CN116842351A - Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment - Google Patents
Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment Download PDFInfo
- Publication number
- CN116842351A CN116842351A CN202311118002.6A CN202311118002A CN116842351A CN 116842351 A CN116842351 A CN 116842351A CN 202311118002 A CN202311118002 A CN 202311118002A CN 116842351 A CN116842351 A CN 116842351A
- Authority
- CN
- China
- Prior art keywords
- features
- meteorological
- data
- carbon sink
- maturity
- 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.)
- Granted
Links
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 title claims abstract description 142
- 229910052799 carbon Inorganic materials 0.000 title claims abstract description 142
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000010276 construction Methods 0.000 title claims abstract description 7
- 238000000605 extraction Methods 0.000 claims abstract description 84
- 230000008859 change Effects 0.000 claims abstract description 68
- 230000003068 static effect Effects 0.000 claims abstract description 43
- 238000013210 evaluation model Methods 0.000 claims abstract description 34
- 238000012549 training Methods 0.000 claims abstract description 22
- 230000035784 germination Effects 0.000 claims description 37
- 238000006243 chemical reaction Methods 0.000 claims description 29
- 238000012417 linear regression Methods 0.000 claims description 19
- 238000011156 evaluation Methods 0.000 claims description 15
- 230000006870 function Effects 0.000 claims description 7
- 230000035613 defoliation Effects 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 2
- 239000004973 liquid crystal related substance Substances 0.000 claims 2
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 15
- 230000008569 process Effects 0.000 description 14
- 244000234281 Tamarix gallica Species 0.000 description 12
- 235000014265 Tamarix gallica Nutrition 0.000 description 12
- 241000196324 Embryophyta Species 0.000 description 11
- 230000012010 growth Effects 0.000 description 11
- 150000003839 salts Chemical class 0.000 description 11
- 239000013598 vector Substances 0.000 description 11
- 230000011218 segmentation Effects 0.000 description 10
- 238000003860 storage Methods 0.000 description 7
- 235000010185 Tamarix canariensis Nutrition 0.000 description 5
- 235000010154 Tamarix ramosissima Nutrition 0.000 description 5
- 238000009825 accumulation Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 4
- 238000005859 coupling reaction Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000009919 sequestration Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- -1 /or Species 0.000 description 2
- 241000124033 Salix Species 0.000 description 2
- 239000003086 colorant Substances 0.000 description 2
- 238000013480 data collection Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000009545 invasion Effects 0.000 description 2
- 230000002045 lasting effect Effects 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004451 qualitative analysis Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 239000002028 Biomass Substances 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 230000004071 biological effect Effects 0.000 description 1
- 238000009933 burial Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000008216 herbs Nutrition 0.000 description 1
- 230000002015 leaf growth Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000006386 neutralization reaction Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 235000015097 nutrients Nutrition 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000013535 sea water Substances 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a carbon reserve monitoring technology, in particular to a coastal wetland carbon sink assessment model construction method, an assessment method and electronic equipment, wherein the carbon sink assessment model comprises the following steps: the system comprises a first feature extraction network, a second feature extraction network, a maturity prediction model and an assessment model; acquiring sample data; inputting the first near image data to a first feature extraction network, and extracting static features and time sequence change features of an evolution region; inputting the meteorological data to a second feature extraction network to extract meteorological features; taking the static characteristics, the time sequence variation characteristics and the meteorological characteristics as inputs of a maturity prediction model, and taking maturity data as outputs to train the maturity prediction model for multiple times until the loss function is smaller than a preset threshold value, so as to obtain a trained maturity prediction model; and inputting the maturity data, the meteorological features and the second near-image data into an evaluation model, and training the evaluation model for multiple times to obtain a trained carbon sink evaluation model.
Description
Technical Field
The invention relates to the technical field of carbon reserve monitoring, in particular to a method for constructing a coastal wetland carbon sink assessment model, an assessment method and electronic equipment.
Background
Research shows that annual carbon burial amount per square kilometer of the coastal wetland can be estimated to be 0.22GgC. Therefore, effectively evaluating the carbon sink capacity of the coastal wetland is an important basis for achieving the objective of carbon neutralization.
Coastal wetlands are sensitive and fragile areas which change globally, are often in a changing state due to the influence of biological activities, human activities, climate and other factors, for example, some coastal wetlands disappear, some artificial coastal wetlands and some types of coastal wetlands change due to human activities, biological invasion and the like, and the carbon sink forms and the carbon sink capacities of different types of coastal wetlands are very different.
Based on this, there are some methods for evaluating the carbon sink effect of the coastal wetland based on the evolution region of the coastal wetland in the related art, for example, in the chinese patent with the patent number of CN202310115601.6, it is pointed out that in addition to the qualitative analysis of the mature wetland region, the evolution data of the coastal wetland to be evaluated is comprehensively considered, the evolution type, the types of the wetland before and after the evolution, the carbon sink form before and after the evolution, the carbon sink capacity before and after the evolution, the carbon sink quantity before and after the evolution and other factors representing the carbon sink are comprehensively evaluated on the coastal wetland to be evaluated, so that the carbon sink of the wetland to be evaluated can be more accurately obtained.
Although the carbon sink evaluation is performed based on the evolution region of the coastal wetland in the related art, the carbon sink evaluation is performed only based on the area of the evolution region and the type of the coastal wetland in the evolution region, and the actual characteristics of the evolution region are not considered, so that the more accurate carbon sink evaluation is achieved.
Therefore, how to perform carbon sink assessment more accurately is a technical problem to be solved.
Disclosure of Invention
The application aims to provide a coastal wetland carbon sink assessment model construction method, an assessment method and electronic equipment, so as to solve the technical problem of how to perform more accurate carbon sink assessment and provide theoretical guidance for management of coastal wetlands.
In order to achieve the above object, the present application provides the following solutions:
according to a first aspect, an embodiment of the present application provides a method for constructing a carbon sink estimation model of a coastal wetland, where the carbon sink estimation model includes: the system comprises a first feature extraction network, a second feature extraction network, a maturity prediction model and an assessment model; acquiring sample data, wherein the sample data comprises meteorological data of a coastal wetland in a first preset period, first near-image data of an evolution region of the coastal wetland, second near-image data of a mature region of the coastal wetland and maturity data of the evolution region in a second preset period, the evolution region comprises conversion regions among different types of wetlands and net change regions of the coastal wetland except the conversion regions, and the first preset period is before the second preset period in a time sequence dimension; inputting the first near-image data to the first feature extraction network, and extracting static features and time sequence change features of the evolution region; inputting the meteorological data to the second feature extraction network to extract the meteorological features; taking the static characteristics, the time sequence change characteristics and the meteorological characteristics as the input of the maturity prediction model, and taking the maturity data as the output to train the maturity prediction model for a plurality of times until the loss function is smaller than a preset threshold value, so as to obtain a trained maturity prediction model; and inputting the maturity data, the meteorological features and the second near image data into the evaluation model, and training the evaluation model for multiple times to obtain a trained carbon sink evaluation model.
Optionally, the first feature extraction network includes a first branch feature extraction network and a second branch feature extraction network, and a first timing feature extraction layer; the first near-field image data comprise near-ground three-dimensional point cloud data and aerial image data; the inputting the first near image data to the first feature extraction network, extracting static features and time sequence variation features of the evolution region comprises: inputting the near-ground three-dimensional point cloud data into a first branch feature extraction network to extract three-dimensional network features of the evolution region; inputting the aerial image data into the second branch feature extraction network to extract color space features of the evolution region; aligning the three-dimensional network feature with the color space feature to obtain the static feature; and inputting the static features into the time sequence feature extraction network to extract the time sequence change features.
Optionally, the aligning the three-dimensional network feature and the color space feature to obtain the static feature further includes: performing first clustering on the three-dimensional network characteristics to obtain first sparsity of plants in the evolution region; performing second clustering on the color space features to obtain second sparsity of the evolution region; and fusing the three-dimensional network feature and the color space feature based on the first sparsity and the second sparsity to obtain the static feature.
Optionally, the second feature extraction network includes a meteorological feature extraction layer, a second timing feature layer, and an attention layer; inputting the meteorological features into the meteorological feature extraction layer to extract multiple types of meteorological features; inputting the multi-type meteorological features to the second time sequence feature layer to extract the meteorological accumulated change features; and capturing the weighting weight of the weather accumulated change feature through the attention layer to obtain the weather accumulated change feature with the weighting weight.
Optionally, the evaluation model includes a third feature extraction network and a linear regression model; inputting the maturity data, the meteorological features and the second near image data into the evaluation model, performing multiple rounds of training on the evaluation model, and obtaining a trained carbon sink evaluation model comprises: extracting vegetation change features in the first preset period in the second near-earth image based on the third feature extraction network; and constructing the linear regression model by using the maturity data, the meteorological features and the vegetation change features as variables.
Optionally, the first preset period includes a preset period in a spring germination period; the meteorological data comprise air temperature, rainwater quantity, duration of raining and storm surge; the second preset time period comprises any time period after the first preset time period and before the defoliation period.
According to a second aspect, an embodiment of the present application provides a method for evaluating carbon sink of a coastal wetland, including: acquiring meteorological data of a coastal wetland to be evaluated in a first preset period, first near-image data of an evolution region of the coastal wetland and second near-image data of a mature region of the coastal wetland; inputting the first near-image data to the first feature extraction network, and extracting static features and time sequence change features of the evolution region; inputting the meteorological data to the second feature extraction network to extract the meteorological features; inputting the static characteristics, the time sequence change characteristics and the meteorological characteristics into the maturity prediction model to obtain the maturity of the evolution region; and inputting the maturity, the meteorological features and the second near-image data into a trained evaluation model to obtain a carbon sink evaluation result of the coastal wetland.
Optionally, the evaluation model includes a third feature extraction network and a linear regression model; inputting the maturity, the meteorological features and the second near-image data into a trained evaluation model, and obtaining a carbon sink evaluation result of the coastal wetland comprises the following steps: inputting the second near-earth image data into the third characteristic extraction network to extract vegetation change characteristics in the first preset period in the second near-earth image; inputting the maturity data, the meteorological features and the vegetation change features into the linear regression model to obtain a carbon sink weight coefficient of the coastal wetland; and determining carbon sinks of the coastal wetland based on the preset carbon sink capacity corresponding to the coastal wetland type and the carbon sink weight coefficient.
Optionally, the method further comprises: acquiring maturity data in a plurality of continuous preset periods of an evolution region in the coastal wetland to be evaluated; determining an evolution rate of the evolution region based on a plurality of consecutive maturity data; and adjusting the carbon sink weight coefficient of the evolution region based on the evolution rate.
According to a third aspect, an embodiment of the present application provides an electronic device, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the coastal wetland carbon sink assessment model construction method according to any one of the first aspects and/or the coastal wetland carbon sink assessment method according to the second aspect.
According to the method and the system, when the coastal wetland to be evaluated is evaluated, particularly when the wetland in the region with frequent change is evaluated, in addition to qualitative analysis of the mature wetland region in the prior art, the evolution data of the coastal wetland to be evaluated is comprehensively considered, the evolution type, the wetland types before and after the evolution, the carbon sink forms before and after the evolution, the carbon sink capacity before and after the evolution, the carbon sink quantity before and after the evolution and other factors representing the carbon sink are comprehensively evaluated, and the carbon sink of the wetland to be evaluated can be more accurately obtained.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and do not constitute a limitation on the invention. In the drawings:
FIG. 1 shows a schematic diagram of a carbon sink evaluation model structure according to the present invention;
FIG. 2 shows a schematic flow chart of a coastal wetland carbon sink assessment model construction method;
FIG. 3 shows a schematic view of the maturity prediction model of the present invention;
FIG. 4 is a flow chart of a method of extracting static features and time series variation features of the present invention;
FIG. 5 is a flow chart of a method of extracting meteorological data of the present invention;
fig. 6 shows a flow diagram of the coastal wetland carbon sink assessment method of the invention.
Detailed Description
For a clearer understanding of the technical features, objects and effects of the present invention, embodiments of the present invention will now be described with reference to the drawings, in which like reference numerals refer to identical or structurally similar but functionally identical components throughout the separate views.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those described herein, and therefore the scope of the present invention is not limited to the specific embodiments disclosed below.
In the following description, various aspects of the present invention will be described, however, it will be apparent to those skilled in the art that the present invention may be practiced with only some or all of the structures or processes of the present invention. For purposes of explanation, specific numbers, configurations and orders are set forth, it is apparent that the invention may be practiced without these specific details. In other instances, well-known features will not be described in detail so as not to obscure the invention.
In the prior art, the evaluation is often based on the area of the evolution area and the type of the coastal wetland in the evolution area, however, the maturity of the evolution area is a key to the carbon sink capacity of the evolution area, and is not related to the duration of the evolution, and is also related to whether the vegetation in the area can grow normally or not. In addition, vegetation self-growth is affected by the coastal environment, so that the evolution time of the evolution region cannot accurately reflect the carbon sink capacity of the evolution region, and the accuracy of carbon sink assessment is reduced.
Based on this, the embodiment of the application provides a method for constructing a coastal wetland carbon sink estimation model, as shown in fig. 1, the carbon sink estimation model may include a first feature extraction network, a second feature extraction network, a maturity prediction model and an estimation model, as shown in fig. 2, the method for constructing the coastal wetland carbon sink estimation model includes:
s10, acquiring sample data, wherein the sample data comprise meteorological data of a coastal wetland in a first preset period, first near-image data of an evolution region of the coastal wetland, second near-image data of a mature region of the coastal wetland and maturity data of the evolution region in a second preset period, the evolution region comprises conversion regions among different types of wetlands and net change regions of the coastal wetland of each type except the conversion regions, and the first preset period is before the second preset period in a time sequence dimension.
As an illustrative example, the coastal wetlands may include mature coastal wetlands, i.e., wetlands of various types that are not easily changed near the central region, and also include modified coastal wetlands, e.g., edge regions of various types of coastal wetlands, which are more easily changed due to influence of environment, artificial activities, or biological invasion, etc. In this embodiment, the evolution region includes a region where a transition has occurred, for example, a beach becomes a salt marsh and/or a tamarix chinensis forest, or a tamarix chinensis forest and/or a salt marsh becomes a beach, or a newly added region, for example, a newly added beach, a salt marsh or a tamarix chinensis forest, or the like.
In this embodiment, the first preset period may include a first preset sub-period in the germination period in spring, and may also include a preset period before the germination period; the meteorological data comprise air temperature, rainwater quantity, duration of raining and storm surge; the second preset time period comprises any time period after the first preset time period and before the defoliation period.
Because of the vegetation in the germination period, if weather is not suitable for germination, the production of the vegetation can be influenced, and even the vegetation can die. For example, vegetation in the germination stage may affect plant germination, or shoot leaf growth, if the amount of rain is large during the germination stage. In addition, before or during the germination period, if the air temperature is low, the plant may be frozen to damage the plant, which affects the growth of the plant.
Thus, in this embodiment, meteorological data is acquired for a first preset period of time to characterize the meteorological data of plant germination.
In some exemplary embodiments, the first preset period may further include a second preset sub-period lasting from before the germination period to the end of the germination period; the first near-image data may be near-image data continuously collected in a second preset sub-period, and in the second preset sub-period, image information continuously changed in the whole period from germination to foliation of the plant may be covered by collecting the first near-image.
As an alternative embodiment, the first preset sub-period may be any duration covering the period of germination of the plant, for example, 3-10 days after the start of emergence of the bud pack, without limitation in this embodiment. The second preset sub-period may be any length of time that the cap has not germinated to grow into leaves, for example, may be a length of time between when or before bud packets appear and leaves grow to more than 80% of the average leaf forming size. In this embodiment, it may be 10-30 days.
And second near-image data of the mature region can be acquired in a second preset subperiod and used for determining the germination state of the mature region so as to determine the specific carbon sequestration capacity and carbon sequestration quantity of the mature region in the next year.
As an exemplary embodiment, the first near image data and the second near image data may be unmanned aerial vehicle aerial image data, for example, image data that may be photographed 5-100 meters above a coastal wetland, and may also be unmanned aerial vehicle laser point cloud data.
S20, inputting the first near image data into the first feature extraction network, and extracting static features and time sequence change features of the evolution region. As an exemplary embodiment, the first feature extraction network phrase extracts first near image data.
In this embodiment, since the first image data and the second image data are all captured on the whole coastal wetland, where the first image data may include a mature region, and the second image data may include an evolution region, in order to more accurately distinguish the image data of the evolution region from the image data of the mature region, in this embodiment, the first image data and the second image data may be classified by using a classification model, and the evolution region and the mature region may be accurately determined in the first image data and the second image data.
As an exemplary embodiment, the classification may be performed using a pre-trained classification network to obtain different types of coastal wetlands, and in an exemplary embodiment, the classification network model may be implemented using Python language, which is trained by a near-earth image of a real mature region and a near-earth image of an evolution region. Of course, implementation of the classification network model in other languages is not limiting in this embodiment. The specific training process comprises the following steps: firstly, randomly initializing all parameters of a model, inputting training data, inputting the training data into the classification network model for forward propagation, and obtaining output; then, calculating the loss of the model at the moment by using the constructed discrimination loss function and the classification loss function respectively; model parameters are updated by back propagation and the accuracy of the current model is tested. And in a certain training round number, model parameters are continuously updated through back propagation, and the model is stored when the current optimal precision is broken through each time, so that the finally trained network model can be obtained. In an alternative embodiment, the training parameters are set as follows: training round is 200, learning rate is 0.005, and random gradient descent is used as optimization function. Thus precisely dividing the development region and the maturation region.
The near-image data corresponding to the evolution regions in the classification result is input into the first feature extraction network to extract static features and time sequence change features of the evolution regions, in this embodiment, the static features may include current growth states of plants in near-ground images of each evolution region, for example, features such as colors, germination rates, growth states of buds of the evolution regions, and the time sequence change features may be features such as color change rates of the evolution regions, growth rates of buds of the evolution regions, and the like.
In this embodiment, the current and growth states of the bud are characterized by static and time-varying features.
S30, inputting the meteorological data to the second feature extraction network, and extracting the meteorological features. As an exemplary embodiment, the weather data may include air temperature, amount of rain, duration of rain, wind speed, and storm surge. As an exemplary embodiment, the actually measured meteorological data may be divided and sampled according to a sliding time window with a first preset length, so as to obtain meteorological data in a plurality of window intervals; determining a two-dimensional feature vector under each preset time scale in each window interval based on the data attribute and the data value of the meteorological data; and carrying out data standardization on the two-dimensional feature vector to obtain meteorological data.
By way of example, the data may be sub-sampled in accordance with a sliding window of length k; after dividing and sampling to obtain a plurality of window intervals, obtaining a two-dimensional feature vector for the actually measured meteorological data of each preset time scale xi in each window interval to represent the meteorological information of the preset time scale; illustratively, one dimension of the two-dimensional feature vector is time xi and the other dimension is a numerical value; specifically, the two-dimensional feature vector may be in the form of [ temperature i-k+1, … …, temperature i ], [ rainfall i-k+1, … …, rainfall i ], [ wind speed i-k+1, … …, wind speed i ], [ storm tide height i-k+1, … …, storm tide height i ].
In this embodiment, the weather features in the weather data of the plant germination period are extracted through the second feature extraction network, and may be classified into weather feature 1, weather feature 2, … … weather feature n according to the weather data type, for example, may include a temperature feature sequence, a rainfall feature sequence, a wind speed feature sequence, a wind high tide height feature sequence, and the like.
In this embodiment, after the first feature extraction network and the second feature extraction network extract the corresponding features, the features extracted by the first feature extraction network and the second feature extraction network are also aligned in time sequence, so as to align the meteorological features and the near-image features with each other in time.
S40, taking the static characteristics, the time sequence change characteristics and the meteorological characteristics as inputs of the maturity prediction model, taking the maturity data as outputs, and training the maturity prediction model for multiple times until the loss function is smaller than a preset threshold value, so as to obtain a trained maturity prediction model.
As an exemplary embodiment, the maturity prediction model may employ one or more of a BP neural network, a recurrent neural network, a convolutional neural network. Fig. 3 illustrates a BP neural network as an example.
Specifically, based on the static characteristics, characteristic values of time sequence change characteristics and meteorological characteristics are used as input layers of the BP neural network, output layers are maturity data, the middle layer is selected according to requirements, a maturity prediction model is built, training of the maturity prediction model is conducted, the maturity prediction model is obtained through training, and therefore maturity in one year in the future or maturity in one quarter or two quarters in the future is predicted.
As shown in fig. 3, in the present embodiment, the static features may include a plurality of static features such as static feature 1, static feature 2 … …, static feature n, etc., for example, features such as color of evolution region, germination rate, growth state of bud, etc. The time sequence variation characteristics can also comprise a plurality of time sequence variation characteristics such as a time sequence variation characteristic 1, a time sequence variation characteristic 2, a time sequence variation characteristic n and the like, for example, characteristics such as color variation rate characteristics of an evolution region, growth rate of buds and the like.
S50, inputting the maturity data, the meteorological features and the second near image data into the evaluation model, and training the evaluation model for multiple times to obtain a trained carbon sink evaluation model.
As an exemplary embodiment, the maturity data may be employed in the sample data during a training stage, and the maturity data may be a prediction result of a maturity prediction model during an evaluation stage. In this embodiment, the second near-field image data input by the evaluation model may be a near-field image of the mature region obtained by classifying the model.
As an exemplary embodiment, the assessment model may include a carbon sink capacity assessment model in which maturity data, characteristic values of meteorological features, and vegetation transformation features extracted from second near image data are variables of carbon sink capacity,
in this embodiment, the evaluation model may be a linear regression model, or a decision tree model or a logistic regression model, and in this embodiment, a linear regression model is used as an example for explanation. For example, a multiple linear regression model may be used for carbon sink assessment, which in this embodiment includes a third feature extraction network and a linear regression model. The third feature extraction network is used for extracting vegetation change features in the first preset time period in the second near-earth image so as to represent the vegetation germination rate or death rate in the mature region. In this embodiment, some special weather conditions affect the germination of vegetation, such as continuous rain during the germination period, cold flow during the germination period or before the germination, storm surge in spring, etc., which may affect the germination of vegetation, and the arrangement may cause a part of vegetation to fail to germinate and die, thereby continuously affecting carbon sequestration capacity and carbon sequestration. Thus, in this embodiment, the vegetation change characteristics of the mature region are also considered.
In this embodiment, the maturity data, the characteristic value of the meteorological feature and the vegetation transformation feature extracted from the second near-image data may be used as regression factors, and the carbon sink evaluation result may be used as a regression factor to perform multiple linear regression fitting, so as to obtain parameters of the multiple linear model, and complete training of the evaluation model. The assessment model may be used as a carbon sink assessment model for the overall coastal wetland region (including the mature region and the evolution region).
As an exemplary embodiment, the multiple linear regression model may employ the following formula:
‘
wherein a is 0 Is an intrinsic coefficient, a 1 、a 2 ……a n And b 1 、b 2 ……b n Respectively to be fitted with coefficient, x 1 、x 2 ……x n In order to evaluate the input data of the model, y1 … … yn corresponds to the carbon sink or the weight coefficient of the carbon sink capacity of each coastal wetland type respectively.
Taking an evolution area as a conversion area, converting the first type of coastal wetland into the second type of coastal wetland, wherein the weight coefficient is the weight coefficient of the carbon sink capacity of the coastal wetland type after conversion, the type after conversion is the first type of coastal wetland, the type before conversion is the second type of coastal wetland, the preset carbon sink capacity of the first type of coastal wetland is A, the preset carbon sink capacity of the second type of coastal wetland is B, when the carbon sink capacity of the conversion area is evaluated, multiplying the weight coefficient by A is needed to obtain the carbon sink capacity of the first type in the conversion process, and multiplying the weight coefficient by B is needed to obtain the carbon sink capacity of the second type in the conversion process, and taking the sum of the carbon sink capacity of the first type and the carbon sink capacity of the second type as the carbon sink evaluation result of the conversion area.
As another alternative embodiment, in order to more accurately evaluate the carbon sink, the evolution area and the mature area may be fitted respectively, in this embodiment, the evolution area may be fitted by using a multiple linear regression model, and by way of example, the doneness data of the evolution area and the characteristic value of the meteorological feature are used as regression factors, and the carbon sink evaluation result of the evolution area is used as a regressed factor, and multiple linear regression is performed to obtain an evaluation model of the evolution area and the mature area; fitting may be performed using a unitary polynomial for the mature region. By way of example, the vegetation change characteristics are taken as independent variables, the carbon sink capacity or the carbon sink effect of the mature region is taken as dependent variables to perform fitting of a unitary polynomial, and after the fitting is completed, the outputs of the vegetation change characteristics and the carbon sink effect of the mature region are added at the same time sequence, so that the carbon sink effect of the whole coastal wetland region can be obtained.
As an alternative embodiment, as shown in fig. 3, the first feature extraction network includes a first branch feature extraction network and a second branch feature extraction network, and a first timing feature extraction layer; the first near-field image data comprise near-ground three-dimensional point cloud data and aerial image data; the first near-image data includes near-ground three-dimensional point cloud data and aerial image data. As shown in fig. 4, the inputting the first near image data into the first feature extraction network, extracting the static feature and the time-series change feature of the evolution region includes:
S201, inputting the near-earth three-dimensional point cloud data into a first branch feature extraction network to extract the three-dimensional network features of the evolution region. As an exemplary embodiment, the near-ground three-dimensional point cloud data may automatically perform point cloud data collection on the established actual coastal wetland by using an automobile, an unmanned plane, or the like equipped with a point cloud data collection device, which may be a laser scanner, a depth camera, a binocular camera, or the like.
In this embodiment, classification of three-dimensional point cloud data may be based on projection methods, such as voxel grid-based methods and multiview-based methods. To accommodate the characteristics of three-dimensional point cloud data, methods based on the original point cloud, such as a multi-layer perceptron, may also be used; a convolutional neural network; a graph convolutional neural network; attention mechanisms, etc. In the embodiment, a point cloud semantic segmentation algorithm based on a deep learning model adopts a RandLA-Net algorithm to segment point cloud semantics. In this embodiment, the semantic division network may be used to perform semantic division on the point cloud data, and in this embodiment, the coastal wetland may be divided into tamarix according to vegetation height, salt marsh and mud flat.
The process of first branch feature extraction network establishment may include:
the semantic segmentation labels of the coastal wetland point cloud data samples are selected, and the coastal wetland cloud data samples can be labeled according to the types of coastal wetlands, such as trees, herbs and beaches, as main elements of the semantic segmentation labels.
And selecting a point cloud semantic segmentation algorithm based on the deep learning model, and training the point cloud segmentation model on the labeled coastal wetland point cloud data sample. By means of the method, the random sampling is adopted to obtain the point sampling and the local characteristic aggregation module is adopted to conduct the local characteristic learning of the point cloud data, the calculation rate is improved on the basis of guaranteeing the semantic segmentation accuracy of the point cloud data, and the method is suitable for the point cloud data processing of the large-scale coastal wetland. And setting a precision threshold value of the point cloud segmentation model training, and stopping calculating when the precision reaches the threshold value to obtain the coastal wetland semantic segmentation model. Three-dimensional point cloud data can be classified based on the semantic segmentation model of the coastal wetland to obtain plaques of the coastal wetland of each type.
Based on the classification result, a coastal wetland spatial network is constructed, in this embodiment, the ground surface may be selected as a reference surface, and the height 1 meter above the highest position of the coastal wetland and above the highest position of the coastal wetland may be selected as the vertical height, thereby determining the coastal wetland space. And after the littoral wet land meshing space is obtained, a certain point in the littoral wet land point cloud data exists in the voxel grid, the voxel is considered to exist in the littoral wet land, and the type of the littoral wet land to which the point belongs is determined by the semantic segmentation result.
After obtaining the voxels, the sparseness of the current type of coastal wetland and the distribution state of vegetation in the current area can be determined based on the distance between the voxels in each classification result.
In this embodiment, the three-dimensional network characteristics may include a type of coastal wetland and a spatial distribution characteristic of vegetation within the coastal wetland under the type, and exemplary spatial distribution characteristics may include a sparseness of the coastal wetland under the type and a distribution state of vegetation within the current area.
S202, inputting the aerial image data into the second branch feature extraction network to extract color space features of the evolution region. As an exemplary embodiment, the second branch feature extraction network may employ a CNN network, which may include 2 convolution layers, 2 activation layers, 1 batch normalization layer, and 1 maximum pooling layer, and extract color space features of each pixel in the aerial image data using one-dimensional operations. In this embodiment, the color of vegetation in the coastal wetland in the defoliation zone during germination varies with the germination status. Thus, in the present embodiment, the color space features are extracted to characterize the vegetation germination status of the current area.
S203, aligning the three-dimensional network feature and the color space feature to obtain the static feature. As an exemplary embodiment, since the three-dimensional point cloud data and the near-image data may not be acquired by the same device, or may have different spatial scales, in this embodiment, the three-dimensional network feature and the color space feature need to be aligned, in this embodiment, the three-dimensional network feature and the color space feature need to be registered in geographic information, and in this embodiment, the time domain registration needs to be registered, and in addition, the resolution registration needs to be performed, so as to ensure that the three-dimensional network feature and the color space feature can be matched in a one-to-one correspondence manner, in this embodiment, the three-dimensional network feature and the color space feature need to be aligned, so as to fuse the three-dimensional network feature and the color space feature to obtain the coastal wetland feature with three-dimensional space color, and it may be understood that the three-dimensional network feature is colored based on the color space feature.
In this embodiment, a three-dimensional point cloud data and an aerial image data are used as a set of near-image data, three-dimensional network features and color space features in the set of near-earth images are extracted through a first branch feature extraction network and a second branch feature extraction network, and then alignment fusion is performed to obtain static features of the set of near-earth images.
Because the near-image information can only acquire two-dimensional features, especially more accurate color features or spectrum features, and the point cloud data can acquire vertical space features, namely three-dimensional features, in the embodiment, three-dimensional network features are extracted based on the three-dimensional point cloud data, two-dimensional color space features are extracted based on the near-image data, and the three-dimensional network features and the color space features are aligned to fuse colors and the three-dimensional network features, so that the color features with three-dimensional space are obtained, and the growth state of the current vegetation is more comprehensively represented. More accurate data is provided for the evaluation and prediction of maturity, carbon sink effect and carbon sink capacity.
S204, inputting the static features into the time sequence feature extraction network, and extracting the time sequence change features. As an exemplary embodiment, the static feature may be a set of static image features having a time-series relationship extracted by a plurality of sets of near image data based on a time-series arrangement. The time sequence feature extraction network may adopt a cyclic neural network, in this embodiment, for example, a long-short-term memory network or a GRU network, in this embodiment, the time sequence change feature of the static feature in time sequence is extracted by the time sequence feature extraction network, in this embodiment, the accumulated feature based on time sequence may be extracted, and the change feature along with time sequence may be extracted.
For example, in this embodiment, the time sequence variation feature may include a feature such as a variation amount and a variation rate of a color feature, and may also include a three-dimensional network feature, that is, a feature such as a variation amount and a variation rate of a three-dimensional volume with time sequence.
As an alternative embodiment, as shown in fig. 5, the second feature extraction network includes a weather feature extraction network, a second time sequence feature network, and an attention network. In this embodiment, the meteorological feature extraction network may comprise a convolutional network. The meteorological data is split and reconstructed into two-dimensional feature vectors to represent the meteorological change information at the moment. Preliminary features in the meteorological data are extracted through a convolution network. Because the time sequence change characteristics of the weather change cannot be considered by the convolution network, a second time sequence characteristic network is added into the second characteristic extraction network to capture and learn the time accumulation effect of the weather change sequence, and the self-adaptive weighting fusion is carried out on different information of the characteristic vector by introducing a multi-head attention mechanism, so that not only can the non-important characteristics be filtered, but also various dependency relations in the sequence can be captured. Specifically, the method comprises the following steps:
S301, inputting the meteorological data into the meteorological feature extraction network to extract multiple types of meteorological features. As an exemplary embodiment, the data may be sub-sampled in accordance with a sliding window of length k; after dividing and sampling to obtain a plurality of window intervals, obtaining a two-dimensional feature vector for the actually measured meteorological data of each preset time scale xi in each window interval to represent the meteorological information of the preset time scale; illustratively, one dimension of the two-dimensional feature vector is time xi and the other dimension is a numerical value; specifically, the two-dimensional feature vector may be in the form of [ temperature i-k+1, … …, temperature i ], [ rainfall i-k+1, … …, rainfall i ], [ wind speed i-k+1, … …, wind speed i ], [ storm tide height i-k+1, … …, storm tide height i ]. And extracting meteorological features in each two-dimensional feature vector through a CNN network.
S302, inputting the multi-type meteorological features into the second time sequence feature network to extract the meteorological accumulated change features. In this embodiment, the weather variation feature of each type of weather sequence data may be extracted by the second time sequence feature extraction network, respectively. Because the germination process and vegetation growth state need to meet a certain climate condition and a time accumulation change process, and complex coupling relations exist under various fine meteorological conditions in the accumulation change process, in order to determine the time accumulation effect of the time accumulation change process and the complex coupling relations for representing each type of meteorological sequence data, in the application, when model training is carried out, the characteristic extraction of the meteorological change characteristics of each type of meteorological sequence data is realized by using a gating circulation unit. Wherein the retention of history information and the forgetting of unnecessary information are realized mainly by the update gate and the reset gate
S303, capturing the weighting weight of the weather accumulated change feature through the attention layer to obtain the weather accumulated change feature with the weighting weight. And learning the accumulated change characteristics in time and the weighting weights of the accumulated change characteristics of the meteorological change under the mutual coupling of a plurality of the meteorological change characteristics. Learning a plurality of weather accumulated change characteristics in the model training process; because of the complex coupling relationship in the germination process and under various meteorological conditions in the vegetation growth state, the complex relationship of the multidimensional sequence characteristic data is processed through a multi-head attention mechanism. And adjusting the weighting weights of the meteorological change characteristics through a plurality of groups of attention mechanisms so as to acquire the relation of different meteorological characteristic sequences. And finally, fusing the weighted subsequences through a linear layer to form an integral output sequence. Thereby improving the significance level of the meteorological features influencing the germination of vegetation and better capturing the correlation of the influence of various types of meteorological features on the germination of the vegetation.
The embodiment of the application also provides a coastal wetland carbon sink assessment method, which can comprise the following steps as shown in fig. 6:
s100, acquiring meteorological data of the coastal wetland to be evaluated in a first preset period, first near-image data of an evolution region of the coastal wetland and second near-image data of a mature region of the coastal wetland.
As an exemplary embodiment, the first preset period may include a first preset sub-period within the spring germination period, and may also include a preset period before the germination period; the meteorological data comprise air temperature, rainwater quantity, duration of raining and storm surge.
The first preset time period may further include a second preset sub-period lasting from before the germination period to after the germination period; the first near-image data may be near-image data continuously collected in a second preset sub-period, and in the second preset sub-period, image information continuously changed in the whole period from germination to foliation of the plant may be covered by collecting the first near-image.
S200, inputting the first near image data into the first feature extraction network, and extracting static features and time sequence change features of the evolution region. Specifically, the description of the static feature and the time sequence variation feature of the first near image data is extracted by using the first feature extraction network in the above embodiment.
S300, inputting the meteorological data to the second feature extraction network, and extracting the meteorological features. See in particular the description of the meteorological features of the meteorological data extracted using the second feature extraction network in the above embodiments.
S400, inputting the static characteristic, the time sequence change characteristic and the meteorological characteristic into the maturity prediction model to obtain the maturity of the evolution region. In this embodiment, the maturity prediction model may be a maturity prediction model obtained based on training in the above embodiment, and the maturity of the evolution region may be obtained by inputting the static feature, the time-series variation feature, and the meteorological feature into the trained maturity prediction model.
In this embodiment, the maturity may include the maturity of tamarix forest in the net change region from the non-coastal wetland to the coastal wetland, for example, the maturity of its salt biogas in the net change region from the non-coastal wetland to the salt biogas, or the maturity of tamarix forest in the net change region from the non-coastal wetland to the tamarix forest. The conversion maturity of the conversion region from the first type to the second type may also be included, and exemplified by the maturity of salt-land in the conversion region from beach to salt-land, or the maturity of beach in the conversion region from salt-land to beach.
In this embodiment, since different types of coastal wetlands often have different carbon sink capacities, the biomass quantity, the stability degree and the like of the new coastal wetlands do not reach the degree of the mature coastal wetlands, the initial carbon sink capacity of the newly increased coastal wetlands may be 40-60% of the preset carbon sink capacity, and as the maturity of the evolving coastal wetlands increases, the carbon sink capacity of the newly increased coastal wetlands is continuously increased until the carbon sink capacity of the mature coastal wetlands is reached, so in this embodiment, the maturity may be used as a weight coefficient of the carbon sink capacity of the coastal wetlands after the evolution in the current evolution area, so as to obtain the real-time carbon sink capacity of the coastal wetlands after the evolution in the current evolution area.
S500, inputting the maturity, the meteorological features and the second near-image data into a trained evaluation model to obtain a carbon sink evaluation result of the coastal wetland. After the maturity of the evolution region is obtained, the maturity, meteorological features, and image features of vegetation of the maturity region can be used as factors affecting the carbon sink effect to evaluate the carbon sink capacity.
In this embodiment, the evaluation model may be a linear regression model, or a decision tree model or a logistic regression model, and is illustrated by taking a multiple linear regression model as an example:
the multiple linear regression model may employ the following formula:
，
wherein a is 0 Is an intrinsic coefficient, a 1 、a 2 ……a n And b 1 、b 2 ……b n Fitting coefficients, x respectively 1 、x 2 ……x n In order to evaluate the input data of the model, y1 … … yn corresponds to the carbon sink effect or the weight coefficient of the carbon sink capacity of each coastal wetland type respectively.
Taking an evolution area as a conversion area, converting the first type of coastal wetland into the second type of coastal wetland, wherein the weight coefficient is the weight coefficient of the carbon sink capacity of the coastal wetland type after conversion, the type after conversion is the first type of coastal wetland, the type before conversion is the second type of coastal wetland, the preset carbon sink capacity of the first type of coastal wetland is A, the preset carbon sink capacity of the second type of coastal wetland is B, when the carbon sink capacity of the conversion area is evaluated, multiplying the weight coefficient by A is needed to obtain the carbon sink capacity of the first type in the conversion process, and multiplying the weight coefficient by B is needed to obtain the carbon sink capacity of the second type in the conversion process, and taking the sum of the carbon sink capacity of the first type and the carbon sink capacity of the second type as the carbon sink evaluation result of the conversion area.
Illustratively, in this embodiment, tamarix chinensis forests and/or salt-biogas are included in a first type of wetland; the second type of wetland comprises a beach as an example, namely, the conversion from salt marsh or tamarix forests to the beach is specifically described as an example:
the weight coefficient of tamarix chinensis willow and/or salt marsh is y1, and the weight coefficient of beach is 1-y1; in the process of converting tamarix chinensis forests or salt marsh to mud flat, the production capacity gradually decreases and gradually changes into the input carbon sink capacity. Therefore, when evaluating the carbon sink, the real-time carbon sink capability of the tamarix chinensis willow and/or the salt marsh in the conversion area can be obtained by multiplying the preset carbon sink capability by the corresponding weight coefficient. The carbon sink capacity of the conversion area can be the sum of the product of the carbon sink capacities A and y1 of the first type of coastal wetland and the product of the carbon sink capacities B and 1-y1 of the second type of coastal wetland.
Other types of coastal wetlands can also be calculated by referring to the embodiment, and finally the carbon sink effect of the whole coastal wetland is obtained.
As an alternative practice, the evolution rate of the evolution region may be faster or slower under the influence of natural environment and human factors, and the carbon sink capacity of the region corresponding to the faster evolution rate is closer to or reaches the preset carbon sink capacity faster than the carbon sink capacity of the type after evolution, and the carbon sink capacity of the region corresponding to the slower evolution rate reaches the preset carbon sink capacity slower or fails to reach the carbon sink capacity corresponding to the predicted maturity in a short time. Therefore, in order to more accurately evaluate or predict the carbon sink capability of the evolution region, in the present embodiment, the weight coefficient of the evolution region may be adjusted based on the evolution rate of the evolution region.
While predicting or determining the evolution rate by natural environment and human factors often requires a large number of data dimensions, e.g., various meteorological data, seawater nutrient salts, etc., as well as various uncertain influencing factors, e.g., factors of human activity. The evolution rate is calculated by using the data, so that the accuracy is not high, the calculated amount and the data amount are huge, and therefore, in order to more accurately determine the evolution rate, in the embodiment, the maturity data in a plurality of continuous preset periods of the evolution region in the coastal wetland to be evaluated are obtained; determining an evolution rate of the evolution region based on a plurality of consecutive maturity data; and adjusting the carbon sink weight coefficient of the evolution region based on the evolution rate.
As an exemplary effort, the preset period may be a preset maturity prediction period of one quarter, two quarters, one year, etc. In this embodiment, an average value of one or more maturity measured data in a plurality of maturity prediction periods may be obtained as average maturity data of a current evolution region, and a maturity growth rate or a change rate may be determined based on the plurality of maturity data. In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.
The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in each embodiment of the present invention may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk or an optical disk, or the like, which can store program codes.
Alternatively, the above-described integrated units of the present invention may be stored in a computer-readable storage medium if implemented in the form of software functional modules and sold or used as separate products. Based on such understanding, the technical solutions of the embodiments of the present invention may be embodied in essence or a part contributing to the prior art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a removable storage device, ROM, RAM, magnetic or optical disk, or other medium capable of storing program code.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. The method for constructing the coastal wetland carbon sink assessment model is characterized by comprising the following steps of: the system comprises a first feature extraction network, a second feature extraction network, a maturity prediction model and an assessment model;
acquiring sample data, wherein the sample data comprises meteorological data of a coastal wetland in a first preset period, first near-image data of an evolution region of the coastal wetland, second near-image data of a mature region of the coastal wetland and maturity data of the evolution region in a second preset period, the evolution region comprises conversion regions among different types of wetlands and net change regions of the coastal wetland except the conversion regions, and the first preset period is before the second preset period in a time sequence dimension;
inputting the first near-image data to the first feature extraction network, and extracting static features and time sequence change features of the evolution region;
inputting the meteorological data to the second feature extraction network to extract meteorological features;
taking the static characteristics, the time sequence change characteristics and the meteorological characteristics as the input of the maturity prediction model, and taking the maturity data as the output to train the maturity prediction model for a plurality of times until the loss function is smaller than a preset threshold value, so as to obtain a trained maturity prediction model;
And inputting the maturity data, the meteorological features and the second near image data into the evaluation model, and training the evaluation model for multiple times to obtain a trained carbon sink evaluation model.
2. The method for constructing a coastal wetland carbon sink assessment model according to claim 1, wherein the first feature extraction network comprises a first branch feature extraction network and a second branch feature extraction network, and a first timing feature extraction layer;
the first near-field image data comprise near-ground three-dimensional point cloud data and aerial image data;
the inputting the first near image data to the first feature extraction network, extracting static features and time sequence variation features of the evolution region comprises:
inputting the near-ground three-dimensional point cloud data into a first branch feature extraction network to extract three-dimensional network features of the evolution region;
inputting the aerial image data into the second branch feature extraction network to extract color space features of the evolution region;
aligning the three-dimensional network feature with the color space feature to obtain the static feature;
and inputting the static features into the time sequence feature extraction network to extract the time sequence change features.
3. The method for constructing a coastal wetland carbon sink assessment model according to claim 2, wherein the aligning the three-dimensional network feature and the color space feature to obtain the static feature further comprises:
performing first clustering on the three-dimensional network characteristics to obtain first sparsity of plants in the evolution region;
performing second clustering on the color space features to obtain second sparsity of the evolution region;
and fusing the three-dimensional network feature and the color space feature based on the first sparsity and the second sparsity to obtain the static feature.
4. The method for constructing a coastal wetland carbon sink assessment model according to claim 1, wherein the second feature extraction network comprises a meteorological feature extraction layer, a second time sequence feature layer and an attention layer;
inputting the meteorological data into the meteorological feature extraction layer to extract multiple types of meteorological features;
inputting the multi-type meteorological features to the second time sequence feature layer to extract meteorological accumulated change features;
capturing the weighting weight of the weather accumulated change feature through the attention layer to obtain the weather accumulated change feature with the weighting weight.
5. The method for constructing a coastal wetland carbon sink assessment model according to claim 1, wherein the assessment model comprises a third feature extraction network and a linear regression model;
inputting the maturity data, the meteorological features and the second near image data into the evaluation model, performing multiple rounds of training on the evaluation model, and obtaining a trained carbon sink evaluation model comprises:
extracting vegetation change features in the first preset period in the second near-earth image based on the third feature extraction network;
and constructing the linear regression model by using the maturity data, the meteorological features and the vegetation change features as variables.
6. The method for constructing a coastal wetland carbon sink assessment model according to claim 1, wherein the first preset period comprises a preset period in a spring germination period; the meteorological data comprise air temperature, rainwater quantity, duration of raining and storm surge;
the second preset time period comprises any time period after the first preset time period and before the defoliation period.
7. A method for evaluating carbon sink of a coastal wetland, characterized in that the method for evaluating adopts a carbon sink evaluation model obtained by the method for constructing a carbon sink evaluation model of a coastal wetland according to any one of claims 1 to 6, and the method for evaluating comprises the following steps:
Acquiring meteorological data of a coastal wetland to be evaluated in a first preset period, first near-image data of an evolution region of the coastal wetland and second near-image data of a mature region of the coastal wetland;
inputting the first near-image data to the first feature extraction network, and extracting static features and time sequence change features of the evolution region;
inputting the meteorological data to the second feature extraction network to extract the meteorological features;
inputting the static characteristics, the time sequence change characteristics and the meteorological characteristics into the maturity prediction model to obtain the maturity of the evolution region;
and inputting the maturity, the meteorological features and the second near-image data into a trained evaluation model to obtain a carbon sink evaluation result of the coastal wetland.
8. The coastal wetland carbon sink assessment method as claimed in claim 7, comprising: the evaluation model comprises a third feature extraction network and a linear regression model;
inputting the maturity, the meteorological features and the second near-image data into a trained evaluation model, and obtaining a carbon sink evaluation result of the coastal wetland comprises the following steps:
Inputting the second near-earth image data into the third characteristic extraction network to extract vegetation change characteristics in the first preset period in the second near-earth image;
inputting the maturity data, the meteorological features and the vegetation change features into the linear regression model to obtain a carbon sink weight coefficient of the coastal wetland;
and determining carbon sinks of the coastal wetland based on the preset carbon sink capacity corresponding to the coastal wetland type and the carbon sink weight coefficient.
9. The coastal wetland carbon sink assessment method as defined in claim 8, further comprising:
acquiring maturity data in a plurality of continuous preset periods of an evolution region in the coastal wetland to be evaluated;
determining an evolution rate of the evolution region based on a plurality of consecutive maturity data;
and adjusting the carbon sink weight coefficient of the evolution region based on the evolution rate.
10. An electronic device, the electronic device comprising:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores a computer program executable by the at least one processor to enable the at least one processor to perform the coastal wetland carbon sink assessment model construction method according to any one of claims 1 to 6 and/or the coastal wetland carbon sink assessment method according to any one of claims 7 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311118002.6A CN116842351B (en) | 2023-09-01 | 2023-09-01 | Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311118002.6A CN116842351B (en) | 2023-09-01 | 2023-09-01 | Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116842351A true CN116842351A (en) | 2023-10-03 |
| CN116842351B CN116842351B (en) | 2023-11-10 |
Family
ID=88172858
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311118002.6A Active CN116842351B (en) | 2023-09-01 | 2023-09-01 | Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116842351B (en) |
Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200184214A1 (en) * | 2018-12-11 | 2020-06-11 | The Climate Corporation | Mapping soil properties with satellite data using machine learning approaches |
| CN111488902A (en) * | 2020-01-14 | 2020-08-04 | 沈阳农业大学 | Method and system for quantitatively estimating carbon reserves of ecosystem of primary coastal wetland |
| CN112765808A (en) * | 2021-01-15 | 2021-05-07 | 黄河勘测规划设计研究院有限公司 | Ecological drought monitoring and evaluating method |
| CN113128134A (en) * | 2021-06-17 | 2021-07-16 | 中国矿业大学(北京) | Mining area ecological environment evolution driving factor weight quantitative analysis method |
| CN113537702A (en) * | 2021-06-04 | 2021-10-22 | 中国石油大学(北京) | Geological dessert evaluation method, device, equipment and storage medium |
| CN114460013A (en) * | 2022-01-28 | 2022-05-10 | 自然资源部第一海洋研究所 | Coastal wetland vegetation ground biomass GAN model self-learning remote sensing inversion method |
| CN114943404A (en) * | 2022-03-30 | 2022-08-26 | 北京师范大学 | Wetland water shortage ecological risk assessment method based on ecosystem service balance |
| CN115019166A (en) * | 2022-05-24 | 2022-09-06 | 深圳大学 | Marsh wetland information extraction method, device, medium and terminal based on deep network model |
| US20220405864A1 (en) * | 2019-11-19 | 2022-12-22 | Zhejiang University | Crop yield estimation method based on deep temporal and spatial feature combined learning |
| CN115758801A (en) * | 2022-12-09 | 2023-03-07 | 成都市环境保护科学研究院 | High-resolution weather-driven planning carbon sink numerical evaluation method, system and terminal |
| CN115937692A (en) * | 2023-02-15 | 2023-04-07 | 山东省海洋资源与环境研究院(山东省海洋环境监测中心、山东省水产品质量检验中心) | Coastal wetland carbon sink effect evaluation method and system |
| CN116128377A (en) * | 2023-04-04 | 2023-05-16 | 山东省海洋资源与环境研究院(山东省海洋环境监测中心、山东省水产品质量检验中心) | Carbon sink effect evaluation method and device for offshore area and electronic equipment |
| CN116448984A (en) * | 2023-04-20 | 2023-07-18 | 北京师范大学 | Multi-module integrated device and method for indoor simulation experiment of universe in wetland |
| CN116485263A (en) * | 2023-04-25 | 2023-07-25 | 中国林业科学研究院亚热带林业研究所 | River wetland carbon sink monitoring and evaluating method |
-
2023
- 2023-09-01 CN CN202311118002.6A patent/CN116842351B/en active Active
Patent Citations (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200184214A1 (en) * | 2018-12-11 | 2020-06-11 | The Climate Corporation | Mapping soil properties with satellite data using machine learning approaches |
| US20220405864A1 (en) * | 2019-11-19 | 2022-12-22 | Zhejiang University | Crop yield estimation method based on deep temporal and spatial feature combined learning |
| CN111488902A (en) * | 2020-01-14 | 2020-08-04 | 沈阳农业大学 | Method and system for quantitatively estimating carbon reserves of ecosystem of primary coastal wetland |
| CN112765808A (en) * | 2021-01-15 | 2021-05-07 | 黄河勘测规划设计研究院有限公司 | Ecological drought monitoring and evaluating method |
| CN113537702A (en) * | 2021-06-04 | 2021-10-22 | 中国石油大学(北京) | Geological dessert evaluation method, device, equipment and storage medium |
| CN113128134A (en) * | 2021-06-17 | 2021-07-16 | 中国矿业大学(北京) | Mining area ecological environment evolution driving factor weight quantitative analysis method |
| CN114460013A (en) * | 2022-01-28 | 2022-05-10 | 自然资源部第一海洋研究所 | Coastal wetland vegetation ground biomass GAN model self-learning remote sensing inversion method |
| CN114943404A (en) * | 2022-03-30 | 2022-08-26 | 北京师范大学 | Wetland water shortage ecological risk assessment method based on ecosystem service balance |
| CN115019166A (en) * | 2022-05-24 | 2022-09-06 | 深圳大学 | Marsh wetland information extraction method, device, medium and terminal based on deep network model |
| CN115758801A (en) * | 2022-12-09 | 2023-03-07 | 成都市环境保护科学研究院 | High-resolution weather-driven planning carbon sink numerical evaluation method, system and terminal |
| CN115937692A (en) * | 2023-02-15 | 2023-04-07 | 山东省海洋资源与环境研究院(山东省海洋环境监测中心、山东省水产品质量检验中心) | Coastal wetland carbon sink effect evaluation method and system |
| CN116128377A (en) * | 2023-04-04 | 2023-05-16 | 山东省海洋资源与环境研究院(山东省海洋环境监测中心、山东省水产品质量检验中心) | Carbon sink effect evaluation method and device for offshore area and electronic equipment |
| CN116448984A (en) * | 2023-04-20 | 2023-07-18 | 北京师范大学 | Multi-module integrated device and method for indoor simulation experiment of universe in wetland |
| CN116485263A (en) * | 2023-04-25 | 2023-07-25 | 中国林业科学研究院亚热带林业研究所 | River wetland carbon sink monitoring and evaluating method |
Non-Patent Citations (4)
| Title |
|---|
| ZE ZHANG等: "Assessment of the impact of wetland changes on carbon storage in coastal urban agglomerations from 1990 to 2035 in support of SDG15.1", 《SCIENCE OF THE TOTAL ENVIRONMENT》, pages 1 - 15 * |
| 刘玉斌: "中国海岸带典型生态系统服务价值评估研究", 《中国博士学位论文全文数据库 工程科技Ⅰ辑》, vol. 2022, no. 1, pages 027 - 19 * |
| 李瑾璞: "基于InVEST模型的土地利用变化与生态系统碳储量研究——以雄安新区为例", 《中国优秀硕士学位论文全文数据库 基础科学辑》, vol. 2021, no. 1, pages 006 - 1547 * |
| 申霞等: "滨海盐沼净碳汇能力研究方法综述", 《生态学杂志》, vol. 41, no. 4, pages 792 - 803 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116842351B (en) | 2023-11-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Halme et al. | Utility of hyperspectral compared to multispectral remote sensing data in estimating forest biomass and structure variables in Finnish boreal forest | |
| CN113221765B (en) | Vegetation phenological period extraction method based on digital camera image effective pixels | |
| CN116128141B (en) | Storm surge prediction method and device, storage medium and electronic equipment | |
| Lee et al. | Applying machine learning methods to detect convection using Geostationary Operational Environmental Satellite-16 (GOES-16) advanced baseline imager (ABI) data | |
| CN111161362A (en) | Tea tree growth state spectral image identification method | |
| CN113033453A (en) | Method and system suitable for remote sensing identification of crop types in landscape crushing area | |
| CN109918449B (en) | Internet of things-based agricultural disaster information remote sensing extraction method and system | |
| CN112434569B (en) | Unmanned aerial vehicle thermal imaging system | |
| Balti et al. | Big data based architecture for drought forecasting using LSTM, ARIMA, and Prophet: Case study of the Jiangsu Province, China | |
| CN115457403A (en) | Intelligent crop identification method based on multi-type remote sensing images | |
| CN111476197A (en) | Oil palm identification and area extraction method and system based on multi-source satellite remote sensing image | |
| CN117114913A (en) | Intelligent agricultural data acquisition system based on big data | |
| CN116227758A (en) | Agricultural product maturity prediction method and system based on remote sensing technology and deep learning | |
| CN110046756B (en) | Short-term weather forecasting method based on wavelet denoising and Catboost | |
| CN114973024A (en) | Method and system for monitoring and evaluating comprehensive evaluation index of mechanical cotton picking and defoliating effect | |
| CN109960972B (en) | Agricultural and forestry crop identification method based on middle-high resolution time sequence remote sensing data | |
| He et al. | Recognition of soybean pods and yield prediction based on improved deep learning model | |
| CN116863341B (en) | Crop classification and identification method and system based on time sequence satellite remote sensing image | |
| CN116842351B (en) | Coastal wetland carbon sink assessment model construction method, assessment method and electronic equipment | |
| Wu et al. | A Dense Litchi Target Recognition Algorithm for Large Scenes | |
| CN111579565B (en) | Agricultural drought monitoring method, system and storage medium | |
| Kumawat et al. | Time-Variant Satellite Vegetation Classification Enabled by Hybrid Metaheuristic-Based Adaptive Time-Weighted Dynamic Time Warping | |
| Gunes et al. | Classification of hazelnuts with CNN based deep learning system | |
| CN113537645A (en) | Soybean yield prediction method based on machine learning fusion satellite and weather data | |
| CN112488230A (en) | Crop water stress degree judging method and device based on machine learning |
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 |