Mangrove forest biomass estimation method and system based on laser radar

Abstract

The invention provides a mangrove forest biomass estimation method and system based on a laser radar, and relates to the technical field of laser radars, and the method comprises the steps: carrying out the noise filtering and normalization processing of an unmanned plane laser radar point cloud; mangrove forest information in the unmanned aerial vehicle laser radar point cloud is extracted; carrying out sample screening and dimension reduction processing on the GEDI L4A satellite-borne laser radar light spot data; the Sentinel-2 surface reflectance data are processed, and mangrove forest reflectance characteristics are extracted; based on GEDI L4A satellite-borne laser radar light spot data, unmanned aerial vehicle laser radar point cloud information and Sentinel-2 surface reflectance data mangrove forest information are extracted; constructing a stack integrated nonlinear coupling model, and establishing a regression relationship; the regression relation is popularized to the whole mangrove forest research area, and a high-precision and continuous mangrove forest biomass result is obtained through estimation. According to the method, the accuracy and the large-range applicability of biomass estimation are improved.

Description

Technical Field

[0001] The present invention relates to the field of laser radar technology, and in particular to a method and system for estimating mangrove biomass based on laser radar. Background Art

[0002] Mangroves are located in the shallow mudflats at the junction of land and sea. They are a type of plant community that can reduce wave erosion and provide key habitats, reproduction and shelter for coastal organisms. In addition, mangroves have a strong ability to adapt to intertidal mudflat environments and play an unparalleled role in purifying seawater and preventing land pollutants from entering the ocean. As a typical coastal blue carbon ecosystem, mangroves have a high carbon sequestration capacity and are typical coastal wetland "blue carbon", playing a key role in the global carbon cycle. Although the biomass of mangroves in coastal vegetation accounts for only 0.05% of the global total, the carbon they store accounts for more than 50% of the carbon reserves in marine sediments.

[0003] However, the sensitivity and vulnerability of mangrove ecosystems to global changes have caused this high-biodiversity, high-economic-value ecosystem to degrade dramatically in the past few decades. Human activities such as tidal flat landfill, forest development and overfishing have directly damaged mangroves; ecological and environmental changes, such as climate change and sea level rise, have also had a significant impact on the growth environment of mangroves; in addition, the threat of invasive species and alien species may cause the degradation or loss of ecological functions of mangroves. These problems have jointly led to a significant reduction in the area of ​​mangroves. Due to the ecological and economic value of mangroves, the protection of mangroves has received increasing attention. Biomass is a key indicator for assessing the ecological environment and carbon storage of forests, which helps to deeply understand the existing volume and carbon storage of mangrove ecosystems. In order to better understand and protect mangroves, it is crucial to reasonably and effectively estimate the biomass of mangroves.

[0004] With the continuous development of sensing technology, the methods of information acquisition have become diversified. As an active remote sensing, LiDAR has the advantages of high resolution, accuracy and strong anti-interference ability. The emitted laser pulses can penetrate mangroves and have strong capabilities in detecting the vertical structure of trees and terrain surfaces, so LiDAR is widely used in forestry surveys. UAV LiDAR is a laser detection and ranging system installed on drones. Its advantages are that it can flexibly adjust the route and observation time, high data accuracy, and can quickly obtain a large amount of mangrove data. However, the equipment and operating costs of UAV LiDAR are high, and the single working time is limited. It can only be used to invert the mangrove biomass in a small area. Satellite LiDAR is a LiDAR system installed on a satellite, which is used to measure the elevation and structure of the earth's surface by emitting laser pulses and recording their reflection time. Although the accuracy of satellite LiDAR in obtaining vertical information of mangroves is slightly lower than that of UAV LiDAR, its advantage of large-scale coverage makes it possible to estimate the biomass of mangroves in large areas.

[0005] By using satellite platforms to provide rich surface detection information and fusing lidar with surface reflection information, mangrove biomass can be estimated more accurately and continuously. However, the traditional multi-data coupling method still has some limitations, such as insufficient response of data to complex environmental changes and the need to improve the accuracy of the model. This is especially true for mangroves with complex vertical structures, heterogeneous horizontal structures, and diverse spectral properties. Therefore, a more efficient and accurate multi-data coupling method is urgently needed to improve the accuracy and efficiency of mangrove biomass estimation, so as to address the limitations of existing technologies and enhance the scientific basis for mangrove protection and management. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a mangrove biomass estimation method and system based on lidar, which realizes mangrove biomass estimation based on the combination of lidar and spectral reflectance information to improve the accuracy of biomass estimation and its wide range applicability.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] A method for estimating mangrove biomass based on laser radar comprises the following steps:

[0009] Step S1, performing noise filtering and normalization processing on the UAV laser radar point cloud;

[0010] Step S2, extracting mangrove height information, point cloud density information and echo intensity information from the UAV laser radar point cloud;

[0011] Step S3, performing sample screening and dimensionality reduction processing on the GEDI L4A satellite-borne laser radar spot data;

[0012] Step S4, performing cloud-free synthesis and multi-time series mosaic processing on the Sentinel-2 surface reflectance data, and extracting the reflectance characteristics of mangroves;

[0013] Step S5, extracting UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data;

[0014] Step S6, constructing a stack integrated nonlinear coupling model, and establishing a regression relationship between the GEDI L4A satellite-borne laser radar spot data and a surface feature set; the surface feature set includes the mangrove height information, point cloud density information and echo intensity information in the UAV laser radar point cloud obtained in step S2 and the mangrove reflectivity characteristics obtained in step S4;

[0015] Step S7: Extend the regression relationship to the entire mangrove study area to estimate high-precision and continuous mangrove biomass results.

[0016] The present invention also provides a mangrove biomass estimation system based on laser radar, comprising the following modules:

[0017] Point cloud processing module, used to perform noise filtering and normalization on the UAV lidar point cloud;

[0018] Mangrove information extraction module, used to extract mangrove height information, point cloud density information and echo intensity information from the UAV lidar point cloud;

[0019] The spot data processing module is used to perform sample screening and dimension reduction processing on the GEDI L4A satellite-borne laser radar spot data;

[0020] Mangrove reflectance feature extraction module, used to perform cloud-free synthesis and multi-time series mosaic processing on Sentinel-2 surface reflectance data, and extract mangrove reflectance features;

[0021] Comprehensive information extraction module, used to extract UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data;

[0022] A surface feature set construction module is used to construct a stack integrated nonlinear coupling model to establish a regression relationship between the GEDI L4A satellite-borne laser radar spot data and the surface feature set; the surface feature set includes mangrove height information, point cloud density information, echo intensity information and mangrove reflectivity characteristics in the UAV laser radar point cloud;

[0023] The estimation module is used to generalize the regression relationship to the entire mangrove study area and estimate high-precision and continuous mangrove biomass results.

[0024] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the above-mentioned method for estimating mangrove biomass based on laser radar are implemented.

[0025] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for estimating mangrove biomass based on laser radar.

[0026] Beneficial effects:

[0027] Compared with other landforms, the most prominent features of mangroves are the complexity of vertical structure, the heterogeneity of horizontal structure and the diversity of spectral properties. LiDAR can provide direct and accurate measurement data of forest vertical structure. High-precision mangrove spatial feature extraction technology can greatly reduce the random positioning error of sensors introduced in the multi-data coupling process. However, the coupling of LiDAR and spectral reflectance information has high complexity and uncertainty. The present invention proposes a LiDAR spot and multi-data collaborative registration technology, which expands the potential shooting range of LiDAR spots, improves the spatial indexing effect of LiDAR and multi-data features, greatly improves the model's ability to explain biomass variability, and reduces estimation uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a flow chart of a method for estimating mangrove biomass based on laser radar of the present invention;

[0029] Figure 2 Schematic diagram of 2D Gaussian spatial weighting method; A is a schematic diagram of the calculation of the potential first shooting range of the spot with only lidar data, and B is a schematic diagram of the calculation of the potential second shooting range of the spot with the coordination of lidar and Sentinel-2 surface reflectivity data;

[0030] Figure 3a This is a comparison chart of the estimation accuracy without involving the biomass results in step 5;

[0031] Figure 3b This is a comparison chart of the estimation accuracy of the biomass results in step 5. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0033] In the description of the present invention, it should be understood that the terms "upper", "lower", "top", "bottom", "inner", "outer", etc. used in the present invention to indicate the orientation or position relationship is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0034] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] The present invention first obtains the mangrove research area Within the time range arrive UAV lidar point cloud data, Sentinel-2 surface reflectivity data, GEDI L4A satellite-borne lidar spot data, among which, The starting time for biomass estimation, is the end time of biomass estimation, where Y, D, and M represent year, month, and day respectively.

[0036] like Figure 1 As shown, a method for estimating mangrove biomass based on laser radar of the present invention comprises the following steps:

[0037] Step S1, performing noise filtering and normalization processing on the UAV laser radar point cloud;

[0038] Step S2, extracting mangrove height information, point cloud density information and echo intensity information from the UAV laser radar point cloud;

[0039] Step S3, performing sample screening and dimensionality reduction processing on the GEDI L4A satellite-borne laser radar spot data;

[0040] Step S4, performing cloud-free synthesis and multi-time series mosaic processing on the Sentinel-2 surface reflectance data, and extracting the reflectance characteristics of mangroves;

[0041] Step S5, extracting UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data to obtain feature set F;

[0042] Step S6, constructing a stack integrated nonlinear coupling model, and establishing a regression relationship between the GEDI L4A satellite-borne laser radar spot data and a surface feature set; the surface feature set includes the mangrove height information, point cloud density information and echo intensity information in the UAV laser radar point cloud obtained in step S2 and the mangrove reflectivity characteristics obtained in step S4;

[0043] Step S7: Extend the regression relationship to the entire mangrove study area to estimate high-precision and continuous mangrove biomass results.

[0044] Specifically, the step S1 includes:

[0045] The point cloud data was cut according to the mangrove study area, and then Grid divided into Block point cloud file, extract the The corresponding value of the 95th percentile of the height within the block point cloud The corresponding value for the height percentile 5 , remove point clouds with height attributes higher than or below Noise points. Then follow Square will Block point cloud segmentation Block point cloud file, Block point cloud file The corresponding value of the height percentile of 1 within the block point cloud extraction range , will The corresponding height of all points in the block point cloud minus the corresponding value , complete the Block point cloud Normalization of block point cloud files. This step is repeated until all The denoising process of the point cloud file After the point cloud normalization process ends, usually .

[0046] Specifically, the step S2 includes:

[0047] The denoised and normalized UAV lidar point cloud is Grid divided into Block point cloud files, then statistics Block point cloud file The mangrove height information, point cloud density information and echo intensity information in the block point cloud are completed. Extract point cloud information from block point cloud files.

[0048] Among them, the mangrove height information includes the maximum elevation , minimum elevation ,average value , standard deviation , the corresponding value of the height percentile 5 , the corresponding value of the height percentile 10 , the corresponding value of the height percentile is 90 , the corresponding value of the height percentile is 95 ; Point cloud density information includes vertical distribution at a height of 1 meter , vertical distribution at a height of 2 meters , vertical distribution at a height of 5 meters ; Echo strength information including first echo percentage , the last echo percentage .

[0049] ;

[0050] ;

[0051] ;

[0052] in, Representative The total number of photons in the point cloud with a height greater than 1 meter, Representative The total number of photons in the point cloud with a height greater than 2 meters, Represents the total number of photons at a height greater than 5 meters, Representative The total number of all photons in the block point cloud.

[0053] ;

[0054] ;

[0055] in, represents the total number of photons in the first echo, represents the total number of photons in the last first echo, Representative The total number of all photons in the block point cloud.

[0056] Specifically, step S3 includes:

[0057] The GEDI L4A satellite-borne laser radar spot data is processed by information dimensionality reduction, and only the four strong beam detection data of "BEAM0101", "BEAM0110", "BEAM1000" and "BEAM1011" are selected. The corresponding data of each beam only retains the following information: detection time , a2 algorithm group latitude , a2 algorithm group longitude , a2 algorithm group biomass , biomass error of a2 algorithm group , a2 algorithm group spot quality mark , a2 algorithm group detection sensitivity , light spot geolocation mark , algorithm running flag . Then only high-quality lidar spots that meet the following conditions are retained: , , , , , , Located between 8pm and 4am. The a2 algorithm group is the content of GEDI L4A data.

[0058] Specifically, step S4 includes:

[0059] Based on the point cloud data of the mangrove study area, the surface reflectance in the blue spectrum of the Sentinel-2 data was selected. , green surface reflectivity , red surface reflectivity , the surface reflectivity of the first red edge e1, surface reflectivity of the second red edge e2, Surface reflectivity of the third red edge e3. Near-infrared surface reflectivity , the surface reflectivity of the 4th red edge e4. Surface reflectivity of the first short-wave infrared , 2nd shortwave infrared surface reflectivity , and cloud mask bands . Perform cloud-free synthesis processing on multi-view Sentinel-2 surface reflectance data to ensure that the cloud cover of a single view is less than 20% ( 20%), and the median method was used to synthesize and mosaic all overlapping data in the study area.

[0060] Subsequently, the following mangrove reflectance features were extracted: Vegetation enhancement features , water body enhancement features .

[0061] ;

[0062] in, represents the surface reflectance in the near infrared, represents the red surface reflectance, and Information comes from each light spot latitude With longitude The corresponding surface reflectivity.

[0063] ;

[0064] in, represents the green surface reflectance, represents the surface reflectance in the near infrared, and Information comes from each light spot latitude With longitude The corresponding surface reflectivity.

[0065] Specifically, step S5 includes:

[0066] In the general use of GEDI data, it is assumed that the GEDI L4A product geolocation information is biased, but in fact the positioning error of satellite platform data (such as GEDI and Sentinel-2 data) cannot be ignored. This step specifically involves a spatial sensing matching method, which expands the potential surface shooting range of the GEDI L4A lidar spot, and uses a 2D Gaussian spatial weighting method to count the UAV lidar point cloud mangrove features and Sentinel-2 surface reflectivity mangrove features that fall within the potential surface shooting range of the spot. The center positioning of each spot is obtained from the GEDI L4A product, including latitude and longitude. With longitude First, set the first potential shooting range radius , the second potential shooting range radius .

[0067] ;

[0068] ;

[0069] in, (like Figure 2 A) is the radius of the potential first shooting range of the light spot, (like Figure 2 (B) is the radius of the potential second shooting range of the determined spot. The average ground radius of the GEDI laser radar spot is , the spot center deviation is , the spatial deviation of Sentinel-2 surface reflectance is . Figure 2 In A, 10.3, 12.5, 60, 22.8, and GEDI represent the laser radar spot deviation, the average spot radius, the average interval between the spot centers, the radius of the potential first shooting range of the spot, and the GEDI laser radar spot example, respectively; Figure 2 In B, 10.3, 28.4, 15.9, and 30 respectively represent the lidar spot deviation, the radius of the potential second shooting range of the spot, the sum of the lidar spot deviation and the spatial deviation of the Sentinel-2 surface reflectivity, and an example of the resolution of the Sentinel-2 surface reflectivity data magnified three times.

[0070] Select The geographical location corresponding to each spot is the latitude With longitude , statistics based on latitude With longitude As the central location, is the Sentinel-2 surface reflectance mangrove characteristic value within the radius The feature is . By The light spot intersects with the Sentinel-2 surface reflectance mangrove feature The intersection feature points are calculated by 2D Gaussian space weighting method, where: , and Corresponding to In the range The eigenvalue, area, and Gaussian weight values ​​corresponding to the intersection point. The area is directly derived from the resolution of the Sentinel-2 surface reflectance product. The light spot Gaussian weights corresponding to the shooting ranges of the features With statistical value It is derived from the following formula:

[0071] ;

[0072] ;

[0073] in, By The latitude and longitude of the intersection feature point (latitude With longitude ) is normalized, exp means exponential meaning, for By The total number of intersection feature points within the intersection range between the light spots and the Sentinel-2 surface reflectance mangrove feature, for Of the intersecting feature points intersection feature points, For the The concentration in the latitude direction of the intersection point, For the The concentration of the intersection points in the longitude direction is generally .

[0074] ;

[0075] ;

[0076] in, Representative The center latitude values ​​of the intersecting feature points, Representative The center longitude value of the intersection feature point.

[0077] Select the The geographical location corresponding to each spot is the latitude With longitude , in latitude With longitude As the central location, is the first mangrove feature value of the drone lidar point cloud within the radius The feature is . By The light spots intersect with the mangrove features of the drone lidar point cloud The intersection feature points are calculated by 2D Gaussian space weighting method, where: , and Corresponding to In the range The eigenvalues, areas and Gaussian weights corresponding to the intersection points. The area is directly obtained from the mangrove feature resolution of the drone lidar point cloud. The light spot Gaussian weights corresponding to the shooting ranges of the features With statistical value It is derived from the following formula:

[0078] ;

[0079] ;

[0080] in, By The latitude and longitude of the intersection feature point (latitude With longitude ) is normalized, exp means exponential meaning, for By The total number of intersection feature points within the intersection range between the light spots and the mangrove features of the drone lidar point cloud, for Of the intersecting feature points intersection feature points, For the The concentration in the latitude direction of the intersection point, For the The concentration of the intersection points in the longitude direction is generally .

[0081] ;

[0082] ;

[0083] in, Representative The center latitude values ​​of the intersecting feature points, Representative The center longitude value of the intersection feature point.

[0084] Feature Set Mangrove features of 13 UAV lidar point clouds obtained in step S2 The mangrove characteristics of the 12 Sentinel-2 surface reflectances obtained in step S4 According to the GEDI L4A spot center geolocation (latitude With longitude This step ensures that the information of each light spot is correctly counted within the potential shooting range.

[0085] Specifically, step S6 includes:

[0086] Constructing a stacked integrated nonlinear coupling model to establish a regression relationship between the lidar and the surface reflection information mainly includes the following steps: Randomly extract the feature set corresponding to the GEDI L4A lidar spot This method uses a stack prediction model The stack prediction model needs to integrate multiple sub-machine learning models with excellent performance (such as good accuracy, stability, etc.) to improve the overall model performance. X is obtained from the feature set F in step 5, and Y is obtained from the biomass in the GEDIL4A data. The training set A is a set of X and Y according to the latitude lat and longitude lon of the center of the GEDI L4A spot.

[0087] Before training the model, you need to set the following parameters: An ensemble of machine learning models ; Set the number of repetitions ; Set the number of cross validation ; Set the number of model stacking layers .exist Each model is loaded into a stacking layer after training. Therefore, before model training, you need to select the stacking layer that has not been loaded with the model to be trained. , as the stacking layer for integrating this training model; start Repeat the training; select the training set The jth feature data block in With the Biomass data-block Participate in model training; in the collection Select the model that has not participated in the training to start training and calculate the cross-validation results of the model ; Until the collection is traversed All model types and N repeated tests are performed. Estimated biomass value calculated after repeated training , and the training model and model weights Load into stacking layers, load all the stacking layers to get the stacking prediction model.

[0088] = ;

[0089] ;

[0090] Among them, the intersection verification result From this formula, we can conclude that The biomass of the GEDI L4A training set; is the biomass of the GEDI L4A test set; n is the number of samples.

[0091] The present invention also provides a mangrove biomass estimation system based on laser radar, comprising the following modules:

[0092] Point cloud processing module, used to perform noise filtering and normalization on the UAV lidar point cloud;

[0093] Mangrove information extraction module, used to extract mangrove height information, point cloud density information and echo intensity information from the UAV lidar point cloud;

[0094] The spot data processing module is used to perform sample screening and dimension reduction processing on the GEDI L4A satellite-borne laser radar spot data;

[0095] Mangrove reflectance feature extraction module, used to perform cloud-free synthesis and multi-time series mosaic processing on Sentinel-2 surface reflectance data, and extract mangrove reflectance features;

[0096] Comprehensive information extraction module, used to extract UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data;

[0097] A surface feature set construction module is used to construct a stack integrated nonlinear coupling model to establish a regression relationship between the GEDI L4A satellite-borne laser radar spot data and the surface feature set; the surface feature set includes mangrove height information, point cloud density information, echo intensity information and mangrove reflectivity characteristics in the UAV laser radar point cloud;

[0098] The estimation module is used to generalize the regression relationship to the entire mangrove study area and estimate high-precision and continuous mangrove biomass results.

[0099] Example:

[0100] The biomass estimation of a mangrove study area A includes the following steps:

[0101] Step S1: In this embodiment, the drone laser radar data set should be in standard las format or laz format, and retain the laser radar point cloud height information, density information, and echo information. In some embodiments, is 100, is 1. In other embodiments, The setting range can be between 50-100. The setting range can be between 0.5-2;

[0102] Step S2: is 2. In other embodiments, The setting range can be between 1-5;

[0103] Step S3, performing sample screening and dimensionality reduction processing on the GEDI L4A satellite-borne laser radar spot data;

[0104] Step S4, performing cloud-free synthesis and multi-time series mosaic processing on the Sentinel-2 surface reflectance data, and extracting the reflectance characteristics of mangroves;

[0105] Step S5: In this embodiment, the average ground radius of the GEDI laser radar spot is 12.5, the deviation of the center of the spot The spatial deviation of the surface reflectance of Sentinel-2 is 10.3. is 4.3;

[0106] Step S6: In this embodiment, C is 5, and the set Including distributed random forest model (DRF), gradient boosting model (GBM), generalized linear model (GLM), extreme random forest model (XRT), extreme gradient boosting model (XGB), N is 3, L is 5. In other embodiments, C, M, N, L are not limited to the above settings, and the speed and accuracy of model training need to be comprehensively considered when setting the parameters.

[0107] Step S7: Obtain the entire feature set of mangrove features from the drone lidar point cloud and the Sentinel-2 surface reflectivity mangrove features , and finally the estimated study area The biomass within. Figure 3a This is a comparison chart of the estimation accuracy without the biomass results in step 5. The coefficient of determination is is 0.11, and RMSE is 140Mg / ha; Figure 3b This is a comparison chart of the estimation accuracy of the biomass results in step 5. The value of the laser radar spot and multi-data collaborative registration technology of the present invention is 0.61, and the RMSE is 93Mg / ha. This indicates that the model's ability to explain biomass variability has been greatly improved ( Mg / ha) and reduced estimate uncertainty (RMSE decreased by 47 Mg / ha).

[0108] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.

[0109] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0110] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0111] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0112] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0113] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for estimating mangrove biomass based on laser radar, characterized in that: The steps include: Step S1, performing noise filtering and normalization processing on the UAV laser radar point cloud; Step S2, extracting mangrove height information, point cloud density information and echo intensity information from the UAV laser radar point cloud; Step S3, performing sample screening and dimensionality reduction processing on the GEDI L4A satellite-borne laser radar spot data; Step S4, performing cloud-free synthesis and multi-time series mosaic processing on the Sentinel-2 surface reflectance data, and extracting the reflectance characteristics of mangroves; Step S5, extracting UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data to obtain feature set F; Step S6, constructing a stack integrated nonlinear coupling model, and establishing a regression relationship between the GEDI L4A satellite-borne laser radar spot data and a surface feature set; the surface feature set includes the mangrove height information, point cloud density information and echo intensity information in the UAV laser radar point cloud obtained in step S2 and the mangrove reflectivity characteristics obtained in step S4; Step S7: Extend the regression relationship to the entire mangrove study area to estimate high-precision and continuous mangrove biomass results.

2. The method for estimating mangrove biomass based on laser radar according to claim 1, characterized in that: The step S1 comprises: According to the point cloud data of the mangrove study area, Grid divided into Block point cloud file, extract the The corresponding value of the 95th percentile of the height within the block point cloud The corresponding value for the height percentile 5 , remove point clouds with height attributes higher than or below Noise points; then according to Square will Block point cloud segmentation Block point cloud file, Block point cloud file The corresponding value of the height percentile of 1 within the block point cloud extraction range , will The corresponding height of all points in the block point cloud minus the corresponding value , complete the Block point cloud Normalization of block point cloud files; traversal The denoising process of the point cloud file After the point cloud normalization process ends, .

3. The method for estimating mangrove biomass based on laser radar according to claim 1, characterized in that: The step S2 comprises: The denoised and normalized UAV lidar point cloud is Grid divided into Block point cloud files, then statistics Block point cloud file The mangrove height information, point cloud density information and echo intensity information in the block point cloud are completed. Extract point cloud information from block point cloud files.

4. The method for estimating mangrove biomass based on laser radar according to claim 1, characterized in that: The step S3 comprises: Only the four strong beam detection data of "BEAM0101", "BEAM0110", "BEAM1000" and "BEAM1011" are selected, and then only the high-quality lidar spots that meet the following conditions are retained: , , , , , The detection time is between 8pm in the evening and 4am in the morning; It is the geographical positioning mark of the light spot. is the detection sensitivity of the a2 algorithm group, It is the flag bit for algorithm operation. is the biomass of the a2 algorithm group, is the biomass error of the a2 algorithm group, It is the spot quality mark of the a2 algorithm group.

5. The method for estimating mangrove biomass based on laser radar according to claim 1, characterized in that: The step S4 comprises: Based on the point cloud data of the mangrove study area, the surface reflectance in the blue spectrum of the Sentinel-2 data was selected. , green surface reflectivity , red surface reflectivity , the surface reflectivity of the first red edge e1, surface reflectivity of the second red edge e2, Surface reflectivity of the third red edge e3. Near-infrared surface reflectivity , the surface reflectivity of the 4th red edge e4. Surface reflectivity of the first short-wave infrared , 2nd short-wave infrared surface reflectivity , and cloud mask bands ; Perform cloud-free synthesis processing on multi-view Sentinel-2 surface reflectance data to ensure that the cloud cover of a single view is less than 20%, and use the median method to synthesize and mosaic all overlapping data in the study area; Subsequently, the following mangrove reflectance features were extracted: Vegetation enhancement features , water body enhancement features ; ； in, represents the surface reflectance in the near infrared, represents the red surface reflectance, and Information comes from each light spot latitude With longitude The corresponding surface reflectivity; ； in, represents the green surface reflectance, represents the surface reflectance in the near infrared, and Information comes from each light spot latitude With longitude The corresponding surface reflectivity.

6. The method for estimating mangrove biomass based on laser radar according to claim 3, characterized in that: The step S5 adopts a spatial sensing matching method to expand the potential surface shooting range of the GEDI L4A laser radar spot, and adopts a 2D Gaussian spatial weighting method to count the mangrove features of the UAV laser radar point cloud and the Sentinel-2 surface reflectance mangrove features that fall within the potential surface shooting range of the spot; The center positioning of each spot is obtained from the GEDI L4A product, including latitude With longitude ; First set the radius of the potential first shooting range of the spot , the radius of the potential second shooting range of the light spot ; ； ； Among them, the average ground radius of the GEDI lidar spot is , the spot center deviation is , the spatial deviation of Sentinel-2 surface reflectance is ; Select The geographical location corresponding to each spot is the latitude With longitude , statistics based on latitude With longitude As the central location, is the Sentinel-2 surface reflectance mangrove characteristic value within the radius The feature is ; By The light spot intersects with the Sentinel-2 surface reflectance mangrove feature The intersection feature points are calculated by 2D Gaussian space weighting method, where: , and Corresponding to In the range The eigenvalues, areas, and Gaussian weights corresponding to the intersection points; the area is directly derived from the resolution of the Sentinel-2 surface reflectance product. The light spot Gaussian weights corresponding to the shooting ranges of the features With statistical value It is derived from the following formula: ； ； in, By Latitude of the intersection feature point With longitude Normalized, exp represents the exponential function, for By The total number of intersection feature points within the intersection range between the light spots and the Sentinel-2 surface reflectance mangrove feature, for Of the intersecting feature points intersection feature points, For the The concentration in the latitude direction of the intersection point, For the The concentration in the longitude direction of the intersection point, ; Feature Set Mangrove features of 13 UAV lidar point clouds obtained in step S2 The mangrove characteristics of the 12 Sentinel-2 surface reflectances obtained in step S4 According to the latitude of the center of the GEDI L4A spot With longitude Combination.

7. The method for estimating mangrove biomass based on laser radar according to claim 6, characterized in that: The step S6 comprises: Proportional Randomly extract the feature set corresponding to the GEDI L4A lidar spot , and The feature set from step 5 is The training set is composed of GEDI L4A biomass according to the geographical location of the GEDI L4A spot center. ; Build An ensemble of machine learning models ; Set the number of repetitions ; Set the number of cross validation ; Set the number of model stacking layers ,exist In the model stacking layers, each model is loaded into a stacking layer after training. Before model training, the stacking layer that has not been loaded with the model to be trained is selected. , as the stacking layer for integrating this training model; start Repeat the training; select the training set The jth feature data block in With the Biomass data-block Participate in model training; in the collection Select the model that has not participated in the training to start training and calculate the cross-validation results of the model ; Until the collection is traversed All model types and N repeated tests are completed; Estimated biomass value calculated after repeated training , and the training model and model weights Load into Stacking layers, load all stacking layers to get the stack prediction model; = ； ； Among them, the intersection verification result From this formula, we can conclude that The biomass of the GEDI L4A training set; is the biomass of the GEDI L4A test set; n is the number of samples.

8. A mangrove biomass estimation system based on laser radar, characterized in that: Includes the following modules: Point cloud processing module, used to perform noise filtering and normalization on the UAV lidar point cloud; Mangrove information extraction module, used to extract mangrove height information, point cloud density information and echo intensity information from the UAV lidar point cloud; The spot data processing module is used to perform sample screening and dimension reduction processing on the GEDI L4A satellite-borne laser radar spot data; Mangrove reflectance feature extraction module, used to perform cloud-free synthesis and multi-time series mosaic processing on Sentinel-2 surface reflectance data, and extract mangrove reflectance features; Comprehensive information extraction module, used to extract UAV lidar point cloud information and Sentinel-2 surface reflectivity data mangrove information based on GEDI L4A satellite-borne lidar spot data; A surface feature set construction module is used to construct a stack integrated nonlinear coupling model to establish a regression relationship between the GEDI L4A satellite-borne laser radar spot data and the surface feature set; the surface feature set includes mangrove height information, point cloud density information, echo intensity information and mangrove reflectivity characteristics in the UAV laser radar point cloud; The estimation module is used to generalize the regression relationship to the entire mangrove study area and estimate high-precision and continuous mangrove biomass results.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of a mangrove biomass estimation method based on laser radar are implemented as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for estimating mangrove biomass based on laser radar as described in any one of claims 1 to 7 are implemented.

Images