the clustering results are correlated with each other, so as to facilitate the extraction of the corresponding cluster labels of the grassland height survey data in the two-dimensional image of the SAR coherence coefficient data in the clustering results.
[0104] Step S400: Obtain representative values of grass heights contained in all clusters in the k-means clustering, and interpolate the grass height representative values corresponding to all clusters to obtain the first grass height G of the area to be detected. k .
[0105] In some embodiments, in step S400, obtaining representative values of grass heights contained in all clusters in the k-means clustering includes the following two cases:
[0106] For the first cluster containing grass height survey data, an average value of all grass height survey data in the first cluster is obtained as a representative value of grass height of the first cluster to reflect the overall level of grass height in the cluster.
[0107] For the second cluster that does not contain grass height survey data, the grass height representative value c of the second cluster is obtained based on the second empirical model. m_h .
[0108] Among them, the second empirical model is:
[0109]
[0110] a and b are empirical constants. Through experimental fitting, the value of a is 1.1, the value of b is 2.5, and median(γ i-c ) represents the median of the coherence coefficient in the cluster.
[0111] In some embodiments, in step S400, obtaining the first grass height of the area to be detected by interpolation method includes:
[0112] Step 401: Generate a mask for each cluster in the k-means clustering and obtain the coordinates of the pixels of each cluster in the two-dimensional image of the SAR coherence coefficient data;
[0113] Step 402: Interpolate the pixels in each cluster based on the natural domain interpolation method to obtain the grass height value of each pixel, so as to obtain the first grass height G of the area to be detected. k .
[0114] The grassland height inversion method of this embodiment proposes a method based on coherence coefficient and model fusion, which solves the problem that the SAR coherence coefficient is difficult to directly associate with the grassland height. At the same time, it improves the efficiency of inverting grass height over a large area using a small amount of field grass height survey data, and to a certain extent improves the accuracy of grass height inversion.
[0115] In some embodiments, the grass height inversion method further includes:
[0116] Step 500: construct a first empirical model based on the effective coherence coefficient in the SAR coherence coefficient data and use the first empirical model to obtain a second grass height G h ;
[0117] Step 501: Based on the second grass height G h Correct the first grass height G k , to get the third grass height G f .
[0118] This embodiment uses an empirical model constructed by using effective coherence coefficients to correct the first grassland height, which can further improve the accuracy of grassland height inversion.
[0119] Furthermore, in some embodiments, the process of constructing the first empirical model includes:
[0120] The effective coherence coefficient data is subjected to a logarithmic square root transformation, and the first empirical model is constructed through a linear transformation relationship.
[0121] Among them, the expression of logarithmic square root transformation of effective coherence coefficient data is as follows:
[0122]
[0123] γ i is the effective coherence coefficient.
[0124] Using NSL, the second grass height G is obtained by the linear transformation formula h , the specific expression is as follows:
[0125]
[0126] α and β are empirical constants determined by experimental fitting, where α is the offset and β is the scaling factor.
[0127] In some embodiments, based on the second grass height G h Correct the first grass height G k ,include:
[0128] The first grass height G k The second grass height G h Weighted average to get the third grass height G f .
[0129] The weighted average expression is as follows:
[0130] G f =w1·G k +w2·G h
[0131] The weights w1 and w2 are determined through experiments, and their weight ratios are 0.5 respectively.
[0132] like Figure 7 As shown, the embodiment of the present application further provides a grassland height inversion device based on coherence coefficient and model fusion, and the grassland height inversion device includes:
[0133] The acquisition module is used to acquire SAR coherence coefficient data and grassland height survey data of the area to be detected.
[0134] A building module is used to build a priori data model based on grass height survey data to obtain a grass height dataset of the priori model.
[0135] The clustering module is used to perform k-means clustering on the SAR coherence coefficient data, assign the label of each cluster in the k-means clustering to the two-dimensional image of the SAR coherence coefficient data, and associate the grass height survey data in the grass height data set with the label of the cluster in the two-dimensional image.
[0136] The interpolation module is used to obtain the representative values of the grass height contained in the cluster in the k-means clustering, and interpolate the representative values of the grass height corresponding to all clusters to obtain the first grass height of the area to be detected.
[0137] In some embodiments, the grass height inversion device further comprises:
[0138] A correction module is used to construct a first empirical model based on the effective coherence coefficient in the SAR coherence coefficient data, and obtain the second grass height G using the first empirical model. h , and then correct the first grass height G based on the second grass height k , to get the third grass height G f .
[0139] The embodiment of the present application uses a specific experiment to demonstrate the grassland height inversion method based on coherence coefficient and model fusion of the present application.
[0140] Among them, the microcomputer used in the experiment has an Intel 9th generation i5 CPU and 4GB memory, and the programming platform is Visual Studio Code. The field survey data used in the experiment are 53 representative areas selected from different geographical locations of the experimental data set of Xiwuzhumuqin Banner, Xilin Gol League, Inner Mongolia Autonomous Region, China. The field survey data covers multiple levels of grassland height, and the collection size of the field survey data is 0.5*0.5㎡. The SAR data used in the experiment is Sentinel-1A data in strip mode of two-scene C-band VV polarization, with a spatial resolution of 5*20m in IW mode. The selected time is July 31, 2023-August 24, 2023, with a time interval of 24 days.
[0141] Firstly, the two-scene SAR data are preprocessed and interferometrically processed with the DEM data. The interference pattern is deflating and filtered, and then the coherence calculation is performed to obtain the SAR coherence coefficient data.
[0142] Secondly, the grassland height survey data was enhanced and a priori model was established. Then, K-Means clustering was performed using the SAR coherence coefficient data and the data set of the priori model. The representative value of the grassland height in each cluster was used to obtain the preliminary grassland height value through the natural field interpolation method.
[0143] Finally, the coherence coefficient is used to establish an empirical model, and the empirical model is used to correct the preliminary grassland height inversion results to obtain the final grassland height inversion data.
[0144] Figure 2 The figure shows a data comparison of the prior model established by using the grassland height survey data through data enhancement. The enhanced grassland height survey data is obviously more diverse and robust than the original grassland height survey data. Figure 3 It shows the number of effective coherence coefficients classified into each cluster after K-Means clustering; Figure 4 It shows that after K-Means clustering, the coherence coefficient in each cluster is mapped back to the two-dimensional image according to different clusters, and a clearly hierarchical cluster visualization image is obtained; Figure 5 The relationship between the coherence coefficient and the grass height survey data is shown.
[0145] Since there is an obvious nonlinear relationship between the coherence coefficient and the measured grass height data, the accuracy of grass height inversion directly through a simple correlation expression between the two is limited. Therefore, the embodiment of the present application adopts an existing empirical model of coherence coefficient and measured grass height to improve the accuracy of the inverted grass height value, rather than inverting the grass height through a direct correlation expression between the two. Figure 6 The final inversion of the grass height value and the field survey data for error fitting analysis, the determination coefficient R 2 It is 0.59, and the mean absolute error is 0.03m.
[0146] The above description is intended to be illustrative rather than restrictive, and those skilled in the art may make changes, modifications, substitutions and variations to the above embodiments within the scope of the present disclosure. Moreover, the above examples (or one or more of them) may be used in combination with each other, and it is contemplated that these embodiments may be combined with each other in various combinations or permutations.
Claims
1. A grassland height inversion method based on coherence coefficient and model fusion, characterized in that: The process includes the following: Obtain SAR coherence coefficient data and grass height survey data of the area to be detected; Constructing a priori data model based on the grassland height survey data to obtain a grassland height data set of the priori model; Performing k-means clustering on the SAR coherence coefficient data so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data, and associating the grass height survey data in the grass height dataset with the label of the cluster in the two-dimensional image; The representative values of grass heights contained in all clusters in the k-means clustering are obtained, and the representative values of grass heights corresponding to all clusters are interpolated to obtain a first grass height of the area to be detected.
2. The grassland height inversion method based on coherence coefficient and model fusion according to claim 1, characterized in that: Also includes: constructing a first empirical model based on the effective coherence coefficient in the SAR coherence coefficient data and obtaining a second grass height using the first empirical model; The first grass height is corrected based on the second grass height to obtain a third grass height.
3. The grassland height inversion method based on coherence coefficient and model fusion as claimed in claim 2, characterized in that: The process of constructing the first empirical model includes: The effective coherence coefficient data is subjected to a logarithmic square root transformation, and the first empirical model is constructed through a linear transformation relationship.
4. The grassland height inversion method based on coherence coefficient and model fusion as claimed in claim 2, characterized in that: Correcting the first grass height based on the second grass height includes: The first grass height and the second grass height are weighted averaged to obtain a third grass height.
5. The grassland height inversion method based on coherence coefficient and model fusion according to claim 1, characterized in that: Building a priori data model based on the grass height survey data to obtain a grass height data set of the priori data model includes: Sampling the acquired grassland height survey data; Add noise to the sampled grassland height survey data; The grass height survey data with added noise is spliced with the acquired original grass height survey data to form a grass height data set of the prior data model.
6. The grassland height inversion method based on coherence coefficient and model fusion as claimed in claim 5, characterized in that: Also includes: The index expression of the grass height dataset obtained from the original grass height survey data in the prior data model is established: m i =argmin j c d -h i | m i is the index of the grass height dataset corresponding to the prior data model of the i-th original grass height survey data; c d is the value of the jth data point in the grass height dataset of the prior data model; |c d -h i | is the absolute difference between the jth prior model data point and the ith original grassland height survey data; argmin j Indicates that the expression |c is found d -h i |The smallest value of j.
7. The grassland height inversion method based on coherence coefficient and model fusion according to claim 2, characterized in that: Performing k-means clustering on the SAR coherence coefficient data so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data, comprising: Extracting effective coherence coefficients from the SAR coherence coefficient data, and performing k-means clustering on the effective coherence coefficients; Get the average value of the effective coherence coefficient within each cluster in k-means clustering; Create labels for each cluster based on the average effective coherence coefficient within each cluster.
8. The grassland height inversion method based on coherence coefficient and model fusion according to claim 7, characterized in that: The first grass height of the area to be detected is obtained by interpolation method, including: Generate a mask for each cluster in the k-means clustering and obtain the coordinates of the pixels of each cluster in the two-dimensional image of the SAR coherence coefficient data; The pixels in each cluster are interpolated based on the natural field interpolation method to obtain the grass height value of each pixel, so as to obtain the first grass height of the area to be detected.
9. The grassland height inversion method based on coherence coefficient and model fusion according to claim 2, characterized in that: Get the representative values of grass height in all clusters in k-means clustering, including: For a first cluster including grass height survey data, obtaining an average value of all grass height survey data in the first cluster as a representative value of grass height of the first cluster; For the second cluster that does not contain grass height survey data, the grass height representative value c of the second cluster is obtained based on the second empirical model. m_h ; The second empirical model is: a and b are empirical constants. Through experimental fitting, the value of a is 1.1, the value of b is 2.5, and median(γ i-c ) represents the median of the coherence coefficient in the cluster.
10. A grassland height inversion device based on coherence coefficient and model fusion, characterized in that: include: An acquisition module is used to acquire SAR coherence coefficient data and grass height survey data of the area to be detected; A construction module, used to construct a priori data model based on the grass height survey data to obtain a grass height data set of the priori model; A clustering module, configured to perform k-means clustering on the SAR coherence coefficient data, so that a label of each cluster in the k-means clustering is assigned to a two-dimensional image of the SAR coherence coefficient data, and associate the grass height survey data in the grass height data set with the label of the cluster in the two-dimensional image; The interpolation module is used to obtain the representative values of grass height contained in all clusters in the k-means clustering, and interpolate the representative values of grass height corresponding to all clusters to obtain the first grass height of the area to be detected.