Method, device and equipment for estimating large-area medium spatial resolution time-series vegetation leaf area index by cooperating space-borne lidar and multi-spectral remote sensing, and medium

Abstract

The application discloses a large-area medium spatial resolution time-series vegetation leaf area index estimation method and device cooperating with a spaceborne laser radar and multispectral remote sensing, equipment and a medium, and relates to the field of vegetation structure parameter estimation. The method comprises the following steps: acquiring multispectral reflectance data and spaceborne laser radar data of a to-be-measured region in a to-be-measured period; constructing a multi-feature data set based on the multispectral reflectance data; matching the multi-feature data set and the spaceborne laser radar data in space and time to obtain matching data; inputting the matching data and the spaceborne laser radar data into an estimation model to obtain an LAI estimation result; the estimation model is obtained by training a deep neural network; and the estimation model is used to determine the mapping relationship between the multi-feature data set matched in space and time and the spaceborne laser radar data, and then determine the LAI estimation result. The application can improve the LAI estimation accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vegetation structure parameter estimation, and in particular to a method and device for estimating medium spatial resolution time-series vegetation leaf area index (LAI) in a large area by combining space-borne lidar and multi-spectral remote sensing, equipment and medium. BACKGROUND

[0002] The Sentinel-2 Level 2 Prototype Processor (SL2P) tool can achieve medium spatial resolution time-series leaf area index (LAI) inversion, but the accuracy is poor for forest areas. Leaf area index, which is half of the total leaf area per unit ground area, is an important vegetation structure parameter. From early one-dimensional radiation transfer models of vegetation canopy such as Suits and SAIL to the spectral invariant theory representing the development direction of forward models, leaf area index has always been the key to parameterizing complex radiation transfer processes. Forward simulation-backward inversion is a classic paradigm for obtaining leaf area index from satellite data. The MODIS LAI product, VIIRS LAI product and SL2P tool based on this paradigm have promoted the rapid development of fields such as carbon cycle, land-atmosphere interaction and vegetation growth monitoring.

[0003] Among them, the SL2P tool establishes a lookup table by combining PROSAIL forward simulation with a neural network, and combining the characteristics of the Sentinel-2A and Sentinel-2B dual-satellite networking with a 5-day revisit period, it can quickly achieve 20-meter spatial resolution time-series LAI inversion with a 5-day revisit period. However, many studies have pointed out that although the LAI produced by S2LP (hereinafter referred to as SL2P LAI product) can meet the accuracy requirements of applications in crop areas, the PROSAIL model used by SL2P is not suitable for spatially discontinuous forests, resulting in large errors in the tool in forest areas.

[0004] Global Ecosystem Dynamics Investigation (GEDI) has a 4-day orbital revisit period and has an accuracy advantage in forest LAI estimation, but it is spatially discontinuous, has a significant positional error, and there are quality differences between light spots.

[0005] GEDI is the first spaceborne large-footprint full-waveform LiDAR data for global forest structure monitoring. Its 25-meter diameter footprint has similar spatial scale as Sentinel-2. Its 4-day orbital revisit period has the potential to reflect the temporal dynamics of forest LAI. Its 60-meter footprint along-track spacing can provide high-density spatial sampling. Meanwhile, the direct observation capability of GEDI on vegetation canopy structure makes it effectively overcome the saturation effect generally existing in passive optical remote sensing data. The global-scale spot LAI product produced by GEDI based on the GORT model has an accuracy advantage over existing passive optical-based inversion results in forest areas. However, GEDI is a spatially discrete sampling observation, and cannot achieve spatially continuous LAI estimation alone. At the same time, GEDI has obvious position error and quality difference between different footprints. SUMMARY

[0006] The purpose of the present application is to provide a large-area medium spatial resolution time-series vegetation leaf area index estimation method, device, equipment and medium that cooperates spaceborne LiDAR and multispectral remote sensing, which can improve the estimation accuracy of LAI.

[0007] To achieve the above purpose, the present application provides the following solutions:

[0008] In a first aspect, the present application provides a large-area medium spatial resolution time-series vegetation leaf area index estimation method that cooperates spaceborne LiDAR and multispectral remote sensing, comprising:

[0009] acquiring multispectral reflectance data and spaceborne LiDAR data of a to-be-measured region in a to-be-measured period; the multispectral reflectance data is data acquired based on the high temporal and spatial resolution characteristics of Sentinel-2 with 20-meter spatial resolution and 5-day revisit period; the spaceborne LiDAR data is all GEDI (Global Ecosystem Dynamics Investigation) data acquired based on the characteristics of 4-day orbital revisit period;

[0010] constructing a multi-feature data set based on the multispectral reflectance data;

[0011] spatially and temporally matching the multi-feature data set and the spaceborne LiDAR data to obtain matching data;

[0012] inputting the matching data and the spaceborne LiDAR data into an estimation model to obtain an LAI estimation result; the estimation model is obtained by training a deep neural network; the estimation model is used to determine the mapping relationship between the spatially and temporally matched multi-feature data set and the spaceborne LiDAR data, and further determine the LAI estimation result.

[0013] Optionally, constructing a multi-feature data set based on the multispectral reflectance data specifically comprises:

[0014] performing extraction processing on the effective pixels based on the multispectral reflectance data to obtain effective pixel data; wherein, cloud shadow scenes are extracted with SCL=3, cloud scenes are extracted with SCL=10, and effective pixel data is extracted with MSK_CLDPRB less than 5%, MSK_SNWPRB less than 5%, and non-cloud shadow scenes and non-cloud scenes; SCL is a scene classification band; MSK_CLDPRB is a cloud probability; and MSK_SNWPRB is a snow probability;

[0015] determining the multi-feature data set according to the effective pixel data.

[0016] Optionally, the multi-feature data set includes reflectance, a plurality of vegetation indices, a solar zenith angle, and geospatial information.

[0017] The reflectance includes visible light, near-infrared, and short-wave infrared; the geospatial information is used to measure spatial proximity; and the geospatial information includes:

[0018]

[0019] wherein, Spatial is geospatial information; R is a constant, set to 1; and θ is a longitude of the effective pixel data. is a latitude of the effective pixel data.

[0020] Optionally, the multi-feature data set and the spaceborne lidar data are matched in space and time to obtain matching data, specifically including:

[0021] performing screening on the GEDI data based on a set screening rule to obtain screened GEDI LAI spots.

[0022] performing search on the GEDI LAI spots for each effective pixel data in the multi-feature data set, with the same position and an interval time of the observation date of the effective pixel data within a set period as a reference, to perform time matching on the multi-feature data set and the GEDI LAI spots, and to obtain searched GEDI LAI spots.

[0023] performing search in the multi-feature data set based on a 3*3 pixel block with the pixel where the searched GEDI LAI spot is located as a center, to perform spatial matching on the multi-feature data set and the GEDI LAI spots, and to obtain multi-feature set near-synchronous observation samples based on the pixel block.

[0024] determining the searched GEDI LAI spots and the multi-feature set near-synchronous observation samples as the matching data.

[0025] Optionally, the determination method of the estimation model specifically includes:

[0026] constructing a deep neural network;

[0027] inputting historical matching data, historical space-borne laser radar data and corresponding LAI results into the deep neural network, training network parameters of the deep neural network with the objective of minimizing mean square error between output of the deep neural network and the LAI results, and obtaining a trained deep neural network; the LAI results are leaf area indexes of GEDI L2B products in GEDI data;

[0028] determining the trained deep neural network as an estimation model.

[0029] Optionally, the deep neural network adopts an encoding-decoding architecture; wherein the encoding part is composed of Residuals with convolution kernel numbers of 64, 128 and 256 in series; and the decoding part is composed of Residuals with convolution kernel numbers of 128 and 64 in series.

[0030] Optionally, a mathematical expression for training the network parameters is:

[0031]

[0032] wherein Φ is a network parameter; N is a total number of samples in near-synchronous observation samples in a multi-feature data set; ω i is a weight of the i-th sample; is an output value corresponding to the i-th sample in the deep neural network; is an LAI result corresponding to the i-th sample.

[0033] In a second aspect, the present application provides a large-area medium spatial resolution time-series vegetation leaf area index estimation device cooperating with space-borne laser radar and multi-spectral remote sensing, comprising:

[0034] a data acquisition module, configured to acquire multi-spectral reflectivity data and space-borne laser radar data of a to-be-measured region in a to-be-measured period; the multi-spectral reflectivity data are data acquired based on 20-meter spatial resolution and 5-day revisit period high temporal and spatial resolution characteristics of Sentinel-2; the space-borne laser radar data are all global ecosystem dynamics investigation GEDI data acquired based on 4-day orbit revisit period characteristics;

[0035] a data set construction module, configured to construct a multi-feature data set based on the multi-spectral reflectivity data;

[0036] a matching module, configured to match the multi-feature data set and the space-borne laser radar data in space and time, and obtain matching data;

[0037] An estimation module is used to input the matching data and the spaceborne lidar data into an estimation model to obtain LAI estimation results. The estimation model is trained using a deep neural network. The estimation model is used to determine the mapping relationship between the spatially and temporally matched multi-feature dataset and the GEDI data, thereby determining the LAI estimation results.

[0038] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method for estimating the temporal vegetation leaf area index in a large area with medium spatial resolution using a collaborative spaceborne lidar and multispectral remote sensing.

[0039] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for estimating the temporal vegetation leaf area index in a large area using a combination of spaceborne lidar and multispectral remote sensing.

[0040] According to the specific embodiments provided in this application, this application has the following technical effects:

[0041] This application provides a method, apparatus, device, and medium for estimating the temporal vegetation leaf area index (LAI) in a large area using a combination of spaceborne lidar and multispectral remote sensing, with medium spatial resolution. The method involves acquiring multispectral reflectance data of the area to be measured during the measured time period; constructing a multi-feature dataset based on the reflectance data; acquiring spaceborne lidar data of the area to be measured during the measured time period; and inputting the spatially and temporally matched multi-feature dataset and the spaceborne lidar data into an estimation model to obtain the LAI estimation result. By fusing GEDI LAI with Sentinel-2 data, the saturation effect commonly found in passive optical remote sensing data can be overcome by GEDI, resulting in higher estimation accuracy than that based on passive optical remote sensing inversion. Furthermore, the estimation model is trained using a deep neural network, exhibiting stronger generalization ability, fully utilizing the effective information of the Sentinel-2 multi-feature set, and leveraging the spatiotemporally dense sampling characteristics of GEDI LAI; thus, this application improves the accuracy of LAI estimation. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1A flow chart of a method for estimating vegetation leaf area index in a large area with medium spatial resolution and time series by cooperating with spaceborne lidar and multispectral remote sensing;

[0044] Figure 2 A network structure diagram of a deep neural network;

[0045] Figure 3 A Residual structure diagram;

[0046] Figure 4 An ECA structure diagram;

[0047] Figure 5 A structure diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0049] The present application aims to play the high spatial and temporal resolution characteristics of Sentinel-2 data, the 4-day orbital revisit period of GEDI and the accuracy advantage in forest LAI estimation, consider the quality factors of GEDI data, propose a method for estimating vegetation leaf area index in a large area with medium spatial resolution and time series by cooperating with spaceborne lidar and multispectral remote sensing, and improve the accuracy of the existing SL2P tool in the forest area.

[0050] The above purposes, features and advantages of the present application can be more obvious and easy to understand. The present application will be described in further detail below with reference to the drawings and specific embodiments.

[0051] In an exemplary embodiment, as shown in Figure 1 A method for estimating vegetation leaf area index in a large area with medium spatial resolution and time series by cooperating with spaceborne lidar and multispectral remote sensing is provided. The method is executed by a computer device, specifically, can be executed by a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiments of the present application, the method is applied to a server as an example, and includes the following steps.

[0052] Step 100: Obtain multispectral reflectance data and spaceborne lidar data of a to-be-tested region in a to-be-tested period. The multispectral reflectance data is obtained based on the high spatiotemporal resolution characteristics of Sentinel-2, 20-meter spatial resolution, and 5-day revisit period; and the spaceborne lidar data is obtained based on the characteristics of all global ecosystem dynamics investigation (GEDI) data with a 4-day orbit revisit period.

[0053] Step 200: Construct a multi-feature data set based on the multispectral reflectance data.

[0054] Step 300: Spatially and temporally match the multi-feature data set and the spaceborne lidar data to obtain matched data.

[0055] Step 400: Input the matched data and the spaceborne lidar data into an estimation model to obtain an LAI estimation result. The estimation model is obtained by training a deep neural network; and the estimation model is used to determine the mapping relationship between the spatially and temporally matched multi-feature data set and the spaceborne lidar data, and further determine the LAI estimation result.

[0056] In an embodiment, the multi-feature data set is constructed based on the multispectral reflectance data, specifically including:

[0057] Based on the multispectral reflectance data, effective pixel extraction processing is performed to obtain effective pixel data; wherein, cloud shadow scenes are extracted with SCL=3, and cloud scenes are extracted with SCL=10; and the effective pixel data is extracted with MSK_CLDPRB less than 5%, MSK_SNWPRB less than 5%, and non-cloud shadow scenes and non-cloud scenes; SCL is a scene classification band; MSK_CLDPRB is a cloud probability; and MSK_SNWPRB is a snow probability.

[0058] The multi-feature data set is determined according to the effective pixel data.

[0059] The multi-feature data set includes reflectance, multiple vegetation indices, solar zenith angle, and geographic spatial information; the reflectance includes visible light, near-infrared, and short-wave infrared; and the geographic spatial information is used to measure spatial proximity. The geographic spatial information includes:

[0060]

[0061] wherein, Spatial is the geographic spatial information; R is a constant and is set to 1; and θ is the longitude of the effective pixel data. is the latitude of the effective pixel data.

[0062] In an embodiment, the multi-feature data set and the spaceborne lidar data are spatially and temporally matched to obtain matched data, specifically including:

[0063] Screening GEDI data based on the set screening rule to obtain screened GEDI LAI spots.

[0064] Searching the GEDI LAI spots for each valid pixel data in the multi-feature data set based on the same position and the interval time within the set period before and after the observation date of the valid pixel data, to perform time matching between the multi-feature data set and the GEDI LAI spots, and to obtain searched GEDI LAI spots.

[0065] Searching in the multi-feature data set based on a 3*3 pixel block with the pixel where the searched GEDI LAI spot is located as the center, to perform spatial matching between the multi-feature data set and the GEDI LAI spots, and to obtain multi-feature set near-synchronous observation samples based on the pixel block.

[0066] Determining the searched GEDI LAI spots and the multi-feature set near-synchronous observation samples as matching data.

[0067] In an embodiment, the method for determining the estimation model specifically comprises:

[0068] Constructing a deep neural network; inputting the historical matching data, the historical spaceborne laser radar data and the corresponding LAI results into the deep neural network; training network parameters of the deep neural network with the objective of minimizing the mean square error between the output of the deep neural network and the LAI results, to obtain the trained deep neural network. The LAI results are leaf area indexes of GEDI L2B products in GEDI data.

[0069] Determining the trained deep neural network as the estimation model.

[0070] The deep neural network adopts an encoding-decoding architecture; wherein the encoding part is composed of Residuals with the number of convolution kernels being 64, 128 and 256 in series; and the decoding part is composed of Residuals with the number of convolution kernels being 128 and 64 in series.

[0071] In an embodiment, the mathematical expression for training the network parameters is:

[0072]

[0073] wherein, Φ is the network parameter; N is the total number of samples in the near-synchronous observation samples in the multi-feature data set; ω i is the weight of the i th sample; is the output value corresponding to the i th sample in the deep neural network; is the LAI result corresponding to the i th sample.

[0074] In practical applications, the specific operation steps of the method mentioned in the application can be as follows:

[0075] Step 1: Obtain all Sentinel-2 reflectance data in the to-be-estimated region and the to-be-estimated period, extract all effective pixels of Sentinel-2 according to the MSK_CLDPRB (cloud probability) and MSK_SNWPRB (snow probability) and SCL (scene classification) bands of Sentinel-2, and obtain effective pixel data. Specifically, first extract cloud shadows and clouds with SCL = 3 and SCL = 10 respectively. Then extract the effective pixels of Sentinel-2 in the non-cloud shadow and cloud scene with MSK_CLDPRB less than 5%, MSK_SNWPRB less than 5%.

[0076] Step 2: Construct a Sentinel-2 multi-feature data set based on Sentinel-2 reflectance data. The Sentinel-2 multi-feature data set includes 51 features of Sentinel-2 reflectance (Bandx), multiple vegetation indices (VIx), solar zenith angle (SZA) and geographic spatial information (Spatial). Among them, the reflectance Bandx includes visible light (Band2, Band3, Band4), near-infrared (Band5, Band6, Band7, Band8, Band 8A ) and short-wave infrared (Band 11 , Band 12 ). The geographic spatial information Spatial contains 3 features to measure spatial proximity. The expression of the geographic spatial information is:

[0077]

[0078] Wherein, Spatial is the geographic spatial information; R is a constant, set to 1; θ is the longitude of the effective pixel data; is the latitude of the effective pixel data.

[0079] Thanks to the unique red edge band of Sentinel-2, the effectiveness of Sentinel-2 vegetation index for LAI inversion has been proven. Therefore, multiple Sentinel-2 vegetation indices VIx are added to the multi-feature set. The calculation formula of VIx is shown in Table 1.

[0080] Table 1 Calculation formula of vegetation index involved in VIx

[0081]

[0082]

[0083] In Table 1, Bx represents the xth reflectance band of Sentinel-2.

[0084] Step 3: Obtain all GEDI L2B products of the global ecosystem dynamics investigation data (GEDI) in the to-be-estimated area from 2019 to 2022, and extract the leaf area index (LAI) of the GEDI L2B product, which is denoted as GEDI LAI. High-quality GEDI LAI spots are screened according to the rules shown in Table 2.

[0085] Table 2: Screening rules for high-quality GEDI LAI spots

[0086]

[0087] Step 4: Search for corresponding GEDI LAI spots for each Sentinel-2 valid pixel data. To ensure the time consistency of GEDI and Sentinel-2 data, in theory, GEDI spots with the same observation date as the Sentinel-2 multi-feature set should be searched (synchronous observation). However, due to the influence of clouds, rain, and other factors, the amount of data for synchronous observation of Sentinel-2 and GEDI is limited.

[0088] Therefore, on the basis of synchronous observation of Sentinel-2 and GEDI, GEDI data with an observation time interval of Sentinel-2 and GEDI less than or equal to two days are added to ensure the representativeness and quantity of subsequent samples. Specifically, for each Sentinel-2 valid pixel, i.e., each valid pixel data in the multi-feature data set, search for GEDI spots with the same position and an observation date interval within two days (including 2 days) before and after the observation date of the Sentinel-2 valid pixel. In the case where there are multiple GEDI spots within the 2-day time interval, the GEDI spot with a smaller time interval is retained.

[0089] Step 5: Construct a GEDI-Sentinel-2 multi-feature set near-synchronous observation sample based on a pixel block. GEDI has a position error of 10.2 meters, and there is a certain spatial scale difference between the 25-meter diameter spot size of GEDI and the 20-meter spatial resolution of Sentinel-2. To weaken the GEDI position error, while considering the spatial scale difference between GEDI and Sentinel-2 and the 60-meter along-track spacing of GEDI, a 3x3 pixel block size of Sentinel-2 multi-feature set is searched with the Sentinel-2 pixel where the searched GEDI spot is located as the center, thereby forming a GEDI-Sentinel-2 multi-feature set near-synchronous observation sample based on a pixel block.

[0090] Step 6: The mapping between GEDI and its corresponding Sentinel-2 multi-feature set based on 3x3 pixel block is constructed by deep learning, that is, the GEDI LAI is taken as the output, the corresponding Sentinel-2 multi-feature set of 3x3 pixel block size is taken as the input, and the deep neural network is used for training. Among them, in order to control the uncertainty caused by the difference in GEDI spot quality, the sample is weighted by taking the sensitivity of GEDI LAI as the weight. Sensitivity represents the minimum proportion of all echoes that penetrate the canopy to reach the ground echoes. In theory, the higher this proportion, the higher the accuracy of GEDI LAI. The above training process can be represented by formula (2).

[0091]

[0092] wherein, Φ is the network parameter; N is the total number of samples in the multi-feature set near-synchronous observation sample; ω i is the weight of the i th sample; is the output value corresponding to the i th sample in the deep neural network; is the LAI result corresponding to the i th sample.

[0093] The network structure of the deep neural network used by deep learning is shown in Figure 2 . The network adopts the Encoder-Decoder architecture. The encoding Encoder and decoding Decoder are both based on residual network (Residual) as the basic unit to avoid the degradation of neural network with the increase of depth. The encoding Encoder part is composed of Residual with convolution kernel number of 64, 128 and 256 in series, and the decoding Decoder part is composed of Residual with convolution kernel number of 128 and 64 in series, which is used to gradually reduce the dimension of the features. Each Residual of the encoding Encoder and decoding Decoder part is connected by jump connection after 3 cycles as shown in Figure 2 to increase the connectivity of the features between layers. The Sentinel-2 multi-feature set is first sequentially subjected to two-dimensional convolution (Conv2D) with convolution kernel number of 64 and convolution kernel size of 3x3, BatchNorm and activation function ReLU, and then input to the Encoder-Decoder structure to obtain the output.

[0094] The Residual structure is shown in Figure 3 , which can be expressed by formula (3).

[0095]

[0096] wherein, X l and X l+1respectively represent the input and output of the l-th residual block, and the activation function is ReLU. The residual part is represented by a one-dimensional convolution Conv1D, batch normalization (Batch normalization, BatchNorm), activation function ReLU, Conv1D, BatchNorm and attention mechanism (Efficient Channel Attention, ECA) in series, W l The parameters of the residual part are represented. h() is the dimension adjustment function of the shortcut, which is composed of Conv1D and Batchnormalization, used to reduce or increase the dimension of the input X l to make its dimension (size()) consistent with The dimension of the output is kept consistent.

[0097] Among them, the ECA attention mechanism is used to further improve the effectiveness of the network in feature extraction, and reduce the collinearity between various vegetation indices, and its structure is shown in Figure 4 The ECA attention mechanism can effectively avoid the information loss caused by dimension reduction of the channel attention mechanism, which can be represented as:

[0098] w = σ (C1D k (y))*y (4)

[0099] Where w is the output of ECA; σ is the Sigmoid function; C1D k () is a one-dimensional convolution with kernel size k; y is the input feature of ECA.

[0100] The convolution kernel size k and the number C are represented as:

[0101]

[0102] C = 2 γ×k-b (6)

[0103] In the formula, |·| odd The nearest odd number is represented by γ and b, which can be set to 2 and 1 respectively.

[0104] Step 7: Based on the learned mapping relationship That is, the trained deep neural network network, with all Sentinel-2 multi-feature sets as input, obtains the LAI estimation result The process can be represented as:

[0105]

[0106] The estimation accuracy of the LAI in the forest region in the present application is higher than that of the existing SL2P tool. The reason is that the point-to-area fusion of GEDI LAI and Sentinel-2 (related to the general idea of GEDI and Sentinel-2 fusion, not limited to a certain step in the present application) takes advantage of the characteristics of GEDI that can overcome the saturation effect commonly existing in passive optical remote sensing data, and the precision advantage of GEDI LAI product in the forest region compared with the existing passive optical remote sensing inversion result.

[0107] The time series LAI estimation result in the forest region in the present application can more effectively reflect the time dynamics of LAI compared with the LAI inversion result in the forest region combined with the measured data and Sentinel. The reason is that the 4-day orbital revisit period of GEDI can more completely depict the time dynamics of LAI compared with the limited time series of measured data. The strategy of searching for the corresponding time-consistent GEDI LAI spot based on each Sentinel-2 effective pixel (step 4) ensures the consistency of GEDI and Sentinel-2 multi-feature set time observation while ensuring the sufficiency of the sample size.

[0108] The estimation accuracy of the LAI in the forest region in the present application is higher than that of the existing SL2P tool. The reason is that the point-to-area fusion of GEDI LAI and Sentinel-2 (related to the general idea of GEDI and Sentinel-2 fusion, not limited to a certain step in the present application) takes advantage of the characteristics of GEDI that can overcome the saturation effect commonly existing in passive optical remote sensing data, and the precision advantage of GEDI LAI product in the forest region compared with the existing passive optical remote sensing inversion result.

[0109] Based on the same inventive concept, the embodiments of the present application also provide an estimation device for implementing the above-mentioned method of cooperatively estimating the medium spatial resolution time series vegetation leaf area index in a large area by using the star-borne laser radar and multi-spectral remote sensing. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more cooperative star-borne laser radar and multi-spectral remote sensing medium spatial resolution time series vegetation leaf area index estimation device embodiments provided below can be referred to the limitations of the cooperative star-borne laser radar and multi-spectral remote sensing medium spatial resolution time series vegetation leaf area index estimation method in the above, which will not be repeated here.

[0110] In one exemplary embodiment, a cooperative star-borne laser radar and multi-spectral remote sensing medium spatial resolution time series vegetation leaf area index estimation device is provided, comprising:

[0111] The data acquisition module is configured to acquire multispectral reflectance data and spaceborne lidar data of a to-be-detected region in a to-be-detected period. The multispectral reflectance data is acquired based on the high spatial-temporal resolution characteristics of Sentinel-2 with a 20-meter spatial resolution and a 5-day revisit period. The spaceborne lidar data is acquired based on the characteristics of all GEDI (Global Ecosystem Dynamics Investigation) data with a 4-day orbit revisit period.

[0112] The data set construction module is configured to construct a multi-feature data set based on the multispectral reflectance data.

[0113] The matching module is configured to perform spatial and temporal matching between the multi-feature data set and the spaceborne lidar data to obtain matched data.

[0114] The estimation module is configured to input the matched data and the spaceborne lidar data into an estimation model to obtain an LAI estimation result. The estimation model is obtained by training a deep neural network. The estimation model is configured to determine a mapping relationship between the spatially and temporally matched multi-feature data set and the spaceborne lidar data, and further determine the LAI estimation result.

[0115] In an exemplary embodiment, a computer device can be provided, which can be a server or a terminal. An internal structure diagram of the computer device can be as shown in Figure 5 The computer device includes a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The database of the computer device is configured to store large-area medium spatial resolution time-series vegetation leaf area index estimation data in cooperation with spaceborne lidar and multispectral remote sensing. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a large-area medium spatial resolution time-series vegetation leaf area index estimation method in cooperation with spaceborne lidar and multispectral remote sensing.

[0116] Those skilled in the art can understand that, Figure 5The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0117] In an exemplary embodiment, a computer readable storage medium is provided, storing a computer program, which is executed by a processor to implement the steps in the above method embodiments.

[0118] In an exemplary embodiment, a computer program product is provided, including a computer program, which is executed by a processor to implement the steps in the above method embodiments.

[0119] In the present application, all actions of obtaining signals, information or data are performed in compliance with the corresponding data protection regulations and policies of the country where the device is located, and with the authorization of the owner of the corresponding device. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data need to comply with relevant regulations.

[0120] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc.

[0121] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0122] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0123] The principles and implementation modes of the present application are described by applying specific examples. The above description of the embodiments is only to help understand the method and its core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range can be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A method for large-area medium spatial resolution temporal vegetation leaf area index estimation synergizing space-borne lidar and multispectral remote sensing, characterized in that, The method comprises: acquiring multispectral reflectance data and spaceborne lidar data of a to-be-tested region in a to-be-tested period; the multispectral reflectance data are data acquired based on the high spatiotemporal resolution characteristics of Sentinel-2, with a 20-meter spatial resolution and a 5-day revisit period; the spaceborne lidar data are all GEDI (Global Ecosystem Dynamics Investigation) data acquired based on the characteristics of a 4-day orbital revisit period; constructing a multi-feature data set based on the multispectral reflectance data; spatially and temporally matching the multi-feature data set and the spaceborne lidar data to obtain matched data; inputting the matched data and the spaceborne lidar data into an estimation model to obtain an LAI estimation result; the estimation model is obtained by training a deep neural network; the estimation model is used to determine the mapping relationship between the spatially and temporally matched multi-feature data set and the spaceborne lidar data, and further determine the LAI estimation result; spatially and temporally matching the multi-feature data set and the spaceborne lidar data to obtain matched data, specifically comprising: filtering the GEDI data based on a set filtering rule to obtain filtered GEDI LAI spots; searching the GEDI LAI spots for each valid pixel data in the multi-feature data set, with the same position and an interval time of the observation date of the valid pixel data within a set period as the reference, to perform time matching on the multi-feature data set and the GEDI LAI spots, and obtain searched GEDI LAI spots; searching in the multi-feature data set based on a 3*3 pixel block with the pixel where the searched GEDI LAI spot is located as the center, to perform spatial matching on the multi-feature data set and the GEDI LAI spots, and obtain multi-feature set near-synchronous observation samples based on the pixel block; determining the searched GEDI LAI spots and the multi-feature set near-synchronous observation samples as the matched data.

2. The synergistic space-borne lidar and multispectral remote sensing large-area medium spatial resolution time-series vegetation leaf area index estimation method according to claim 1, characterized in that, constructing a multi-feature data set based on the multispectral reflectance data, specifically comprising: performing extraction processing on valid pixels based on the multispectral reflectance data to obtain valid pixel data; wherein cloud shadow scenes are extracted with SCL=3, and cloud scenes are extracted with SCL=10; valid pixel data are extracted with MSK_CLDPRB less than 5%, MSK_SNWPRB less than 5%, and non-cloud shadow scenes and non-cloud scenes; SCL is a scene classification band; MSK_CLDPRB is cloud probability; and MSK_SNWPRB is snow probability; determining the multi-feature data set according to the valid pixel data.

3. The synergistic space-borne lidar and multispectral remote sensing large-area medium spatial resolution time-series vegetation leaf area index estimation method according to claim 2, characterized in that, The multi-feature data set comprises reflectance, multiple vegetation indices, solar zenith angle, and geographic spatial information; the reflectance comprises visible light, near-infrared, and short-wave infrared; and the geographic spatial information is used to measure spatial proximity; the geographic spatial information comprises: wherein Spatial is geographic spatial information; R is a constant, set to 1; Θ is the longitude of the effective pixel data; is the latitude of the effective pixel data.

4. The synergistic space-borne lidar and multispectral remote sensing based large area medium spatial resolution temporal vegetation leaf area index estimation method according to claim 2, characterized in that, a method for determining the estimation model, specifically comprising: constructing a deep neural network; The historical matching data, the historical spaceborne laser radar data, and corresponding LAI results are input into a deep neural network, network parameters of the deep neural network are trained with the least mean square error between an output of the deep neural network and the LAI results as the target, and a trained deep neural network is obtained; the LAI results are leaf area indexes of GEDI L2B products in GEDI data. The trained deep neural network is determined as an estimation model.

5. The synergistic space-borne lidar and multispectral remote sensing large-area medium spatial resolution temporal vegetation leaf area index estimation method according to claim 1, characterized in that, The deep neural network adopts an encoding-decoding architecture; wherein the encoding part is composed of Residuals with convolution kernel numbers of 64, 128, and 256 in series; and the decoding part is composed of Residuals with convolution kernel numbers of 128 and 64 in series.

6. The synergistic space-borne lidar and multispectral remote sensing large-area medium spatial resolution temporal vegetation leaf area index estimation method according to claim 1, characterized in that, The mathematical expression for training the network parameters of the deep neural network is: wherein, Φ is a network parameter; N is the total number of samples in the near-synchronous observation samples in the multi-feature data set; ω i is the weight of the i th sample; is the output value corresponding to the i th sample in the deep neural network; is the LAI result corresponding to the i th sample.

7. A device for synergic spaceborne lidar and multispectral remote sensing based large area medium spatial resolution time series vegetation leaf area index estimation, characterized in that, The device comprises: A data acquisition module is configured to acquire multispectral reflectance data and spaceborne laser radar data of a to-be-measured region in a to-be-measured period; the multispectral reflectance data are acquired based on the high spatiotemporal resolution characteristics of Sentinel-2 with a 20-meter spatial resolution and a 5-day revisit period; and the spaceborne laser radar data are all GEDI (Global Ecosystem Dynamics Investigation) data acquired based on the characteristics of a 4-day orbital revisit period. A data set construction module is configured to construct a multi-feature data set based on the multispectral reflectance data. A matching module is configured to match the multi-feature data set and the spaceborne laser radar data in space and time to obtain matching data. An estimation module is configured to input the matching data and the spaceborne laser radar data into an estimation model to obtain LAI estimation results; the estimation model is obtained by training a deep neural network; and the estimation model is used to determine the mapping relationship between the multi-feature data set matched in space and time and the spaceborne laser radar data, and further determine the LAI estimation results. The matching of the multi-feature data set and the spaceborne laser radar data in space and time to obtain matching data specifically comprises: The GEDI data are screened based on a set screening rule to obtain screened GEDI LAI spots. Each valid pixel data in the multi-feature data set is used as a reference, and the GEDI LAI spots are searched in the multi-feature data set to perform time matching of the multi-feature data set and the GEDI LAI spots, and the searched GEDI LAI spots are obtained. The searched GEDI LAI spots are used as a center, and a 3*3 pixel block in the multi-feature data set is searched to perform spatial matching of the multi-feature data set and the GEDI LAI spots, and a multi-feature set near-synchronous observation sample based on the pixel block is obtained. The searched GEDI LAI spots and the multi-feature set near-synchronous observation sample are determined as the matching data.

8. A computer device comprising: Memory, processor, and computer program stored on the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for estimating large-area medium spatial resolution temporal vegetation leaf area index by synergic space-borne lidar and multispectral remote sensing according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the method for estimating large-area medium spatial resolution temporal vegetation leaf area index by synergic space-borne lidar and multispectral remote sensing according to any one of claims 1-6.

Images