CN114819112A - Land use suitability probability generation method considering spatial partition - Google Patents

Land use suitability probability generation method considering spatial partition Download PDF

Info

Publication number
CN114819112A
CN114819112A CN202210756949.9A CN202210756949A CN114819112A CN 114819112 A CN114819112 A CN 114819112A CN 202210756949 A CN202210756949 A CN 202210756949A CN 114819112 A CN114819112 A CN 114819112A
Authority
CN
China
Prior art keywords
land
layer
driving
land use
change
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
Application number
CN202210756949.9A
Other languages
Chinese (zh)
Other versions
CN114819112B (en
Inventor
杜志强
李柏延
王超
李沐春
王伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Beidou Innovation And Application Technology Research Institute Co ltd
Original Assignee
Nanjing Beidou Innovation And Application Technology Research Institute Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Beidou Innovation And Application Technology Research Institute Co ltd filed Critical Nanjing Beidou Innovation And Application Technology Research Institute Co ltd
Priority to CN202210756949.9A priority Critical patent/CN114819112B/en
Publication of CN114819112A publication Critical patent/CN114819112A/en
Application granted granted Critical
Publication of CN114819112B publication Critical patent/CN114819112B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • 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
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • 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
    • G06Q50/02—Agriculture; Fishing; Forestry; Mining
    • 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
    • G06Q50/16—Real estate
    • G06Q50/165—Land development
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/771—Feature selection, e.g. selecting representative features from a multi-dimensional feature space
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Tourism & Hospitality (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Mathematical Physics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Computational Linguistics (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Biomedical Technology (AREA)
  • Primary Health Care (AREA)
  • Probability & Statistics with Applications (AREA)
  • Development Economics (AREA)
  • Marine Sciences & Fisheries (AREA)
  • Mining & Mineral Resources (AREA)
  • Algebra (AREA)
  • Mathematical Analysis (AREA)
  • Agronomy & Crop Science (AREA)
  • Animal Husbandry (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computational Mathematics (AREA)

Abstract

The invention discloses a land use suitability probability generation method considering spatial zoning, which considers the spatial difference of a land use change driving mechanism, quantitatively describes the contribution degree of each driving factor in land use change modeling and is beneficial to understanding the influence of each driving factor on land use change; the problem of reduced simulation precision caused by insufficient consideration of time-space heterogeneity in land use change simulation is solved by constructing a space partition, a driving factor data set, calculating contribution degrees of all driving factors and training suitability probability; based on the convolutional neural network, the complex nonlinear association of the natural geography and social economy driving factors and the land utilization change is constructed, so that the suitability probability of each land utilization is obtained, the understanding of the space-time heterogeneity of the land utilization change of a research area is deepened, the perception capability of the multi-land utilization driving factors in the neighborhood space is improved, and the preparation is also made for long-time-sequence simulation and prediction of the land utilization change space-time.

Description

Land use suitability probability generation method considering spatial partition
Technical Field
The invention relates to the technical field of land use change simulation, in particular to a land use suitability probability generation method considering spatial zoning.
Background
The study on Land Use Change suitability probability aims to disclose one pole of intrinsic dynamic mechanism and action mechanism of Land Use/coverage Change (LUCC) and further predict the dynamic evolution trend of Land Use, so that the study is widely concerned by domestic and foreign scholars, the selection probability of each Land Use type is proportional to the suitability probability, the larger the suitability probability is, the stronger the Land Use competitiveness is, and the larger the probability of Land Use to be selected as the conversion state of a cellular at the next moment is.
The driving factors of land utilization change are mainly divided into two categories, namely natural environment factors and social and economic factors, and the natural environment factors represented by gradient, terrain, climate, rainfall and the like are an indispensable part of long-time land utilization change research; socio-economic factors affecting LUCC include population growth, economic development, traffic accessibility, and land use management policies.
The existing research shows that along with the continuous improvement of the urbanization level, social and economic factors represented by traffic accessibility exert influence far exceeding the natural environment factors on the land utilization change; internal correlation also exists among driving factors with different time and space scales, so that space-time heterogeneity of land utilization change is caused, and complexity and uncertainty are added to an analysis process and a construction process of the LUCC driving mechanism; on a long time scale, the influence of natural environment factors on land use change is obvious; on the medium and small scale and the short time scale, as the influence of human activities on land use change is increasingly prominent, social and economic factors gradually take a leading position.
In recent years, a plurality of scholars begin to pay attention to the space-time heterogeneity phenomenon of land utilization change, try to divide the whole research area in a space partition mode, further explore land utilization change modes of different sub-areas, and with the breakthrough of geographic metering concepts and the introduction of intelligent algorithms, the research of a land utilization change driving mechanism breaks through a traditional qualitative method with dominant practical experience, and gradually progresses to the quantification of high efficiency of a simulation process and high precision of a simulation result.
The introduction of the intelligent algorithm provides a more efficient means for the research of the land utilization change driving mechanism, and in recent years, the research of the driving mechanism based on the intelligent algorithm is continuously increased, so that the space-time dependence characteristics of the driving factors of the research area and the space-time evolution difference of the land utilization are considered, a reasonable driving factor data set for the research of the land utilization change is constructed, the association between the land utilization change and the driving factors of the natural geography and social economic dimensionality is mined, and the importance and the key of the research of the land utilization change driving mechanism are reached.
Disclosure of Invention
In order to solve the above technical problems, the present invention provides a land use suitability probability generation method considering spatial zoning, comprising the steps of:
s1, land utilization data preprocessing: reclassifying the land utilization data, and setting the encoding of cultivated land as 1; the code of the woodland is set to 2; the coding of shrubs and grass is set to 3; the code of the water body is set to be 4; the code of the construction land is set to 5, and the construction land comprises a watertight surface; the code of the unused land is set as 6, and the unused land comprises snow land, ice surface, wet land and bare land;
s2, constructing land utilization data with the resolution of 100m multiplied by 100m of a plurality of land utilization area information;
s3, preprocessing driving factor data: selecting 12 driving factors to construct a driving factor data set, wherein the driving factor data set comprises natural geography driving factors and social economy driving factors, and the number of the natural geography driving factors is set to be 3, namely elevation, gradient and distance water body; the number of the socio-economic driving factors is set to 9, and the driving factors are respectively the distance from the government, population, GDP, expressway, main road, primary road, secondary road, tertiary road and railway; constructing a traffic accessibility data set of driving factors of the land utilization change socioeconomic driving factors according to the expressway, the main road, the primary road, the secondary road, the tertiary road and the railway; selecting Euclidean distance data from the interest points of the government construction area; the driver data set is resampled and the results are normalized, in the mathematical form:
Figure 100002_DEST_PATH_IMAGE002
in the formula (I), the compound is shown in the specification,
Figure 100002_DEST_PATH_IMAGE004
represents the traffic reachability after normalization, d represents the euclidean distance of the current cell from the above-mentioned driving factors,
Figure 100002_DEST_PATH_IMAGE006
and
Figure 100002_DEST_PATH_IMAGE008
respectively representing the minimum value and the maximum value of the Euclidean distance between the current cellular and the driving factor;
s4, space division based on the land utilization change visual angle and the landscape pattern change visual angle: generating a land use change data set according to key time nodes of a research area, and carrying out space division on the research area from two visual angles of land use change and landscape pattern change;
s5, selecting a driving factor;
s6, constructing a land utilization driving mechanism based on the CNN network, and training the land suitability probability by matching with a mixed cellular automaton;
s7, calculating the suitability probability, wherein the mathematical form is as follows:
Figure 100002_DEST_PATH_IMAGE010
in the formula (I), the compound is shown in the specification,
Figure 100002_DEST_PATH_IMAGE012
indicating the state of the cell, i.e. the ground type,
Figure 100002_DEST_PATH_IMAGE014
representing a cellular transformation to
Figure 100002_DEST_PATH_IMAGE015
The probability of (a) of (b) being,
Figure 100002_DEST_PATH_IMAGE017
indicating the state of the cell
Figure 811285DEST_PATH_IMAGE015
The probability of suitability of the target,
Figure 100002_DEST_PATH_IMAGE019
representing the cumulative probability;
and S8, calculating convolution network training results based on the confusion matrix of each land according to the space partition, and outputting a multi-band suitability probability training result graph with consistent land use types.
The invention further defines the technical scheme that:
further, in step S2, the method for constructing the land use data of the 100m × 100m resolution of the plurality of land use area information includes the steps of:
s2.1, selecting a resolution of 100m multiplied by 100m as a basis for land use change simulation;
s2.2, constructing a fishing net with the resolution of 100m multiplied by 100m based on land utilization data with the original resolution of 30m multiplied by 30 m;
s2.3, calculating the proportion of each land area in the fishing net space with the resolution of 100m multiplied by 100m to the area of the fishing net;
and S2.4, obtaining land use data with the resolution of 100m multiplied by 100m and comprising a plurality of land use area information.
In the land use suitability probability generation method considering the spatial partition, in step S4, the method for spatially partitioning the research area from two perspectives of land use change and landscape pattern change includes the following steps:
s4.1, carrying out space division on the research area from a soil utilization change view angle: partitioning according to the land utilization degree and the change rate of the land utilization degree, and dividing into a first echelon, a second echelon and a third echelon;
s4.2, carrying out space division on the research area from the view angle of landscape pattern change: and selecting a maximum plaque index (LPI) and an Edge Density (ED) as evaluation indexes, and dividing the space into a main urban area and a far urban area.
In the land use suitability probability generation method considering the spatial partition, in step S5, the selection factor of the driving factor includes the following steps:
s5.1, judging whether the driving factors have easy acquireability, comprehensiveness, space-time information consistency, space difference and quantifiability at the same time, and if so, selecting the driving factors;
s5.2, constructing a driving factor data set;
and S5.3, evaluating the importance of the driving factors on land use change.
In the land use suitability probability generation method considering the spatial division, in step S6, the method for training the land use suitability probability includes the following steps:
s6.1, a convolutional layer: the method comprises the following steps that input data of a convolution network model firstly pass through convolution layers, feature representation of input variables is learned through convolution kernels, the convolution network model gradually analyzes abstract features of different dimensions of an image through a plurality of convolution layers, deeper feature representation is obtained, and the mathematical form of the convolution layers is as follows:
Figure 100002_DEST_PATH_IMAGE021
in the formula (I), the compound is shown in the specification,
Figure 100002_DEST_PATH_IMAGE023
an ith dimension input variable representing the first convolutional layer,
Figure 100002_DEST_PATH_IMAGE025
and
Figure 100002_DEST_PATH_IMAGE027
respectively representing the weight vector and the bias term of the jth convolution kernel group in the ith convolution layer,
Figure 100002_DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the first convolution layer, n representing the total dimension of input variables of the first convolution layer, and f representing a nonlinear activation function;
s6.2, a pooling layer: selecting the data characteristics after convolution extraction, screening out the characteristics meeting the requirements according to the difference of information quantity, removing unimportant characteristics, simultaneously reducing data dimensionality and avoiding overfitting, calculating a function value for the data characteristics in a pooling region, taking the arithmetic mean value of all elements in the region as output through mean pooling, and extracting the most significant characteristic value of a local characteristic plane as output through maximum pooling for extracting shallow characteristics, wherein the mathematical form of a pooling layer is as follows:
Figure 100002_DEST_PATH_IMAGE031
in the formula (I), the compound is shown in the specification,
Figure 100002_DEST_PATH_IMAGE033
an ith dimension input variable representing the ith pooling layer,
Figure 100002_DEST_PATH_IMAGE035
and
Figure 100002_DEST_PATH_IMAGE037
respectively representing a multiplication offset term and an addition offset term of a j dimension variable in the l pooling layer,
Figure 452220DEST_PATH_IMAGE029
an output variable representing a j dimension variable in the l pooling layer, and down representing a pooling function;
s6.3, a full connection layer: the convolution network inputs the abstract expression of the original input data into a full connection layer for feature fusion through feature extraction and filtering layer by layer, and forms a one-dimensional feature vector corresponding to a target space, wherein the mathematical form of the full connection layer is as follows:
Figure 100002_DEST_PATH_IMAGE039
in the formula (I), the compound is shown in the specification,
Figure 57644DEST_PATH_IMAGE033
the ith dimension neuron representing the l < th > fully connected layer,
Figure 100002_DEST_PATH_IMAGE041
represents weights of neuron i to neuron j in the first convolutional layer,
Figure 100002_DEST_PATH_IMAGE043
represents the bias term from neuron i to neuron j in the convolution layer of the l layer,
Figure 257682DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the ith convolution layer, wherein n represents the total number of input neurons of the fully-connected layer, and f represents a nonlinear activation function;
s6.4, an active layer: taking the land coverage ratio of the mixed cells as a convolutional network training target, setting an activation function adopted by the full connection layer as a Sigmoid activation function, wherein the mathematical form of the activation function is as follows:
Figure 100002_DEST_PATH_IMAGE045
in the formula (I), the compound is shown in the specification,
Figure 657570DEST_PATH_IMAGE029
representing the input of the active layer.
In the land use suitability probability generation method considering the spatial partition, in step S3, the driver data set is resampled by the software arcgis 10.1.
In the land use suitability probability generation method considering the spatial division as described above, the driver data set is set to the original 30m30m resolution driver data set in step S3.
In the land use suitability probability generation method considering the space partition, in step S2.3, a calculation tool of the proportion of each land area to the area of the fishing net is set as a partition statistical tool of the arcgis10.1 software.
The invention has the beneficial effects that:
(1) according to the method, the spatial difference of the land use change driving mechanism can be considered, the contribution of each driving factor in land use change modeling is quantitatively described, and the method is helpful for understanding how each driving factor influences the land use change;
(2) the method disclosed by the invention surrounds space partition, construction of a driving factor data set, calculation of contribution degrees of all driving factors and training of suitability probability, and solves the problem of reduced simulation precision caused by insufficient consideration of time-space heterogeneity in land use change simulation;
(3) the method is based on a Convolutional Neural Network (CNN), and constructs the complex nonlinear association of natural geography and social economy driving factors and land utilization change, so that the suitability probability of each land utilization is obtained, the understanding of the space-time heterogeneity of the land utilization change in a research area is deepened, the sensing capability of the natural geography and social economy driving factors in a neighborhood space is improved, and preparation is made for long-time-sequence land utilization change space-time simulation and prediction.
Drawings
FIG. 1 is a time-space variation diagram of land utilization in a test area according to the present invention;
FIG. 2 is a diagram of the temporal and spatial variation of landscape indexes in a test area according to the present invention;
FIG. 3 is the contribution (SHAP value) of each driving factor to the cultivated land and the forest land in the test area according to the present invention;
FIG. 4 is a test area convolution network structure of the present invention;
FIG. 5 is a test area main urban area convolution network suitability training confusion matrix in the invention;
FIG. 6 is a confusion matrix for testing the suitability of a far-urban convolutional network in a test area;
FIG. 7 is a diagram illustrating suitability probability training for various fields of experiments in the present invention.
Detailed Description
In order to facilitate understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be further described in detail by taking wuhan city with large land use variation range and large scale in the last decade as an example.
A land use suitability probability generation method considering spatial zoning comprises the following steps:
s1, land utilization data preprocessing: reclassifying the land utilization data according to relevant specifications of 'contents of general survey of geographical national conditions and indexes GDPJ 01-2013', and setting the codes of cultivated land as 1; the code of the woodland is set to 2; the coding of shrubs and grass is set to 3; the code of the water body is set to be 4; the code of the construction land is set to 5, and the construction land comprises a watertight surface; the code of the unused land is set to 6, and the unused land includes snow, ice, wet land and bare land.
S2, constructing land use data with the resolution of 100m multiplied by 100m of a plurality of land use area information, and comprising the following steps:
s2.1, comprehensively considering research targets and model operation efficiency, and selecting a resolution of 100m multiplied by 100m as a basis for land use change simulation;
s2.2, constructing a fishing net with the resolution of 100m multiplied by 100m based on the land utilization data with the original resolution of 30m multiplied by 30 m;
s2.3, calculating the proportion of each land area in a fishing net space with the resolution of 100m multiplied by 100m to the area of the fishing net through a partition statistical tool of ArcGIS10.1 software, thereby aggregating the spatial distribution characteristics of high-resolution land utilization data into a low-resolution grid and effectively preventing all high-resolution data information from being lost;
and S2.4, obtaining land use data with the resolution of 100m multiplied by 100m and comprising a plurality of land use area information.
S3, preprocessing driving factor data: selecting 12 driving factors to construct a driving factor data set, wherein the driving factor data set comprises natural geography driving factors and social economy driving factors, and the number of the natural geography driving factors is set to be 3, namely elevation, gradient and distance water body; the number of the socio-economic driving factors is set to 9, and the driving factors are respectively the distance from the government, population, GDP, expressway, main road, primary road, secondary road, tertiary road and railway; constructing a traffic accessibility data set of driving factors of the land utilization change socioeconomic driving factors according to the expressway, the main road, the primary road, the secondary road, the tertiary road and the railway; selecting Euclidean distance data from the interest points of the government construction area; the original 30m30m resolution driver data set was resampled by the software ArcGISI 10.1 and the results were normalized as follows:
Figure 943058DEST_PATH_IMAGE002
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE046
representing normalized traffic reachabilityD represents the Euclidean distance of the current cell from the above-mentioned driving factors,
Figure 680070DEST_PATH_IMAGE006
and
Figure DEST_PATH_IMAGE008A
respectively representing the minimum and maximum Euclidean distances of the current cell from the above-mentioned driving factors.
S4, according to the characteristics of space-time evolution of land utilization in a research area, selecting representative 4 years (2005, 2010, 2015 and 2020) as key time nodes for research on land utilization change in Wuhan city, generating research data sets of 2005-2010, 2010-2015 and 2015-2020 at intervals of 6 years, analyzing the characteristics of space-time evolution of land utilization in Wuhan city from two perspectives of land utilization change and landscape pattern change, and further performing space division on the research area, wherein the landscape pattern is a drawing and description of the structural characteristics of landscape space in the research area, is the combination and distribution of landscape elements with different shapes and sizes, and is also the result of the combined action of various elements such as nature, society and humanity, and the like, and the method for space division comprises the following steps:
s4.1, carrying out space division on the research area from a soil utilization change visual angle: partitioning is performed according to the land utilization degree and the change rate of the land utilization degree, and as shown in fig. 1, the partitioning is divided into a first echelon, a second echelon and a third echelon:
the first echelon comprises Jianghhan, mouths, Qingshan and Wuchang, the land utilization degree is higher and is more than 300, meanwhile, in 2005 + 2020, the land utilization degree changes in three areas except mouths are not big, and the average land utilization degree change rates of Jianghhan, Qingshan and Wuchang are only 1.35%, 1.37% and 0.66% respectively;
the second echelon comprises a river bank, Hanyang and a Hongshan, although the land utilization degree is lower than that of the first echelon in the initial period (2005), the second echelon shows a rapid growth trend in the research period (2005 + 2020), and the average land utilization degree change rates are respectively 5.09%, 9.01% and 8.15%, wherein the river bank land utilization degree breaks through 300 in 2020 and spans into the range of higher land utilization degree;
the third echelon comprises east-west lakes, Chuaidian, Jiangxa and Hannan, the change trends of the land utilization degrees of four regions are similar, the regions are mainly peripheral administrative regions of Wuhan cities, compared with the township development of main urban regions, the land utilization degrees of the regions in the initial period (2005) of research are all lower than 140, and the change ranges in the research period (2005) 2020 are not large, the average land utilization degree change rates of Chuaidian, Jiangxa and Hannan in 2005 2020 are 5.45%, 4.80% and 4.72% respectively, although the east-west lakes show relatively high average land utilization degree increase rate of 7.49%, the regions are still divided into the third echelon due to the lower land utilization degree in the initial period (2005) of research;
s4.2, carrying out space division on the research area from the view angle of landscape pattern change: the method comprises the steps of selecting a maximum plaque index (LPI) and an Edge Density (ED) as evaluation indexes, and calculating to obtain landscape pattern change degrees of Wuhan cities in 2005, 2010, 2015 and 2020, wherein as shown in FIG. 2, edge density changes of cultivated land, forest land, water body and construction land of the Wuhan cities are most obvious, and a main urban area and a far urban area show obvious space-time difference characteristics.
S5, selecting a driving factor, comprising the following steps:
s5.1, judging whether the driving factors have easy acquireability, comprehensiveness, space-time information consistency, space difference and quantifiability at the same time, and if so, selecting the driving factors;
easy availability: after comprehensively considering driving force factors which may affect land use change, firstly, screening the land use driving force by using the easy acquirability of the driving factors, wherein the driving factors have reliable acquisition ways and can be applied to research, and the accuracy and the usability of the acquired driving factors are checked;
the comprehensiveness: the driving factors influencing the land use change are numerous, but the influence degrees of the driving factors on the land use change are different, the driving factors influence the land use change from different aspects, and typical influence factors in the aspects of society, nature, economy and the like are comprehensively selected in the driving factor selection process so as to comprehensively reflect the land use change of a research area;
consistency of space-time information: the research of each driving factor has scientificity only if the driving factors are in the same space-time scale, if the driving factors are in different time-space scales, the necessity of simulation research is lost, the driving factors are required to be kept to have the same time information and space coordinate information, and the space-time information consistency of the driving factors is determined by the model characteristics and the research scientificity; the selected driving factors should keep consistent with the spatial scale information, otherwise, analysis is difficult to continue, so that research is feasible only if the driving factors are in the same time scale;
spatial diversity: the final land utilization type of the grid is determined by the land type distribution probability determined by the driving factors on the grid, if the driving factors do not have spatial difference and the values of each grid are the same, the result of each grid is consistent, so that the research result has no significance, and therefore the driving factors have to have difference in spatial distribution;
the method can quantify: in the simulation process, the convolution network is used for measuring and calculating the relation between the driving factors and the land type change, the influence of the driving factors on the land use change needs to be input into the convolution network, and the simulation result can be obtained only by quantitative analysis, so that the driving factors need to be quantifiable; in the process of selecting the driving factors, although factors such as laws have great influence on land use change, quantitative representation is difficult, so the factors are not generally included in a driving force system.
S5.2, constructing a driving factor data set: the land utilization change is influenced by various driving factors of natural geography and social economy, and researches show that the gradient, the elevation and the like are considered as important natural geography driving factors influencing the land utilization change, and the distance from rivers and lakes can influence the spatial distribution of cultivated land and construction land; from the socio-economic point of view, population density, total domestic production value and distance to government play a crucial role in the progress of urbanization, as human activities are the determining factors affecting the change in land use.
In addition, the traffic element is an important dimension for evaluating the strength and the breadth of the space interaction, and the accessibility is used as a main product of a traffic system and directly determines the advantages of regions in each region, so that the traffic element is not compatible with the economic development level, social welfare and ecological environment of the region; in order to quantify the influence of traffic elements on land use changes, the existing research generally adopts a traffic accessibility mode to calculate, such as the distance from a highway, the distance from a city center and the like.
S5.3, evaluating the contribution degree of the driving factors to land use change: the contribution degree of each driving factor in land utilization change modeling is quantitatively described, so that urban planners can further know how the factors influence the evolution of land utilization, and the method is important for making a land utilization planning policy; and (4) taking the division maps of the main urban area and the far urban area of the Wuhan city obtained in the step (S4.2) as an example of the space division, and respectively and quantitatively describing the contribution degrees of the driving factors of all the land in the main urban area and the far urban area based on a convolution network.
The SHAP is an effective method for carrying out heuristic interpretation on a machine learning network architecture by calculating the contribution degree of each feature to a model, and the method estimates the contribution degree of a driving factor to the land utilization change by calculating the total SHAP value of the driving factor of each land utilization type of a research subarea (a main urban area and a far urban area). The process of quantitatively describing the contribution of the driving factors for each land in the first, second and third echelons in step S4.2 is similar to the process of calculating the contribution of the driving factors for each land in the spatial partition in step S5.3, and is not described again here.
As shown in fig. 3, it can be seen that the land utilization change driving mechanism of the main urban area and the far urban area in wuhan city has difference, the spatial distribution of the intertillage land in the main urban area is greatly influenced by population and GDP, and the SHAP values are 10.27% and 9.85% respectively; the influence of natural geography driving factors is second, the influence of the distance, the gradient and the elevation from the water body to cultivated land is relatively close, and the SHAP values are 6.75%, 6.61% and 6.56% respectively; compared with the main urban area, the natural geographic driving factors in the far urban area have larger contribution degree to the spatial distribution condition of cultivated land, the SHAP values of the elevation and the gradient in the far urban area are respectively improved by 3.98 percent and 2.70 percent, and the contribution degree of the water body is not greatly changed.
In addition, SHAP values of the forest lands basically show similar variation trends in main urban areas and far urban areas, and natural geographic driving factors represented by elevations have more influence on spatial distribution of the forest lands and far exceed traffic accessibility, population and GDP; taking a forest land as an example, the height SHAP values of the main urban area and the far urban area are respectively 2.79% and 3.33%, and the SHAP values of the socioeconomic driving factors, such as population, GDP, distance from a main trunk road and the like, are less than 0.5%; in general, the evolution of agriculture and woodland is mainly influenced by the natural environment, so that natural geographic driving factors contribute most to the change of the two land utilization.
S6, constructing a land utilization driving mechanism based on the CNN network, and training the land suitability probability by matching with a mixed cellular automaton, wherein the training method comprises the following steps:
s6.1, a convolutional layer: the method comprises the steps that input data of a convolution network model firstly pass through a convolution layer, and feature representation of input variables is learned through convolution kernels; the convolution layer consists of a plurality of convolution kernels used for extracting different characteristics of input data and is the core of a convolution neural network, the characteristic mapping output of the convolution network needs convolution operation through the convolution kernels, and the characteristic extraction output is obtained through a nonlinear activation function; the convolution network model gradually analyzes the abstract characteristics of different dimensionalities of the image through a plurality of convolution layers to obtain deeper characteristic representation; in addition, the convolution layer reduces the number of parameters of operation through a weight sharing mechanism, effectively reduces the time complexity and the space complexity of network calculation, and the mathematical form of the convolution layer is as follows:
Figure 108515DEST_PATH_IMAGE021
in the formula (I), the compound is shown in the specification,
Figure 538359DEST_PATH_IMAGE023
an ith dimension input variable representing the first convolutional layer,
Figure 311143DEST_PATH_IMAGE025
and
Figure 586267DEST_PATH_IMAGE027
respectively representing the weight vector and the bias term of the jth convolution kernel group in the ith convolution layer,
Figure 229737DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the first convolution layer, n representing the total dimension of input variables of the first convolution layer, and f representing a nonlinear activation function;
s6.2, a pooling layer: after the convolutional layer operation is completed, although the parameter magnitude is reduced, the data volume is still large, so that the data features extracted by convolution are continuously selected, the features meeting the requirements are screened out according to different information volumes, unimportant features are removed, meanwhile, the data dimensionality is reduced, overfitting is avoided, a function value is calculated for the data features in a pooling region, the function value is different due to different parameters, so that the arithmetic mean value of all elements in the region is taken as output through mean pooling, the most significant feature value of a local feature plane is extracted as output through maximum pooling, and the most significant feature value is used for extracting shallow features, and the mathematical form of the pooling layer is as follows:
Figure DEST_PATH_IMAGE031A
in the formula (I), the compound is shown in the specification,
Figure 502587DEST_PATH_IMAGE033
an ith dimension input variable representing the ith pooling layer,
Figure 231509DEST_PATH_IMAGE035
and
Figure 310323DEST_PATH_IMAGE037
respectively representing a multiplication offset term and an addition offset term of a j dimension variable in the l pooling layer,
Figure 214825DEST_PATH_IMAGE029
an output variable representing a j dimension variable in the l pooling layer, and down representing a pooling function;
s6.3, a full connection layer: the full connection layer is usually positioned at the tail end of the convolution network and connects all output variables of the previous layer of network; the convolution network inputs the abstract expression of the original input data into a full connection layer for feature fusion through feature extraction and filtering layer by layer, and forms a one-dimensional feature vector corresponding to a target space, wherein the mathematical form of the full connection layer is as follows:
Figure DEST_PATH_IMAGE039A
in the formula (I), the compound is shown in the specification,
Figure 517630DEST_PATH_IMAGE033
the ith dimension neuron representing the l < th > fully connected layer,
Figure 468269DEST_PATH_IMAGE041
represents weights of neuron i to neuron j in the first convolutional layer,
Figure 350774DEST_PATH_IMAGE043
represents the bias term from neuron i to neuron j in the convolution layer of the l layer,
Figure 109783DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the ith convolution layer, wherein n represents the total number of input neurons of the fully-connected layer, and f represents a nonlinear activation function;
s6.4, an active layer: the convolutional network is generally added with a nonlinear activation function after a convolutional layer and a full connection layer so as to simulate the inhibitory action of human neurons, thereby improving the image recognition capability of the network; common activation functions comprise a Sigmoid activation function, a Tanh activation function and a Relu activation function, wherein the Sigmoid is used as the activation function with the widest application range, and the output is mapped between (0, 1); the invention gives consideration to mixed cells, in the process of training the suitability of the convolutional network, the land coverage ratio of the mixed cells is taken as a training target of the convolutional network, an activation function adopted by a full connection layer is set as a Sigmoid activation function, and the mathematical form is as follows:
Figure 52331DEST_PATH_IMAGE045
in the formula (I), the compound is shown in the specification,
Figure 21424DEST_PATH_IMAGE029
representing the input of the active layer, the input data can be compressed to the range of 0-1 through a Sigmoid function.
S7, calculating the suitability probability: the selection probability of each land is proportional to the suitability probability, the larger the suitability probability is, the stronger the land competitiveness is, and the larger the probability of the conversion state of the land selected as the next moment of the cell is; its mathematical form is as follows:
Figure 442041DEST_PATH_IMAGE010
in the formula (I), the compound is shown in the specification,
Figure 914611DEST_PATH_IMAGE012
indicating the state of the cell, i.e. the ground type,
Figure 933120DEST_PATH_IMAGE014
representing a cellular transformation to
Figure 858351DEST_PATH_IMAGE015
The probability of (a) of (b) being,
Figure 82659DEST_PATH_IMAGE017
indicating the state of the cell
Figure 675314DEST_PATH_IMAGE015
The probability of suitability of the target,
Figure 959665DEST_PATH_IMAGE019
representing the cumulative probability.
In order to obtain the land use change suitability probability of the cellular automata model conversion rule, the invention refers to a LetNet-5 model to construct a convolutional neural network model with a seven-layer network structure, the model takes a land use change driving factor as input, and takes each land coverage ratio as output, and the specific network structure is shown in figure 4:
selecting a central unit of N driving factors and a three-dimensional tensor of a spatial neighborhood from the upper left corner of a research area by taking each pixel in a land utilization classification map of the research area as a central unit; in combination with a LetNet-5 model, 33 cells are selected as the height and the width of input data of a convolutional network, namely a three-dimensional tensor of the cells with the length of Nx 35 x 35, wherein N represents the number of driving factors, and N = 12.
The model comprises three convolution layers, a pooling layer and two full-connection layers, wherein the first layer consists of 12 multiplied by 5 convolution kernels and outputs a 12 multiplied by 29 characteristic space; the second layer is a 2 × 2 pooling layer and outputs a 12 × 14 × 14 feature space; the third layer is a convolution kernel of 16 × 5 × 5; the fourth layer is a 2 multiplied by 2 pooling layer, and a 16 multiplied by 5 feature space is obtained; the fifth layer is a convolution kernel of 120 × 5 × 5; the sixth layer is a full connection layer and comprises 120 neurons; the seventh layer is a connecting layer containing 84 neurons; and finally, calculating conversion suitability probability of various places by adopting a Sigmoid activation function.
S8, calculating convolution network training results based on the confusion matrix for each land according to space partition, as shown in FIGS. 5 to 6, respectively displaying the convolution network training results based on the confusion matrix for each land in the main urban area and the far urban area, wherein the simulation result of the mixed land has high consistency with the real mixed land and is generally more than 0.8; the accuracy of the convolutional network training is reduced due to fewer samples of grassland and unused land; in addition, the spatial distribution of the suitability of each land use type is obtained based on the convolutional network, and data support is provided for the land use dynamic evolution simulation of the coupling model;
outputting a multiband suitability probability training result graph with consistent land use types, wherein the high suitability probability means that the land use type is easy to expand in the neighborhood to occupy larger land coverage area, so that other lands are easy to convert to the land use state in the land use evolution process; as shown in fig. 7, in order to test the probability training diagram of suitability of each land, it can be found that the suitability of the spatial distribution of cultivated land is mainly affected by water and construction land, but in general, socioeconomic driving factors mainly including population and GDP and natural geographic driving factors mainly including elevation and gradient play a leading role in spatial distribution of each land utilization type.
In addition to the above embodiments, the present invention may have other embodiments. All technical solutions formed by adopting equivalent substitutions or equivalent transformations fall within the protection scope of the present invention.

Claims (8)

1. A land use suitability probability generation method considering spatial zoning is characterized by comprising the following steps: the method comprises the following steps:
s1, land utilization data preprocessing: reclassifying the land utilization data, and setting the encoding of cultivated land as 1; the code of the woodland is set to 2; the coding of shrubs and grass is set to 3; the code of the water body is set to be 4; the code of the construction land is set to 5, and the construction land comprises a watertight surface; the code of the unused land is set as 6, and the unused land comprises snow land, ice surface, wet land and bare land;
s2, constructing land utilization data with the resolution of 100m multiplied by 100m of a plurality of land utilization area information;
s3, preprocessing driving factor data: selecting 12 driving factors to construct a driving factor data set, wherein the driving factor data set comprises natural geography driving factors and social economy driving factors, and the number of the natural geography driving factors is set to be 3, namely elevation, gradient and distance water body; the number of the socio-economic driving factors is set to 9, and the driving factors are respectively the distance from the government, population, GDP, expressway, main road, primary road, secondary road, tertiary road and railway; constructing a traffic accessibility data set of driving factors of the land utilization change socioeconomic driving factors according to the expressway, the main road, the primary road, the secondary road, the tertiary road and the railway; selecting Euclidean distance data from the interest points of the government construction area; the driver data set is resampled and the results are normalized, in the mathematical form:
Figure DEST_PATH_IMAGE002
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE004
represents the traffic reachability after normalization, d represents the euclidean distance of the current cell from the above-mentioned driving factors,
Figure DEST_PATH_IMAGE006
and
Figure DEST_PATH_IMAGE008
respectively representing the minimum value and the maximum value of the Euclidean distance between the current cellular and the driving factor;
s4, space division based on the land utilization change visual angle and the landscape pattern change visual angle: generating a land use change data set according to key time nodes of a research area, and carrying out space division on the research area from two visual angles of land use change and landscape pattern change;
s5, selecting a driving factor;
s6, constructing a land utilization driving mechanism based on the CNN network, and training the land suitability probability by matching with a mixed cellular automaton;
s7, calculating the suitability probability, wherein the mathematical form is as follows:
Figure DEST_PATH_IMAGE010
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE012
indicating the state of the cell, i.e. the ground type,
Figure DEST_PATH_IMAGE014
representing a cellular transformation to
Figure DEST_PATH_IMAGE015
The probability of (a) of (b) being,
Figure DEST_PATH_IMAGE017
indicating the state of the cell
Figure 502377DEST_PATH_IMAGE015
The probability of suitability of the target,
Figure DEST_PATH_IMAGE019
representing the cumulative probability;
and S8, calculating convolution network training results based on the confusion matrix of each land according to the space partition, and outputting a multi-band suitability probability training result graph with consistent land use types.
2. The land use suitability probability generation method considering spatial zoning according to claim 1, characterized in that: in step S2, the method for constructing land use data of 100m × 100m resolution of a plurality of land use area information includes the steps of:
s2.1, selecting a resolution of 100m multiplied by 100m as a basis for land use change simulation;
s2.2, constructing a fishing net with the resolution of 100m multiplied by 100m based on the land utilization data with the original resolution of 30m multiplied by 30 m;
s2.3, calculating the proportion of each land area in the fishing net space with the resolution of 100m multiplied by 100m to the area of the fishing net;
and S2.4, obtaining land use data with the resolution of 100m multiplied by 100m and comprising a plurality of land use area information.
3. The land use suitability probability generation method considering spatial zoning according to claim 1, characterized in that: in step S4, the method for spatially dividing the research area from two perspectives of land use change and landscape pattern change includes the following steps:
s4.1, carrying out space division on the research area from a soil utilization change visual angle: partitioning according to the land utilization degree and the change rate of the land utilization degree, and dividing into a first echelon, a second echelon and a third echelon;
s4.2, carrying out space division on the research area from the view angle of landscape pattern change: and selecting a maximum plaque index (LPI) and an Edge Density (ED) as evaluation indexes, and dividing the space into a main urban area and a far urban area.
4. The land use suitability probability generation method considering spatial zoning according to claim 1, characterized in that: in step S5, the selection of the driving factor includes the following steps:
s5.1, judging whether the driving factors have easy acquireability, comprehensiveness, space-time information consistency, space difference and quantifiability at the same time, and if so, selecting the driving factors;
s5.2, constructing a driving factor data set;
and S5.3, evaluating the importance of the driving factors on land use change.
5. The land use suitability probability generation method considering spatial zoning according to claim 1, characterized in that: in step S6, the method for training the land suitability probability includes the following steps:
s6.1, a convolutional layer: the method comprises the following steps that input data of a convolution network model firstly pass through convolution layers, feature representation of input variables is learned through convolution kernels, the convolution network model gradually analyzes abstract features of different dimensions of an image through a plurality of convolution layers, deeper feature representation is obtained, and the mathematical form of the convolution layers is as follows:
Figure DEST_PATH_IMAGE021
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE023
an ith dimension input variable representing the first convolutional layer,
Figure DEST_PATH_IMAGE025
and
Figure DEST_PATH_IMAGE027
respectively representing the weight vector and the bias term of the jth convolution kernel group in the ith convolution layer,
Figure DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the first convolution layer, n representing the total dimension of input variables of the first convolution layer, and f representing a nonlinear activation function;
s6.2, a pooling layer: selecting the data characteristics after convolution extraction, screening out the characteristics meeting the requirements according to the difference of information quantity, removing unimportant characteristics, simultaneously reducing data dimensionality and avoiding overfitting, calculating a function value for the data characteristics in a pooling region, taking the arithmetic mean value of all elements in the region as output through mean pooling, and extracting the most significant characteristic value of a local characteristic plane as output through maximum pooling for extracting shallow characteristics, wherein the mathematical form of a pooling layer is as follows:
Figure DEST_PATH_IMAGE031
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE033
an ith dimension input variable representing the ith pooling layer,
Figure DEST_PATH_IMAGE035
and
Figure DEST_PATH_IMAGE037
respectively representing a multiplication offset term and an addition offset term of a j dimension variable in the l pooling layer,
Figure 324577DEST_PATH_IMAGE029
an output variable representing a j dimension variable in the l pooling layer, and down representing a pooling function;
s6.3, a full connection layer: the convolution network inputs the abstract expression of the original input data into a full connection layer for feature fusion through feature extraction and filtering layer by layer, and forms a one-dimensional feature vector corresponding to a target space, wherein the mathematical form of the full connection layer is as follows:
Figure DEST_PATH_IMAGE039
in the formula (I), the compound is shown in the specification,
Figure 510839DEST_PATH_IMAGE033
the ith dimension neuron representing the l < th > fully connected layer,
Figure DEST_PATH_IMAGE041
represents weights of neuron i to neuron j in the first convolutional layer,
Figure DEST_PATH_IMAGE043
represents the bias term from neuron i to neuron j in the convolution layer of the l layer,
Figure 437207DEST_PATH_IMAGE029
representing an output variable corresponding to a jth convolution kernel in the ith convolution layer, wherein n represents the total number of input neurons of the fully-connected layer, and f represents a nonlinear activation function;
s6.4, an active layer: taking the land coverage ratio of the mixed cells as a convolutional network training target, setting an activation function adopted by the full connection layer as a Sigmoid activation function, wherein the mathematical form is as follows:
Figure DEST_PATH_IMAGE045
in the formula (I), the compound is shown in the specification,
Figure 658103DEST_PATH_IMAGE029
representing the input of the active layer.
6. The land use suitability probability generation method considering spatial zoning according to claim 1, characterized in that: in step S3, the driver data set is resampled by the software arcgis 10.1.
7. The land use suitability probability generation method considering spatial zoning according to claim 6, wherein: in the step S3, the driving factor data set is set to the original 30m30m resolution driving factor data set.
8. The land use suitability probability generation method considering spatial zoning according to claim 2, characterized in that: in the step S2.3, the calculation tool of the ratio of each area to the area of the fishing net is set as a partition statistical tool of the arcgis10.1 software.
CN202210756949.9A 2022-06-30 2022-06-30 Land use suitability probability generation method considering spatial partition Active CN114819112B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210756949.9A CN114819112B (en) 2022-06-30 2022-06-30 Land use suitability probability generation method considering spatial partition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210756949.9A CN114819112B (en) 2022-06-30 2022-06-30 Land use suitability probability generation method considering spatial partition

Publications (2)

Publication Number Publication Date
CN114819112A true CN114819112A (en) 2022-07-29
CN114819112B CN114819112B (en) 2022-09-16

Family

ID=82523497

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210756949.9A Active CN114819112B (en) 2022-06-30 2022-06-30 Land use suitability probability generation method considering spatial partition

Country Status (1)

Country Link
CN (1) CN114819112B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117172094A (en) * 2023-07-25 2023-12-05 内蒙古师范大学 Positive and negative influence visualization and quantification method for land utilization change driving factors

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107807909A (en) * 2017-10-12 2018-03-16 中国地质大学（北京） A kind of Land in Regional Land purposes policy of regulation and control simulation and effect analysis method
CN108537710A (en) * 2018-03-26 2018-09-14 重庆交通大学 A kind of urban growth boundary demarcation method based on Markov-FLUS models
CN109359166A (en) * 2018-10-10 2019-02-19 广东国地规划科技股份有限公司 A kind of space increasing trends simulate calculation method synchronous with driving force factors contribution degree
CN109543278A (en) * 2018-11-16 2019-03-29 华南农业大学 Simulation of land use changes method and system
CN112069573A (en) * 2020-08-24 2020-12-11 深圳大学 City group space simulation method, system and equipment based on cellular automaton
CN112182951A (en) * 2020-08-31 2021-01-05 广东工业大学 Land utilization planning method considering multiple uncertainties
CN113240257A (en) * 2021-04-30 2021-08-10 陕西华地勘察设计咨询有限公司 Territorial space partitioning method and device based on minimum cumulative resistance model
US20210398289A1 (en) * 2020-06-18 2021-12-23 International Business Machines Corporation Terrain-based automated detection of well pads and their surroundings

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107807909A (en) * 2017-10-12 2018-03-16 中国地质大学（北京） A kind of Land in Regional Land purposes policy of regulation and control simulation and effect analysis method
CN108537710A (en) * 2018-03-26 2018-09-14 重庆交通大学 A kind of urban growth boundary demarcation method based on Markov-FLUS models
CN109359166A (en) * 2018-10-10 2019-02-19 广东国地规划科技股份有限公司 A kind of space increasing trends simulate calculation method synchronous with driving force factors contribution degree
CN109543278A (en) * 2018-11-16 2019-03-29 华南农业大学 Simulation of land use changes method and system
US20210398289A1 (en) * 2020-06-18 2021-12-23 International Business Machines Corporation Terrain-based automated detection of well pads and their surroundings
CN112069573A (en) * 2020-08-24 2020-12-11 深圳大学 City group space simulation method, system and equipment based on cellular automaton
CN112182951A (en) * 2020-08-31 2021-01-05 广东工业大学 Land utilization planning method considering multiple uncertainties
CN113240257A (en) * 2021-04-30 2021-08-10 陕西华地勘察设计咨询有限公司 Territorial space partitioning method and device based on minimum cumulative resistance model

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
XINXIN WU等: "Simulating mixed land-use change under multi-label concept by integrating a convolutional neural network and cellular automata: a case study of Huizhou, China", 《GISCIENCE & REMOTE SENSING》 *
崔学森: "基于地理加权CLUE－S模型的土地利用变化分区模拟", 《江苏农业科学》 *
李柏延: "银川盆地土地变化生态效应时空差异分析", 《中国优秀硕士学位论文全文数据库 (经济与管理科学辑)》 *
谢志文: "利用深度学习构建城市扩展元胞自动机模型", 《中国优秀硕士学位论文全文数据库 (经济与管理科学辑)》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117172094A (en) * 2023-07-25 2023-12-05 内蒙古师范大学 Positive and negative influence visualization and quantification method for land utilization change driving factors

Also Published As

Publication number Publication date
CN114819112B (en) 2022-09-16

Similar Documents

Publication Publication Date Title
Liao et al. Land use optimization of rural production–living–ecological space at different scales based on the BP–ANN and CLUE–S models
CN109508360B (en) Geographical multivariate stream data space-time autocorrelation analysis method based on cellular automaton
CN106780089B (en) Permanent basic farmland planning method based on neural network cellular automaton model
CN110458048A (en) Take population distribution Spatio-temporal Evolution and the cognition of town pattern feature into account
Motlagh et al. Spatial modeling of land-use change in a rapidly urbanizing landscape in central Iran: Integration of remote sensing, CA-Markov, and landscape metrics
CN109359166B (en) Space growth dynamic simulation and driving force factor contribution degree synchronous calculation method
CN102360458B (en) Fuzzy evaluation method for selection of immigrant settlement areas
Zuo et al. Impacts of future land use changes on land use conflicts based on multiple scenarios in the central mountain region, China
CN110298411A (en) A kind of impaired identification and evaluation method of group of cities ecological space
CN107403253A (en) The method and apparatus for monitoring farmland quality
CN115310806A (en) Flood disaster loss evaluation method based on spatial information grid
CN107134141A (en) Consider the expression of large-scale road network space-time traffic behavior and the analysis method of spatial structure characteristic
CN106909899A (en) A kind of analysis method and analysis system of wetland landscape evolution process
CN107766825A (en) Land-based area province cities and counties&#39; space planning 3rd area recognition methods based on space function unit
CN102646164A (en) Land use change modeling method and system implemented in combination with spatial filtering
Aithal et al. Characterization and visualization of spatial patterns of urbanisation and sprawl through metrics and modeling
CN114723283A (en) Ecological bearing capacity remote sensing evaluation method and device for urban group
CN114819112B (en) Land use suitability probability generation method considering spatial partition
CN107665376A (en) A kind of Wetland Space changes in distribution framework analogue and Forecasting Methodology
Xu et al. Land-use change modeling with cellular automata using land natural evolution unit
CN112070304A (en) City group element interaction measuring method, equipment and storage medium
CN115049160A (en) Method and system for estimating carbon emission of plain industrial city by using space-time big data
CN110222133A (en) A kind of multiclass land use simulating and predicting method and device considering landscape index
CN113282572B (en) Gridding division method and system for farmland and terminal equipment
Zhai et al. Using spatial heterogeneity to strengthen the neighbourhood effects of urban growth simulation models

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
PE01 Entry into force of the registration of the contract for pledge of patent right
PE01 Entry into force of the registration of the contract for pledge of patent right

Denomination of invention: A Probability Generation Method for Land Use Suitability Considering Spatial Zoning

Granted publication date: 20220916

Pledgee: China Construction Bank Corporation Nanjing Jiangbei new area branch

Pledgor: Nanjing Beidou innovation and Application Technology Research Institute Co.,Ltd.

Registration number: Y2024980005311