CN117893919B - Method for generating high-space-time resolution leaf area index product in cloudy and foggy region - Google Patents
Method for generating high-space-time resolution leaf area index product in cloudy and foggy region Download PDFInfo
- Publication number
- CN117893919B CN117893919B CN202410057473.9A CN202410057473A CN117893919B CN 117893919 B CN117893919 B CN 117893919B CN 202410057473 A CN202410057473 A CN 202410057473A CN 117893919 B CN117893919 B CN 117893919B
- Authority
- CN
- China
- Prior art keywords
- landsat
- cloudy
- leaf area
- foggy
- area index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000005516 engineering process Methods 0.000 claims abstract description 31
- 230000004927 fusion Effects 0.000 claims abstract description 17
- 238000002310 reflectometry Methods 0.000 claims description 121
- 239000011159 matrix material Substances 0.000 claims description 39
- 239000011521 glass Substances 0.000 claims description 23
- 238000001914 filtration Methods 0.000 claims description 6
- 230000002123 temporal effect Effects 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims 1
- 238000013135 deep learning Methods 0.000 abstract description 12
- 230000000295 complement effect Effects 0.000 abstract description 3
- 238000012549 training Methods 0.000 description 12
- 238000005070 sampling Methods 0.000 description 10
- 230000008569 process Effects 0.000 description 5
- 238000012216 screening Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 230000007547 defect Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 2
- 238000010923 batch production Methods 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/28—Measuring arrangements characterised by the use of optical techniques for measuring areas
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Astronomy & Astrophysics (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
本发明涉及定量遥感卫星产品生成技术领域,具体地说是一种多云多雾地区高时空分辨率叶面积指数产品生成方法，本发明集成时空融合、数据同化与深度学习技术，其中时空融合技术为深度学习技术提供更多的学习信息，数据同化技术为深度学习技术提供高精度的学习目标，而深度学习技术有效解决多云多雾地区数据缺失问题，这三种技术相辅相成，填补了多云多雾地区高时空分辨率LAI产品的空白。
The present invention relates to the technical field of quantitative remote sensing satellite product generation, and specifically to a method for generating leaf area index products with high spatiotemporal resolution in cloudy and foggy areas. The present invention integrates spatiotemporal fusion, data assimilation and deep learning technologies, wherein the spatiotemporal fusion technology provides more learning information for the deep learning technology, and the data assimilation technology provides a high-precision learning target for the deep learning technology, and the deep learning technology effectively solves the problem of missing data in cloudy and foggy areas. The three technologies complement each other and fill the gap of high spatiotemporal resolution LAI products in cloudy and foggy areas.
Description
Technical Field
The invention relates to the technical field of quantitative remote sensing satellite product generation, in particular to a method for generating a leaf area index product with high space-time resolution in a cloudy and foggy region.
Background
Leaf Area Index (LAI) is an important biophysical parameter describing vegetation coverage structure and function, changes in LAI reflect vegetation response to environmental changes, and its dynamic monitoring is of great significance to global change research, ecosystem process simulation, and agricultural production management.
Currently, obtaining LAI based on satellite data has become a major means of regional scale or even global scale LAI acquisition; however, the existing LAI product has lower spatial resolution, generally 500 meters to 1 km, the coarse spatial resolution hardly reflects the spatial heterogeneity of the surface vegetation, the change of the function and structure of an ecosystem cannot be accurately described, and the application of the LAI product in the process of simulating and managing regional scale surface is limited.
Landsat data with 30m resolution in the prior art has the unique advantage of describing vegetation spatial heterogeneity, and the LAI product with high space-time resolution is generated from the Landsat data, so that the simulation and management capability of an ecosystem process can be greatly improved.
However, available Landsat observation data in cloudy and foggy areas is very limited, which brings great challenges to inverting high-precision time-continuous LAI products from Landsat, and the existing data assimilation algorithm needs high-quality and high-density Landsat observation to adjust the trend of the LAI dynamic model, otherwise, the estimation result has larger uncertainty. The Landsat and MODIS space-time fusion algorithm also has limitation, and typical FSDAF and ESTARFM algorithms all need high-quality Landsat reference images of completely clear sky, which are difficult to acquire in cloudy and foggy areas. In addition, the existing space-time fusion method is difficult to deal with the mutation and strong heterogeneity of the earth surface.
Therefore, for the reasons, a method for generating the high space-time resolution leaf area index product in the cloudy and foggy area needs to be designed, the problem of data deletion in the cloudy and foggy area is effectively solved by integrating space-time fusion, data assimilation and deep learning technologies, and the three technologies complement each other, so that the blank of the high space-time resolution LAI product in the cloudy and foggy area is filled.
Disclosure of Invention
The invention aims to overcome the defects of the prior art, provides a method for generating a high space-time resolution leaf area index product in a cloudy and foggy region, and effectively solves the problem of data missing in the cloudy and foggy region by integrating space-time fusion, data assimilation and deep learning technologies, and fills the blank of a high space-time resolution LAI product in the cloudy and foggy region.
In order to achieve the above purpose, the invention provides a method for generating a high space-time resolution leaf area index product in a cloudy and foggy region, which comprises the following steps:
S1：
S1-1, calculating a structural similarity index according to GLASS leaf area indexes aiming at images of three years before and after a Landsat target image polluted by cloud, searching a reference Landsat image which is most similar to a spatial structure of the target image, specifically, a Landsat image with the structural similarity index higher than 0.85 is considered to be similar to the target image structure, and a Landsat image which is completely clear and has the highest structural similarity index is selected as an optimal reference image;
S1-2, when the target influence is polluted by the cloud, if the optimal reference image of the S1-1 is provided, using FSDAF space-time technology to fuse Landsat with MODIS surface reflectivity to predict Landsat surface reflectivity density of the target date; if the optimal reference image conforming to the S1-1 is not provided, the original Landsat surface reflectivity of the target date is reserved;
S2, selecting pixels with the land surface reflectivity density higher than 60% as sampling points, uniformly selecting the sampling points for each surface coverage type by means of Global land 30 surface classification data, constructing a dynamic model from GLASS leaf area indexes and assimilating land surface area index reference sequences with 8-day time resolution for each sampling point by using a set Kalman filtering technology by taking the land surface reflectivity as observation;
S3, extracting a 2-year Landsat surface reflectivity time sequence matrix in a 1.5km multiplied by 1.5km area taking the sample point as a center from all the sample points as input, wherein the dimension of the matrix is 46 multiplied by 50, and the missing value is assigned to 0; according to the real missing condition of Landsat data after space-time fusion, the proportion of effective values in a Landsat earth surface reflectivity time sequence matrix is redistributed, and a sample point Landsat leaf area index reference sequence is taken as output to construct a CNN-BiGRU model;
S4, in the cloudy and foggy region, estimating pixels in all the S2, and sequentially applying FSDAF technology and a CNN-BiGRU model to estimate the high space-time resolution leaf area index of the cloudy and foggy region.
And S3, when the Landsat data of the cloudy and foggy region is obtained according to the step S1, the proportion of the effective reflectivity value in the Landsat earth surface reflectivity matrix is redistributed according to the actual missing condition, the reflectivity matrix is used as input, the Landsat leaf area index reference sequence corresponding to the sample point is used as output, and a CNN-BiGRU time sequence regression model comprehensively utilizing the space-time information is trained.
The beneficial technical effects of the invention are as follows:
The invention integrates space-time fusion, data assimilation and deep learning technologies, wherein the space-time fusion technology provides more learning information for the deep learning technology, the data assimilation technology provides a high-precision learning target for the deep learning technology, the deep learning technology effectively solves the problem of data missing in a cloudy and foggy area, and the three technologies complement each other to fill the blank of a high space-time resolution LAI product in the cloudy and foggy area.
According to the invention, space-time information constraint is introduced, so that the time change rule and the space dependence of the LAI are better modeled, the estimation precision of the LAI with high space-time resolution in the cloud and fog region is improved, and the estimation result has higher consistency with the ground reference LAI. Meanwhile, the LAI product with a long time sequence of 30m and 8 days can be accurately produced under all conditions in the cloudy and foggy region with high efficiency, the advantages of space-time fusion, data assimilation and deep learning technology can be integrated, the problem of Landsat data long time sequence deletion caused by the cloudy and foggy region is solved, the leaf area index product with long time sequence, high precision and high space-time resolution in the cloudy and foggy region is generated, and data support is provided for ecological environment monitoring and ecological process simulation in the cloudy and foggy region.
Drawings
FIG. 1 is a flow chart of a method for generating a high space-time resolution leaf area index product in a cloudy and foggy region of the present invention.
Detailed Description
Referring to fig. 1, the invention provides a method for generating a high space-time resolution leaf area index product in a cloudy and foggy region, which comprises the following steps:
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The embodiment of the invention provides a method for generating a high space-time resolution leaf area index product in a cloudy and foggy region, which comprises the following steps:
S1: aiming at a cloud polluted Landsat target image, calculating a structural similarity index by using a GLASS leaf area index, searching a reference Landsat image which is most similar to the spatial structure of the GLASS leaf area index, and fusing Landsat with MODIS ground surface reflectivity by using FSDAF space-time technology to improve the available Landsat ground surface reflectivity density;
S2: selecting available sample points with higher land surface reflectivity density, and generating a land surface index reference sequence from the GLASS leaf area index and the land surface reflectivity by using a set Kalman filtering technology;
S3: extracting a Landsat earth surface reflectivity time sequence matrix of 2 years in a region with 1.5km multiplied by 1.5km taking the sample point as the center as input, redistributing the proportion of effective values in the Landsat earth surface reflectivity time sequence matrix according to the real missing condition of Landsat data after space-time fusion, and constructing a CNN-BiGRU model by taking a sample point Landsat leaf area index reference sequence as output;
S4: and sequentially applying FSDAF technology and a CNN-BiGRU model in the cloudy and foggy region to estimate the high space-time resolution leaf area index of the cloudy and foggy region.
In one embodiment, step S1 further comprises:
When the GLASS leaf area index is used for calculating the structural similarity index, the time window is set to be three years before and after the target image, the Landsat image with the structural similarity index higher than 0.85 is considered to be similar to the target image structure, and the Landsat image with the complete clear sky and the highest structural similarity index is selected as the optimal reference image.
In the implementation process, a certain cloudy and foggy area is divided into a plurality of sinusoidal projection subareas, the Lansat earth surface reflectivity of a subarea, which is polluted by clouds, is called target Lansat earth surface reflectivity, the date is called target date, and the GLASS V6250 meter LAI matched with the target date and time and the target Lansat earth surface reflectivity space is called target GLASS LAI. And (3) taking the target date as the center, defining a time window of three years (seven years in total), acquiring GLASS LAI of other dates and calculating a structural similarity index with the target GLASS LAI. And (3) sequentially screening Landsat earth surface reflectivities of other dates according to the structural similarity structural indexes from high to low until the Landsat earth surface reflectivities which are completely clear sky and have the structural similarity indexes not smaller than 0.85 are found, and taking the Landsat earth surface reflectivities as reference Landsat earth surface reflectivities, wherein the corresponding dates are called reference dates.
And obtaining MODIS 500 m earth surface reflectivities of the reference date and the target date, which are spatially matched with the reference Landsat earth surface reflectivities and correspond to the wave bands, respectively serving as the reference and target MODIS earth surface reflectivities, applying FSDAF space-time fusion algorithm to input the reference Landsat earth surface reflectivities, the reference MODIS earth surface reflectivities and the target MODIS earth surface reflectivities, and predicting the target Landsat earth surface reflectivities.
If a sub-area polluted by the cloud on a specific date has a reference image which is completely clear and empty and has a structural similarity index not smaller than 0.85, a FSDAF space-time fusion algorithm is applied to predict the target Landsat ground surface reflectivity of the sub-area.
In one embodiment, step S2 includes:
and counting the density of available Landsat ground surface reflectivity in the time sequence of each year of each pixel in the whole cloudy and foggy area 2000-2022, screening pixels with the available Landsat ground surface reflectivity density higher than 60%, and uniformly selecting the same number of pixels with the available Landsat ground surface reflectivity density higher than 60% for each ground surface type as sampling points according to the classified products of 30 meters of the Global land 30.
And extracting GLASS V6 LAI time sequences of corresponding years of the sampling points for constructing a LAI dynamic model, simulating Landsat ground surface reflectivity by adopting an A Two-Layer Canopy Reflectance Model (ACRM) canopy reflectivity model, assimilating the GLASS V6 LAI and Landsat ground surface reflectivity by using a set Kalman filtering technology, and generating a 30 m 8 day LANDSAT LAI reference time sequence.
In one embodiment, step S3 includes:
and (3) extracting a corresponding Landsat surface reflectivity time series matrix for each sample point in the step S2. The time dimension is the current year and the previous year for two years, the space dimension is 1.5km×1.5km, the spectrum dimension is red, green, blue, near infrared and 2 short wave infrared for 6 wave bands, and the final reflectivity matrix is 6×46×50×50.
And (3) counting 20% -90% of the land surface reflectivity missing conditions of the cloudy and foggy area obtained in the step (S1), adjusting the use proportion of the effective reflectivity value in the land surface reflectivity matrix according to the real missing conditions, so as to simulate various missing conditions, uniformly distributing the adjusted available land surface reflectivity density at 10% -80%, and setting the invalid value in the reflectivity matrix to 0.
The reflectance matrix is used as input, LANDSAT LAI reference time sequences are used as output to form training samples, 85% of samples are randomly selected for model training, and the remaining 15% are used for model verification. Training LANDSAT LAI time series regression models by adopting a CNN-BiGRU algorithm which simultaneously considers space and time information, respectively training a model for the three sensors of TM, ETM and OLI, wherein the three models use the same parameter setting: the Adam gradient descent algorithm has a batch size of 100, a maximum number of iterations of 1000, an initial learning rate of 0.01, and a learning rate of 0.001 after 500 training, each training upsets the data set.
In one embodiment, step S4 includes:
When LANDSAT LAI time series of a certain pixel in a certain year are estimated, a corresponding Landsat surface reflectivity time series matrix is extracted. The time dimension is the current year and the previous year for two years, the space dimension is 1.5km×1.5km, the spectrum dimension is red, green, blue, near infrared and 2 short wave infrared for 6 wave bands, the final reflectivity matrix is 6×46×50×50, and the invalid value in the reflectivity matrix is set to 0. And (3) inputting a reflectivity matrix by using the CNN-BiGRU model obtained by training in the step (3) to obtain a LANDSAT LAI time series estimated value of the pixel for one year. The steps are repeatedly executed for each year of each pixel, so that the batch time sequence leaf area index batch production of each pixel is realized.
Finally, the Landsat leaf area index products of 2000-2022 years, 30 meters and 8 days in the cloudy and foggy region are obtained by using the steps S1 to S4.
The invention integrates the space-time fusion, data assimilation and deep learning technologies, fully utilizes multi-source remote sensing data, considers the space-time information constraint of vegetation LAI, and generates a long-time sequence high-precision space-time continuous 30-meter leaf area index product in a cloudy and foggy region.
The implementation flow of the method provided by the invention is shown in fig. 1, and a specific implementation mode is further described with reference to an embodiment. A method for generating a leaf area index product with high space-time resolution in a cloudy and foggy region specifically comprises the following steps:
Step 1, aiming at a Landsat target image polluted by cloud, calculating a structural similarity index by using a GLASS leaf area index, searching a reference Landsat image most similar to the spatial structure of the GLASS leaf area index, and fusing Landsat with an MODIS ground surface reflectivity by using FSDAF space-time technology to improve the available Landsat ground surface reflectivity density;
Taking an Chuan Dian region as an example, dividing the whole Chuan Dian region into 28.8km multiplied by 28.8km sinusoidal projection subregions, regarding a subregion with cloud pollution on a certain date as a target date, taking Landsat ground surface reflectivity images (6 wave bands of red, green, blue, near red, short wave infrared 1 and short wave infrared 2) with cloud pollution on the target date as target Landsat ground surface reflectivity, and acquiring GLASS V6250 m LAI images with the same size as the target Landsat ground surface reflectivity on the target date as target GLASS LAI. And (3) taking the target date as the center, defining a time window of three years (seven years in total), acquiring GLASS V6250 m LAI images with the same size as the target Landsat earth surface reflectivity on other dates (other dates except the target date in the time window), calculating a structural similarity index with the target GLASS LAI, and sequencing from large to small according to the structural similarity, wherein the structural similarity index is larger, so that the earth surface structural characteristics of the other dates are similar to the earth surface structural characteristics of the target date. And obtaining Landsat earth surface reflectivities with the same size as the target Landsat earth surface reflectivities of other dates, identifying cloud and clear sky pixels by using a Fmak 4.6 algorithm, sequentially screening Landsat earth surface reflectivities of other dates according to structural similarity structural indexes from high to low until the Landsat earth surface reflectivities of the complete clear sky with the structural similarity indexes not less than 0.85 are found, and taking the Landsat earth surface reflectivities as reference Landsat earth surface reflectivities, wherein the corresponding dates are called reference dates.
Obtaining an MODIS 500-meter earth surface reflectivity image with the same size as the reference Landsat earth surface reflectivity and corresponding to a wave band, taking the image as the reference MODIS earth surface reflectivity, obtaining an MODIS 500-meter earth surface reflectivity image with the same size as the target Landsat earth surface reflectivity and corresponding to the wave band, taking the image as the target MODIS earth surface reflectivity, applying FSDAF space-time fusion algorithm to input the reference Landsat earth surface reflectivity, the reference MODIS earth surface reflectivity and the target MODIS earth surface reflectivity, and predicting the target Landsat earth surface reflectivity.
In practice, for a target subarea polluted by cloud, if a completely clear sky is found and the Landsat surface reflectivity with the structural similarity index not smaller than 0.85 is used as a reference image, executing FSDAF space-time fusion algorithm to predict the Landsat surface reflectivity of the target date; if the land surface reflectivity meeting the condition is not used as the reference image, the original land surface reflectivity of the target date is reserved.
After the step is finished, the density of the available Landsat surface reflectivity in the Chuan Yunnan area is improved.
Step 2, selecting a sample point with higher available Landsat surface reflectivity density, and generating a Landsat leaf area index reference sequence from the GLASS leaf area index and the Landsat surface reflectivity by using a set Kalman filtering technology;
Counting the density of available Landsat ground surface reflectivity in a time sequence of each year of each pixel in 2000-2022 years of the whole Chuan Yunnan region, screening pixels with available Landsat ground surface reflectivity density higher than 60%, classifying products according to Global land30 ground surface 30 m, uniformly selecting equal quantity of pixels with available Landsat ground surface reflectivity density higher than 60% for each ground surface type as sampling points, and finally collecting 1257795 sampling points.
And extracting GLASS V6 LAI time sequences of corresponding years of the sampling points for constructing a LAI dynamic model, simulating Landsat ground surface reflectivity by adopting an A Two-Layer Canopy Reflectance Model (ACRM) canopy reflectivity model, assimilating the GLASS V6 LAI and Landsat ground surface reflectivity by using a set Kalman filtering technology, and generating a 30 m 8 day LANDSAT LAI reference time sequence.
Step 3, extracting a Landsat earth surface reflectivity time sequence matrix of 2 years in a region with 1.5km multiplied by 1.5km by taking the sample point as a center as input, redistributing the proportion of effective values in the Landsat earth surface reflectivity time sequence matrix according to the real missing condition of Landsat data of the whole Chuan Yunnan region after space-time fusion, and constructing a CNN-BiGRU model by taking a Landsat leaf area index reference sequence of the sample point as output;
extracting a corresponding Landsat earth surface reflectivity time sequence matrix of each sample point in the step 2, and particularly extracting Landsat earth surface reflectivity data in a region with 1.5km multiplied by 1.5km taking the sample point as a center in space, wherein the reflectivity matrix is 50 multiplied by 50; selecting a time sequence of two years of the year of the sample point and the previous year, wherein the number of Landsat ground surface reflectivities in each year is 23, and the reflectivity matrix is 46 multiplied by 50; spectrally, red, green, blue, near infrared and 2 short wave infrared are selected for a total of 6 band surface reflectivities, where the reflectivity matrix becomes 6 x 46 x 50.
The sampling points in the step 2 are obtained by screening pixels with the available Landsat surface reflectivity density higher than 60%, the available Landsat surface reflectivity density in the corresponding reflectivity matrix is also higher than 60%, and LANDSAT LAI reference time sequences generated on the sampling points in the step 2 have higher precision. However, the use proportion of the effective reflectivity values in the reflectivity matrix of the sample points needs to be adjusted according to the actual lack of the land surface reflectivity of Landsat in the Chuan-Dian region. Specifically, the defect condition of Landsat surface reflectivity of the Chuan Yunnan area obtained in the step 1 is counted to be 20% -90%, the use proportion of the effective reflectivity value in the Landsat surface reflectivity matrix is adjusted according to the actual defect condition, so that various defect conditions are simulated, and the adjusted available Landsat surface reflectivity density is uniformly distributed at 10% -80%.
And assigning 0 to the invalid earth surface reflectivity in the reflectivity matrix, and reserving the valid earth surface reflectivity value. And taking the earth surface reflectivity matrix as a model input, taking a corresponding LANDSAT LAI reference time sequence as a model output, and combining the model training samples. The CNN-BiGRU algorithm, which can consider both spatial and temporal information, is chosen to train a regression model that estimates LANDSAT LAI the time series from the surface reflectivity matrix, with model parameters set to: the Adam gradient descent algorithm has a batch size of 100, a maximum number of iterations of 1000, an initial learning rate of 0.01, and a learning rate of 0.001 after 500 training, each training upsets the data set. Training samples were randomly divided into two groups, of which 85% were used for model training and the remaining 15% were used for model independent verification.
For different Landsat sensors, including TM, ETM+, OLI, we trained three CNN-BiGRU models using the same procedure and parameter settings.
And 4, sequentially applying FSDAF technology and a CNN-BiGRU model in the Chuan-Dian region, and estimating the leaf area index with high space-time resolution in the Chun-Dian region.
And (3) executing the step 1 in the Chuan Yunnan region to improve the available Landsat ground surface reflectivity density.
When LANDSAT LAI time series of a certain pixel in a certain year are estimated, extracting a corresponding Landsat earth surface reflectivity time series matrix, specifically, extracting Landsat earth surface reflectivity data in a region with the pixel as a center of 1.5km multiplied by 1.5km in space, wherein the reflectivity matrix is 50 multiplied by 50; selecting a time sequence of two years from the year and the previous year, wherein the number of Landsat surface reflectivities in each year is 23, and the reflectivity matrix is 46×50×50; spectrally, red, green, blue, near infrared and 2 short wave infrared are selected for a total of 6 band surface reflectivities, where the reflectivity matrix becomes 6 x 46 x 50. And assigning an invalid earth surface reflectivity value in the earth surface reflectivity matrix to be 0, and reserving an effective value. And (3) inputting the final earth surface reflectivity matrix into the CNN-BiGRU model trained in the step (3) to obtain a LANDSAT LAI time series estimated value of the pixel in one year. The steps are repeatedly executed for each year of each pixel, so that the batch time sequence leaf area index batch production of each pixel is realized.
Based on the steps, finally, landsat leaf area index products of 2000-2022 years, 30m and 8 days in Chuan Yunnan region are generated.
The above is only a preferred embodiment of the present application, only for helping to understand the method and the core idea of the present application, and the scope of the present application is not limited to the above examples, and all technical solutions belonging to the concept of the present application belong to the scope of the present application. It should be noted that modifications and adaptations to the present application may occur to one skilled in the art without departing from the principles of the present application and are intended to be within the scope of the present application.
The invention fundamentally solves the problems that Landsat observation data available in a cloudy and foggy area in the prior art are very limited, the existing data assimilation algorithm needs high-quality and high-density Landsat observation to adjust the trend of the LAI dynamic model, otherwise, the estimation result has larger uncertainty to influence the estimation result, the time change rule and the space dependence of the LAI are better modeled by introducing space-time information constraint, the estimation precision of the LAI with high space-time resolution in the cloudy and foggy area is improved, and the estimation result has higher consistency with the ground reference LAI.
Claims (2)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410057473.9A CN117893919B (en) | 2024-01-12 | 2024-01-12 | Method for generating high-space-time resolution leaf area index product in cloudy and foggy region |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410057473.9A CN117893919B (en) | 2024-01-12 | 2024-01-12 | Method for generating high-space-time resolution leaf area index product in cloudy and foggy region |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117893919A CN117893919A (en) | 2024-04-16 |
| CN117893919B true CN117893919B (en) | 2024-06-18 |
Family
ID=90650603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410057473.9A Active CN117893919B (en) | 2024-01-12 | 2024-01-12 | Method for generating high-space-time resolution leaf area index product in cloudy and foggy region |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117893919B (en) |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20220061236A1 (en) * | 2020-08-25 | 2022-03-03 | The Board Of Trustees Of The University Of Illinois | Accessing agriculture productivity and sustainability |
| CN113192117B (en) * | 2021-02-08 | 2022-03-22 | 南京农业大学 | Method for improving estimation of leaf area index of early growth stage of wheat based on Sentinel-2 satellite image red edge band |
| CN113378476B (en) * | 2021-06-28 | 2022-07-19 | 武汉大学 | A global 250-meter resolution spatiotemporal continuous leaf area index satellite product generation method |
| CN115063676A (en) * | 2022-05-27 | 2022-09-16 | 中国人民解放军战略支援部队信息工程大学 | Ship target classification method based on AIS data |
| CN116956537A (en) * | 2022-12-30 | 2023-10-27 | 中国地质大学（武汉） | Surface temperature spatiotemporal fusion method and device based on MODIS-Landsat remote sensing data |
| CN116776651B (en) * | 2023-08-23 | 2023-11-14 | 中国科学院空天信息创新研究院 | Method and device for measuring and calculating surface evapotranspiration, electronic equipment and storage medium |
-
2024
- 2024-01-12 CN CN202410057473.9A patent/CN117893919B/en active Active
Non-Patent Citations (2)
| Title |
|---|
| Simultaneous estimation of five temporally regular land variables at seven spatial resolutions from seven satellite data using a multi-scale and multi-depth convolutional neural network;Guodong Zhang等;Remote Sensing of Environment;20231209;全文 * |
| 基于字符级联合网络特征融合的中文文本情感分析;王丽亚;刘昌辉;蔡敦波;赵彤洲;王梦;;微电子学与计算机;20200105(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117893919A (en) | 2024-04-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Racoviteanu et al. | Challenges and recommendations in mapping of glacier parameters from space: results of the 2008 Global Land Ice Measurements from Space (GLIMS) workshop, Boulder, Colorado, USA | |
| Khoshnood Motlagh et al. | Analysis and prediction of land cover changes using the land change modeler (LCM) in a semiarid river basin, Iran | |
| Panagos et al. | Estimating the soil organic carbon content for European NUTS2 regions based on LUCAS data collection | |
| CN113902580B (en) | Historical farmland distribution reconstruction method based on random forest model | |
| Al-shalabi et al. | Manifestation of remote sensing data in modeling urban sprawl using the SLEUTH model and brute force calibration: a case study of Sana’a city, Yemen | |
| Pásztor et al. | Compilation of novel and renewed, goal oriented digital soil maps using geostatistical and data mining tools | |
| CN113205014B (en) | A method for extracting cultivated land from time series data based on image sharpening | |
| WO2024152865A1 (en) | Meteorological-station-based ecosystem carbon-water flux calculation method and system | |
| Alahmadi et al. | A comparison of small-area population estimation techniques using built-area and height data, Riyadh, Saudi Arabia | |
| Jain et al. | Modeling urban land cover growth dynamics based on land change modeler (LCM) using remote sensing: a case study of Gurgaon, India | |
| Le et al. | Deep learning model development for detecting coffee tree changes based on Sentinel-2 imagery in Vietnam | |
| Holmlund et al. | Constraining 135 years of mass balance with historic structure-from-motion photogrammetry on Storglaciären, Sweden | |
| Das et al. | Assessment and prediction of urban expansion using CA-based SLEUTH urban growth model: A case study of Kolkata Metropolitan area (KMA), West Bengal, India | |
| CN115797501A (en) | Time-series forest age mapping method combining forest disturbance and recovery events | |
| Zeitfogel et al. | Soil information on a regional scale: Two machine learning based approaches for predicting saturated hydraulic conductivity | |
| CN117893919B (en) | Method for generating high-space-time resolution leaf area index product in cloudy and foggy region | |
| Omar et al. | Modelling land-use and land-cover changes using Markov-CA, and multiple decision making in Kirkuk city | |
| Li et al. | A framework for quantifying the uncertainty in upscaling evapotranspiration from homogeneous to heterogeneous underlying surface | |
| Lv et al. | Threshold effect of data amount and grid size on urban land use type identification using multi-source data fusion | |
| Huang et al. | Simultaneous update of high-resolution land-cover mapping attempt: Wuhan and the surrounding satellite cities cartography using L2HNet | |
| CN117217075A (en) | Grid scale ecosystem carbon water flux assessment method based on flux observation | |
| Linares et al. | Application of remote sensing and cellular automata model to analyze and simulate urban density changes | |
| Lennert et al. | Creating Wallonia's new very high resolution land cover maps: combining grass GIS Obia and OTB pixel-based results | |
| Raju et al. | Object Recognition in Remote Sensing Images Based on Modified Backpropagation Neural Network. | |
| Hackländer et al. | Land potential assessment and trend-analysis using 2000–2021 FAPAR monthly time-series at 250 m spatial resolution |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |