Rural livable data dynamic acquisition and management method based on multi-source data

Abstract

The invention relates to the technical field of rural livable evaluation, and discloses a rural livable data dynamic acquisition and management method based on multi-source data, comprising the following steps: step S101, obtaining coverage rates of different land types through an image semantic segmentation model; step S102, preprocessing to obtain a feature matrix; s103, obtaining the carbon reserve through the carbon reserve prediction model; step S104, obtaining the carbon flux through the carbon flux prediction model; s105, calculating livable scores according to the coverage rates, the carbon reserves and the carbon fluxes of different land use types of the country to be evaluated; the coverage rates of different land use types are identified through the image semantic segmentation model, time-space analysis is performed on the remote sensing data of the multiple sub-regions through the carbon reserve prediction model and the carbon flux prediction model, and the mapping relation between the remote sensing data and the carbon reserve and the carbon flux is established, so that the data acquisition difficulty is reduced, and the data acquisition efficiency is improved. And the livable score is obtained through comprehensive calculation, so that the evaluation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rural livability assessment, and more specifically, to a method for dynamically collecting and managing rural livability data based on multi-source data. Background Art

[0002] With the acceleration of global urbanization, the development and improvement of rural areas, as important areas for agricultural production and ecological environment protection, have gradually become the focus of social attention.

[0003] Traditional rural livability assessments usually rely on static data statistics and limited field surveys. Although they can provide information about the current situation of rural areas to a certain extent, they have the following defects: 1. Static data statistics have time lags and are difficult to reflect changes in the rural environment in real time, resulting in deviations in the assessment of rural livability; 2. Manual field surveys require a lot of manpower and material resources. This data collection method is not only inefficient, but also has limited coverage, especially in remote or environmentally complex areas. Data collection is difficult, which also leads to deviations in the assessment of rural livability; 3. Traditional rural livability assessments usually only focus on a certain indicator, such as vegetation coverage, land utilization rate, etc., lack of comprehensive analysis of the distribution of data in time and space, and it is difficult to fully reflect the livability of rural areas.

[0004] Therefore, a dynamic collection and management method of rural livability data based on multi-source data is urgently needed to solve the above problems. Summary of the invention

[0005] The present invention provides a method for dynamically collecting and managing rural livability data based on multi-source data, which solves the technical problems in the above-mentioned background technology.

[0006] The present invention provides a method for dynamically collecting and managing rural livability data based on multi-source data, comprising the following steps:

[0007] Step S101, collecting a bird's-eye view image of the village to be evaluated, and identifying the coverage of different land use types in the bird's-eye view image through an image semantic segmentation model;

[0008] Land use types include: housing land, agricultural land, industrial land, water resources and green land;

[0009] Step S102, dividing the village to be evaluated into M sub-areas, collecting remote sensing data of the M sub-areas at preset time intervals t within a preset time period T, and preprocessing the data to obtain a feature matrix;

[0010] The feature matrix includes M rows and N columns, and the element values ​​of the feature matrix are represented by feature vectors. The feature vector of the mth row and nth column is obtained by preprocessing the remote sensing data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0011] The feature vector includes 32 dimensional values, corresponding to the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil-adjusted vegetation index and normalized water index of the near-infrared band, blue band, green band and red band respectively;

[0012] Step S103, inputting the characteristic matrix into the carbon storage prediction model, and the output value represents the carbon storage of the village to be evaluated;

[0013] Step S104, inputting the characteristic matrix into the carbon flux prediction model, and the output value represents the carbon flux of the village to be evaluated;

[0014] Step S105, calculating the livability score of the village to be assessed based on the coverage rate, carbon storage and carbon flux of different land use types in the village to be assessed;

[0015] The livability score ranges from 0 to 1;

[0016] Step S106, if it is determined that the livability score of the village to be evaluated is greater than or equal to a preset livability score threshold, it means that the village to be evaluated is livable, otherwise it means that the village to be evaluated is not livable.

[0017] Furthermore, the number of sub-areas M, the preset time period T, the preset time interval t and the preset livability score threshold are all custom parameters.

[0018] Furthermore, the image semantic segmentation model is built based on the Mask2Former model, the sample labels of the training samples of the image semantic segmentation model are obtained through manual annotation, and pre-training is performed on a public remote sensing dataset before training the image semantic segmentation model.

[0019] Furthermore, the remote sensing data is preprocessed to obtain a feature vector, including the following steps:

[0020] Step S201, respectively obtaining the reflectance average value, maximum value, minimum value, standard deviation, and coefficient of variation of the near infrared band, blue light band, green light band, and red light band of the remote sensing data;

[0021] Where the coefficient of variation is equal to the standard deviation of reflectance divided by the mean reflectance;

[0022] Step S202, obtaining an enhanced vegetation index by calculating the average reflectance of the near infrared band, the blue light band and the red light band of the remote sensing data;

[0023] The calculation formula of the Enhanced Vegetation Index (EVI) is as follows:

[0024]

[0025] Among them, NIR, BLUE and RED represent the average reflectance of the near-infrared band, blue band and red band of remote sensing data respectively;

[0026] Step S203, calculating and obtaining the soil adjusted vegetation index according to the average reflectance of the near infrared band and the red light band of the remote sensing data;

[0027] The calculation formula of soil adjusted vegetation index SAVI is as follows:

[0028]

[0029] Where L afforest Indicates the coverage rate of green land;

[0030] Step S204, calculating a normalized water index based on the average reflectance of the near infrared band and the green band of the remote sensing data;

[0031] The calculation formula of normalized water index NDWI is as follows:

[0032]

[0033] Where GREEN represents the average reflectance of the green band of remote sensing data;

[0034] Step S205, normalize the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil adjusted vegetation index and normalized water index of the near infrared band, blue light band, green light band and red light band by Z-score normalization method to obtain a feature vector.

[0035] Furthermore, the carbon stock prediction model includes M hidden layers, each hidden layer includes N hidden units, the vth hidden unit of the uth hidden layer inputs the element value of the uth row and vth column of the feature matrix, and outputs an update vector, where 1≤u≤M, 1≤v≤N, and the update vector output by the Nth hidden unit of the Mth hidden layer is input to the first classifier, and the classification space of the first classifier represents the carbon stock of the village to be evaluated.

[0036] Furthermore, the calculation formula of the carbon storage prediction model includes:

[0037]

[0038] in represents the update vector of the output of the vth hidden unit in the uth hidden layer, Represents the element value of the u-th row and v-th column of the input feature matrix of the v-th hidden unit of the u-th hidden layer, and They represent the output gating vector, intermediate vector, candidate intermediate vector, input gating vector and forgetting gating vector of the vth hidden unit output of the uth hidden layer, respectively. represents the intermediate vector output by the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the transformation vector, represents the update vector of the output of the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the update vector of the output of the v-1th hidden unit in the uth hidden layer, and The same size, The element values ​​of are all assigned to 0. and denote the first weight parameter, the second weight parameter, the third weight parameter and the fourth weight parameter of the vth hidden unit of the uth hidden layer, respectively. b2 v u 、b3 v u and b4 v u They respectively represent the first bias parameter, second bias parameter, third bias parameter and fourth bias parameter of the vth hidden unit of the uth hidden layer, tanh represents the hyperbolic tangent activation function, sigmoid represents the sigmoid activation function, MLP represents the multilayer perceptron, concat represents the concatenation function, and ⊙ represents element-by-element multiplication.

[0039] Furthermore, the vegetation carbon stock of the sub-region is obtained by the biomass measurement method, and the soil carbon stock of the sub-region is obtained by the soil sampling method. The vegetation carbon stock and soil carbon stock of all sub-regions are added together as sample labels of training samples for training the carbon stock prediction model.

[0040] Furthermore, the composition and calculation formula of the carbon flux prediction model are the same as the carbon flux prediction model.

[0041] Furthermore, the carbon flux of the sub-region is obtained by the eddy covariance method as the sample label of the training sample for training the carbon flux prediction model.

[0042] Furthermore, the calculation formula for the livability score of the village to be evaluated is as follows:

[0043]

[0044] Type i represents the coverage rate of the i-th land use type among housing land, agricultural land, water resources and green land, r i Indicates the weight coefficient corresponding to the coverage rate of the i-th land use type and is a custom parameter. industry represents the coverage rate of industrial land, U represents the weight coefficient corresponding to the coverage rate of industrial land, and U is a custom parameter with a value range of 0.5 to 1. 储量 and They represent the carbon storage of the village to be assessed and the maximum value of historical carbon storage, C 通量 and They represent the maximum carbon flux and historical carbon flux of the village to be evaluated, w1 and w2 represent the first weight coefficient and the second weight coefficient, respectively. Both w1 and w2 are custom parameters, and their sum is 1. σ represents the sigmoid activation function, which is used to control the value range of the livability score between 0 and 1.

[0045] The beneficial effects of the present invention are as follows: the present invention identifies the coverage rates of different land use types in overhead images through an image semantic segmentation model, and performs spatiotemporal analysis on remote sensing data of multiple sub-regions through a carbon reserve prediction model and a carbon flux prediction model, thereby establishing a mapping relationship between remote sensing data and carbon reserves and carbon fluxes, thereby reducing the difficulty of data collection, and obtaining a livability score for the village to be evaluated based on the above parameters through comprehensive calculation, thereby achieving a comprehensive evaluation of the livability of the village and improving the accuracy of the evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a flow chart of a method for dynamically collecting and managing rural livability data based on multi-source data of the present invention;

[0047] Figure 2 It is a flow chart of the preprocessing of the present invention to obtain the feature vector. DETAILED DESCRIPTION

[0048] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0049] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0050] like Figure 1-2 As shown, a method for dynamically collecting and managing rural livability data based on multi-source data includes the following steps:

[0051] Step S101, collecting a bird's-eye view image of the village to be evaluated, and identifying the coverage of different land use types in the bird's-eye view image through an image semantic segmentation model;

[0052] Land use types include: housing land, agricultural land, industrial land, water resources and green land;

[0053] Step S102, dividing the village to be evaluated into M sub-areas, collecting remote sensing data of the M sub-areas at preset time intervals t within a preset time period T, and preprocessing the data to obtain a feature matrix;

[0054] The feature matrix includes M rows and N columns, and the element values ​​of the feature matrix are represented by feature vectors. The feature vector of the mth row and nth column is obtained by preprocessing the remote sensing data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0055] The feature vector includes 32 dimensional values, corresponding to the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil-adjusted vegetation index and normalized water index of the near-infrared band, blue band, green band and red band respectively;

[0056] Step S103, inputting the characteristic matrix into the carbon storage prediction model, and the output value represents the carbon storage of the village to be evaluated;

[0057] Step S104, inputting the characteristic matrix into the carbon flux prediction model, and the output value represents the carbon flux of the village to be evaluated;

[0058] Step S105, calculating the livability score of the village to be assessed based on the coverage rate, carbon storage and carbon flux of different land use types in the village to be assessed;

[0059] The livability score ranges from 0 to 1;

[0060] Step S106, if it is determined that the livability score of the village to be evaluated is greater than or equal to a preset livability score threshold, it means that the village to be evaluated is livable, otherwise it means that the village to be evaluated is not livable.

[0061] It should be noted that residential land is mainly used to reflect the population distribution of the village to be evaluated; agricultural land includes: farmland, orchards, etc., which is used to reflect the agricultural economic activities of the village to be evaluated; industrial land includes: factories, mining areas, etc., which are used to reflect the industrial economic activities of the village to be evaluated; water resources include: rivers, lakes, etc., which are used to reflect the water resource area of ​​the village to be evaluated; green land includes: forests, grasslands, wetlands, etc., which are used to reflect the green land area of ​​the village to be evaluated. In addition, land types can also be subdivided, such as industrial land can be subdivided, including light industry and heavy industry, etc. The coverage rate of different land types is closely related to the rural livability assessment.

[0062] It should be noted that the overhead image of the village to be evaluated can be obtained by taking pictures with a high-resolution camera mounted on an unmanned aerial vehicle. Since the coverage rate of different land use types will not change significantly in a short period of time, the present invention only collects the overhead image of the village to be evaluated once; and the carbon storage and carbon flux vary greatly over time and are affected by the weather. Therefore, the present invention divides the village to be evaluated into multiple sub-areas and collects remote sensing data multiple times to improve the prediction accuracy of carbon storage and carbon flux. In addition, the remote sensing data is represented by the reflectivity corresponding to the near-infrared band (760nm~900nm), blue light band (450nm~520nm), green light band (520nm~600nm) and red light band (630nm~690nm).

[0063] In one embodiment of the present invention, the number of sub-areas M, the preset time period T, the preset time interval t and the preset livability score threshold are all custom parameters. Preferably, the default value of the number of sub-areas M is 50, the default value of the preset time period T is 30 days, and the default value of the preset time interval t is 1 day, then N=T / t=30, and the default value of the preset livability score threshold is 0.5.

[0064] In one embodiment of the present invention, an image semantic segmentation model is constructed based on a Mask2Former model, and sample labels of training samples of the image semantic segmentation model are obtained through manual annotation. Before training the image semantic segmentation model, the model is pre-trained on a public remote sensing dataset. After pre-training, the model parameters are fine-tuned before training, which can improve the training speed of the model. The public remote sensing dataset can be Sentinel-2, Landsat, Google Earth, etc., and the manual annotation software can be QGIS, Labelbox, LabelImg, etc., which will not be repeated here.

[0065] In one embodiment of the present invention, the image semantic segmentation model can also be constructed based on models such as U-Net, DeepLab series (such as DeepLabV3+), SegNet, etc., and before the overhead image of the village to be evaluated is input into the image semantic segmentation model, it can also be pre-processed, such as unifying the resolution and normalizing it, which will not be elaborated here.

[0066] In one embodiment of the present invention, Figure 2 As shown, preprocessing the remote sensing data to obtain the feature vector includes the following steps:

[0067] Step S201, respectively obtaining the reflectance average value, maximum value, minimum value, standard deviation, and coefficient of variation of the near infrared band, blue light band, green light band, and red light band of the remote sensing data;

[0068] Where the coefficient of variation is equal to the standard deviation of reflectance divided by the mean reflectance;

[0069] Step S202, obtaining an enhanced vegetation index by calculating the average reflectance of the near infrared band, the blue light band and the red light band of the remote sensing data;

[0070] The calculation formula of the Enhanced Vegetation Index (EVI) is as follows:

[0071]

[0072] Among them, NIR, BLUE and RED represent the average reflectance of the near-infrared band, blue band and red band of remote sensing data respectively;

[0073] Step S203, calculating and obtaining the soil adjusted vegetation index according to the average reflectance of the near infrared band and the red light band of the remote sensing data;

[0074] The calculation formula of soil adjusted vegetation index SAVI is as follows:

[0075]

[0076] Where L afforest Indicates the coverage rate of green land;

[0077] Step S204, calculating a normalized water index based on the average reflectance of the near infrared band and the green band of the remote sensing data;

[0078] The calculation formula of normalized water index NDWI is as follows:

[0079]

[0080] Where GREEN represents the average reflectance of the green band of remote sensing data;

[0081] Step S205, normalize the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil adjusted vegetation index and normalized water index of the near infrared band, blue light band, green light band and red light band by Z-score normalization method to obtain a feature vector.

[0082] In one embodiment of the present invention, the carbon stock prediction model includes M hidden layers, each hidden layer includes N hidden units, the vth hidden unit of the uth hidden layer inputs the element value of the uth row and vth column of the feature matrix, and outputs an update vector, where 1≤u≤M, 1≤v≤N, and the update vector output by the Nth hidden unit of the Mth hidden layer is input to the first classifier, and the classification space of the first classifier represents the carbon stock of the village to be evaluated.

[0083] In one embodiment of the present invention, the calculation formula of the carbon storage prediction model includes:

[0084]

[0085] in represents the update vector of the output of the vth hidden unit in the uth hidden layer, Represents the element value of the u-th row and v-th column of the input feature matrix of the v-th hidden unit of the u-th hidden layer, and They represent the output gating vector, intermediate vector, candidate intermediate vector, input gating vector and forgetting gating vector of the vth hidden unit output of the uth hidden layer, respectively. represents the intermediate vector output by the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the transformation vector, represents the update vector of the output of the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the update vector of the output of the v-1th hidden unit in the uth hidden layer, and The same size, The element values ​​of are all assigned to 0. and denote the first weight parameter, the second weight parameter, the third weight parameter and the fourth weight parameter of the vth hidden unit of the uth hidden layer, respectively. and They respectively represent the first bias parameter, second bias parameter, third bias parameter and fourth bias parameter of the vth hidden unit of the uth hidden layer, tanh represents the hyperbolic tangent activation function, sigmoid represents the sigmoid activation function, MLP represents the multilayer perceptron, concat represents the concatenation function, and ⊙ represents element-by-element multiplication.

[0086] It should be noted that the weight parameters and bias parameters in the carbon storage prediction model are all learnable hyperparameters, and the sizes of the weight parameters and bias parameters in the carbon storage prediction model can be customized. For example, H v u Designed as a 1×16 size vector, then and are all 1×16 vectors, and Also a vector of size 1×16, It can be designed as a vector of size 1×8, then and After concatenation, a vector of size 1×40 is obtained, and the weight parameters in the carbon storage prediction model are all designed as a matrix of size 40×16.

[0087] In one embodiment of the present invention, the vegetation carbon stock of the sub-region is obtained by a biomass measurement method, and the soil carbon stock of the sub-region is obtained by a soil sampling method, and the vegetation carbon stock and soil carbon stock of all sub-regions are added together as sample labels for training samples for training a carbon stock prediction model.

[0088] It should be noted that the biomass measurement method uses biomass data such as vegetation height and diameter at the sampling points to calculate vegetation carbon storage through regional carbon conversion factors. The soil sampling method collects soil samples at different depths, measures the soil organic carbon content, and then infers the soil carbon storage in the sub-region. The distribution of sampling points should cover different terrains, vegetation types, and land use types as much as possible. In order to speed up the training of carbon storage prediction models and carbon flux prediction models, pre-training can be performed on public training data sets, which will not be elaborated here.

[0089] In one embodiment of the present invention, the composition and calculation formula of the carbon flux prediction model are the same as the carbon flux prediction model, which will not be described in detail here.

[0090] In one embodiment of the present invention, the carbon flux of the sub-region is obtained by the eddy covariance method as a sample label of a training sample for training a carbon flux prediction model.

[0091] In one embodiment of the present invention, the calculation formula of the livability score of the village to be evaluated is as follows:

[0092]

[0093] Type i represents the coverage rate of the i-th land use type among housing land, agricultural land, water resources and green land, r i Indicates the weight coefficient corresponding to the coverage rate of the i-th land use type and is a custom parameter. industry represents the coverage rate of industrial land, U represents the weight coefficient corresponding to the coverage rate of industrial land, and U is a custom parameter with a value range of 0.5 to 1. 储量 and They represent the carbon storage of the village to be assessed and the maximum value of historical carbon storage, C 通量 and They represent the maximum carbon flux and historical carbon flux of the village to be evaluated, w1 and w2 represent the first weight coefficient and the second weight coefficient, respectively. Both w1 and w2 are custom parameters, and their sum is 1. σ represents the sigmoid activation function, which is used to control the value range of the livability score between 0 and 1.

[0094] Preferably, the weight coefficients corresponding to the coverage rates of housing land, agricultural land, water resources and green land are set to 0.2, 0.2, 0.2 and 0.4 respectively, U is set to 0.8, w1 is set to 0.5, and w2 is set to 0.5.

[0095] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A method for dynamically collecting and managing rural livability data based on multi-source data, characterized in that: The following steps are involved: Step S101, collecting a bird's-eye view image of the village to be evaluated, and identifying the coverage of different land use types in the bird's-eye view image through an image semantic segmentation model; Land use types include: housing land, agricultural land, industrial land, water resources and green land; Step S102, dividing the village to be evaluated into M sub-areas, collecting remote sensing data of the M sub-areas at preset time intervals t within a preset time period T, and preprocessing the data to obtain a feature matrix; The feature matrix includes M rows and N columns, and the element values ​​of the feature matrix are represented by feature vectors. The feature vector of the mth row and nth column is obtained by preprocessing the remote sensing data of the mth sub-region at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t; The feature vector includes 32 dimensional values, corresponding to the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil-adjusted vegetation index and normalized water index of the near-infrared band, blue band, green band and red band respectively; Step S103, inputting the characteristic matrix into the carbon storage prediction model, and the output value represents the carbon storage of the village to be evaluated; Step S104, inputting the characteristic matrix into the carbon flux prediction model, and the output value represents the carbon flux of the village to be evaluated; Step S105, calculating the livability score of the village to be assessed based on the coverage rate, carbon storage and carbon flux of different land use types in the village to be assessed; The livability score ranges from 0 to 1; Step S106, if it is determined that the livability score of the village to be evaluated is greater than or equal to a preset livability score threshold, it means that the village to be evaluated is livable, otherwise it means that the village to be evaluated is not livable.

2. According to claim 1, a method for dynamically collecting and managing rural livability data based on multi-source data is characterized in that: The number of sub-areas M, the preset time period T, the preset time interval t, and the preset livability score threshold are all custom parameters.

3. According to the method of claim 1, the method is characterized by: The image semantic segmentation model is built based on the Mask2Former model. The sample labels of the training samples of the image semantic segmentation model are obtained through manual annotation, and pre-training is performed on a public remote sensing dataset before training the image semantic segmentation model.

4. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 1 is characterized in that: Preprocessing remote sensing data to obtain feature vectors includes the following steps: Step S201, respectively obtaining the reflectance average value, maximum value, minimum value, standard deviation, and coefficient of variation of the near infrared band, blue light band, green light band, and red light band of the remote sensing data; Where the coefficient of variation is equal to the standard deviation of reflectance divided by the mean reflectance; Step S202, obtaining an enhanced vegetation index by calculating the average reflectance of the near infrared band, the blue light band and the red light band of the remote sensing data; The calculation formula of the Enhanced Vegetation Index (EVI) is as follows: Among them, NIR, BLUE and RED represent the average reflectance of the near-infrared band, blue band and red band of remote sensing data respectively; Step S203, calculating and obtaining the soil adjusted vegetation index according to the average reflectance of the near infrared band and the red light band of the remote sensing data; The calculation formula of soil adjusted vegetation index SAVI is as follows: Where L afforest Indicates the coverage rate of green land; Step S204, calculating a normalized water index based on the average reflectance of the near infrared band and the green band of the remote sensing data; The calculation formula of normalized water index NDWI is as follows: Where GREEN represents the average reflectance of the green band of remote sensing data; Step S205, normalize the reflectance average, maximum, minimum, standard deviation, coefficient of variation, enhanced vegetation index, soil adjusted vegetation index and normalized water index of the near infrared band, blue light band, green light band and red light band by Z-score normalization method to obtain a feature vector.

5. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 1 is characterized in that: The carbon stock prediction model includes M hidden layers, each hidden layer includes N hidden units. The vth hidden unit of the uth hidden layer inputs the element value of the uth row and vth column of the feature matrix and outputs an update vector, where 1≤u≤M, 1≤v≤N. The update vector output by the Nth hidden unit of the Mth hidden layer is input to the first classifier, and the classification space of the first classifier represents the carbon stock of the village to be evaluated.

6. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 5 is characterized in that: The calculation formula of the carbon storage prediction model includes: in represents the update vector of the output of the vth hidden unit in the uth hidden layer, Represents the element value of the u-th row and v-th column of the input feature matrix of the v-th hidden unit of the u-th hidden layer, and They represent the output gating vector, intermediate vector, candidate intermediate vector, input gating vector and forgetting gating vector of the vth hidden unit output of the uth hidden layer, respectively. represents the intermediate vector output by the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the transformation vector, represents the update vector of the output of the vth hidden unit in the u-1th hidden layer, and The same size, The element values ​​of are all assigned to 0. represents the update vector of the output of the v-1th hidden unit in the uth hidden layer, and The same size, The element values ​​of are all assigned to 0. and denote the first weight parameter, the second weight parameter, the third weight parameter and the fourth weight parameter of the vth hidden unit of the uth hidden layer, respectively. and They respectively represent the first bias parameter, second bias parameter, third bias parameter and fourth bias parameter of the vth hidden unit of the uth hidden layer, tanh represents the hyperbolic tangent activation function, sigmoid represents the sigmoid activation function, MLP represents the multilayer perceptron, concat represents the concatenation function, and ⊙ represents element-by-element multiplication.

7. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 1 is characterized in that: The vegetation carbon stock of the sub-region was obtained by the biomass measurement method, and the soil carbon stock of the sub-region was obtained by the soil sampling method. The vegetation carbon stock and soil carbon stock of all sub-regions were added together as the sample labels of the training samples for training the carbon stock prediction model.

8. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 5 is characterized in that: The composition and calculation formula of the carbon flux prediction model are the same as those of the carbon flux prediction model.

9. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 1 is characterized in that: The carbon flux of the sub-region is obtained by the eddy covariance method as the sample label of the training sample for training the carbon flux prediction model.

10. The method for dynamically collecting and managing rural livability data based on multi-source data according to claim 1 is characterized in that: The calculation formula for the livability score of the village to be evaluated is as follows: Type i represents the coverage rate of the i-th land use type among housing land, agricultural land, water resources and green land, r i Indicates the weight coefficient corresponding to the coverage rate of the i-th land use type and is a custom parameter. industry represents the coverage rate of industrial land, U represents the weight coefficient corresponding to the coverage rate of industrial land, and U is a custom parameter with a value range of 0.5 to 1. 储量 and They represent the carbon storage of the village to be assessed and the maximum value of historical carbon storage, C 通量 and They represent the maximum carbon flux and historical carbon flux of the village to be evaluated, w1 and w2 represent the first weight coefficient and the second weight coefficient, respectively. Both w1 and w2 are custom parameters, and their sum is 1. σ represents the sigmoid activation function, which is used to control the value range of the livability score between 0 and 1.

Images