Multi-source heterogeneous earth big data spatio-temporal seamless monitoring method for urban-rural settlement integration transformation

Abstract

The present application relates to a kind of based on the seamless monitoring method of urban-rural settlement fusion mode transformation of multi-source heterogeneous earth big data fusion, the purpose is to solve the technical problem of the collaborative observation vacancy of current urban-rural settlement fusion mode, technical scheme is: collection earth big data, data preprocessing is carried out, machine learning algorithm of random forest classifier is used, and seamless urban-rural settlement type atlas of time and space is produced;Reveal the trajectory analysis of settlement type change and population movement between urban and rural areas, carry out urban-rural settlement fusion mode identification, using time series change detection method, determine the occurrence year, time sequence track, occurrence frequency of urban-rural settlement fusion mode transformation;The present application integrates existing multi-source heterogeneous earth big data, machine learning method, can carry out the seamless monitoring of all types of urban-rural settlement fusion mode transformation, has the ability of long time sequence extension and fine portrayal of spatial distribution, can enrich existing land survey system and provide solid data for national urban-rural integration regulation and control decision.

Description

Technical Field

[0001] This invention belongs to the field of Earth big data fusion and extraction technology, specifically involving a spatiotemporal seamless monitoring method for multi-source heterogeneous Earth big data in the integration and transformation of urban and rural settlements. Background Technology

[0002] Currently, traditional data sources used by scholars both domestically and internationally to analyze the spatial characteristics and evolution of urban and rural settlements mainly include land survey data, UAV aerial photography, and satellite imagery. While the first two types of data can determine the spatial scale and morphological characteristics of rural settlements, the periodicity of surveys and the uninterrupted nature of flights mean that the continuity of results cannot be guaranteed. Furthermore, the high human and financial costs required for data acquisition limit the coverage of such data. In contrast, long-term, high-precision satellite imagery, combined with the data processing capabilities of cloud computing platforms, can completely compensate for the shortcomings of the above data. Urban and rural settlements are inherently characterized by land as the carrier and population as the core. However, past research has largely focused on interpreting settlements from the perspective of land carriers, lacking representation of population information. In addition, there is currently a lack of collaborative observation of urban-rural settlement integration models such as urban-rural segmentation, urban-rural dualism, urban-rural integration, and urban-rural fusion. Therefore, from the perspective of human-land relations, it is essential to integrate heterogeneous Earth observation data, including land cover remote sensing products and geographic information data, and combine this with machine learning algorithms to deepen the quantitative measurement of the transformation of urban-rural settlement integration models. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems and provide a spatiotemporal seamless monitoring method for multi-source heterogeneous Earth big data in the transformation and integration of urban and rural settlements.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0005] A spatiotemporal seamless monitoring method for multi-source heterogeneous Earth big data in the integration and transformation of urban and rural settlements includes the following steps:

[0006] Step S1: Collect big data about Earth and perform data preprocessing;

[0007] The specific steps of step S1 are as follows:

[0008] Step S1.1: Acquire Earth big data from different data sources with different data structures, and use geographic information system software to perform data preprocessing operations such as georegistration, projection transformation, and raster clipping on the collected multi-source heterogeneous Earth big data in sequence.

[0009] Step S1.2: Examine and correct the misclassified pixels in the land cover type remote sensing product obtained from the data preprocessing in step S1.1 to obtain an accurate year-by-year land cover type raster layer.

[0010] Step S1.3: Clean the mobile signaling data using data mining and machine learning algorithms;

[0011] Step S1.4: Upload the preprocessed Earth big data to the GEE cloud processing platform to build a basic geographic database;

[0012] Step S2: Using annual land cover mapping products and population density geographic grids as input data, a machine learning algorithm with a random forest classifier is used to produce a spatiotemporally seamless urban and rural settlement type map for spatialized urban settlements, rural settlements and other land use types.

[0013] The specific steps of step S2 are as follows:

[0014] Step S2.1: Select samples from existing settlement mapping products for the corresponding year, and then use a generative adversarial neural network to transfer the data to other time periods to obtain long-term classification samples of urban and rural settlement types covering the entire research period.

[0015] Step S2.2: Using the migrated samples, land cover remote sensing products, and population density geographic grids as input data, train a random forest classifier to achieve the subdivision of primary and secondary classes for urban settlements, rural settlements, and other land use types;

[0016] Step S3: Using the results of the urban and rural settlement type map and mobile phone signaling data as input data, reveal the trajectory analysis of changes in settlement types and population movement between urban and rural areas, and also use machine learning methods to identify the urban and rural settlement integration pattern.

[0017] The specific steps of step S3 are as follows:

[0018] Step S3.1: Use the spatial overlay method to identify the trajectory of settlement type change between urban and rural areas using the urban and rural settlement type map cube;

[0019] Step S3.2: Using prefecture-level cities as units, calculate the interannual movement trajectory indicators between urban and rural settlements in the prefecture-level city using mobile phone signaling data, and calculate the explicit and implicit indicators of urban-rural settlement integration based on the indicators.

[0020] Step S3.3: Using prefecture-level cities as units, the explicit and implicit indicators of urban-rural settlement integration obtained in steps S3.1 and S3.2 are divided into thresholds to identify the urban-rural settlement integration patterns year by year.

[0021] S4. Using time-series change detection methods, determine the year, time-series trajectory, and frequency of occurrence of the transformation of the urban-rural settlement integration model;

[0022] S4.1 Using the time-series change detection method, identify the year and frequency of the transformation of the urban-rural settlement integration pattern in each prefecture-level city of the urban-rural settlement integration pattern cube;

[0023] S4.2 Compare the urban-rural settlement integration patterns before and after the year of occurrence to identify its temporal transformation trajectory.

[0024] Furthermore, the Earth big data in step S1.1 includes land cover remote sensing mapping products, population density geographic rasters, building distribution data, mobile phone signaling data, and administrative boundary areal vector data.

[0025] Furthermore, the movement trajectory indicators in step S3.2 include the direction of population movement and the frequency of movement.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] This invention integrates existing multi-source heterogeneous Earth big data and machine learning methods, enabling seamless spatiotemporal monitoring of the transformation of urban-rural settlement integration models across all types. It possesses the ability to extend long-term series and finely characterize spatial distribution, enriching the existing land survey system and providing solid data for subsequent national urban-rural integration regulation and decision-making. Attached Figure Description

[0028] Figure 1 This is a schematic diagram of the process of the present invention; Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0030] A spatiotemporal seamless monitoring method for multi-source heterogeneous Earth big data in the integration and transformation of urban and rural settlements includes the following steps:

[0031] Step S1: Collect big data about Earth and perform data preprocessing;

[0032] The specific steps of step S1 are as follows:

[0033] Step S1.1: Using tools such as Georeferencing, Projections and Transformations, and Raster Processing in ArcGIS software, perform georegistration, projection transformation, and raster clipping on the acquired Earth big data from different data sources with different data structures, including land cover remote sensing mapping products, population density geographic rasters, building distribution data, mobile phone signaling data, and administrative boundary areal vector data, to obtain the corresponding Earth big data vector files (.shp) or raster images (.tif).

[0034] Step S1.2: Examine and correct the misclassified pixels in the land cover type remote sensing product obtained from the data preprocessing in step S1.1 to obtain an accurate year-by-year land cover type raster layer.

[0035] For example, the specific steps for checking and correcting misclassified pixels in the obtained land cover type remote sensing product are as follows: Suppose a pixel is in t i-1 t i-2 and t i+1 t i+2 The average annual construction land area is only t i If the year is grassland, then the pixel in t i The year was considered to have been incorrectly classified as grassland. Therefore, the land cover type of that pixel was changed from grassland to construction land consistent with the adjacent year. The corrected land cover data was obtained by checking the temporal consistency of each pixel.

[0036] Step S1.3: Use functions such as duplicated(), drop_duplicates(), pandas.isnull(), dropna(), and fillna() from the Pandas library in Python to perform cleaning operations on the collected mobile signaling data, including removing duplicate records, handling missing and outlier values, anonymizing, and removing invalid data, to obtain clean mobile signaling data.

[0037] Step S1.4: Upload the vector and raster format Earth big data obtained from the preprocessing in steps S1.1, S1.2, and S1.3 to the GEE cloud processing platform to build a basic geographic database;

[0038] Step S2: Using annual land cover mapping products and population density geographic grids as input data, a machine learning algorithm with a random forest classifier is used to produce a spatiotemporally seamless urban and rural settlement type map for spatialized urban settlements, rural settlements and other land use types.

[0039] The specific steps of step S2 are as follows:

[0040] Step S2.1: Stretch the limited urban and rural settlement type samples by year span. Specifically, on the GEE cloud processing platform, select a sufficient number of different types of settlement samples corresponding to the year from existing urban and rural settlement mapping products with limited years, such as GHS_SMOD R2022A and FROM_GLC.

[0041] Using the selected samples, statistically analyze the land cover type (f1), shape characteristics (f2), population density (f3), building density (f4), and topography (f5) for the corresponding location in the corresponding year. nThe sample feature space F(p) = {f1(p), f2(p), f3(p), f4(p), ..., f...} constitutes a certain sample point p. n (p)};Calculate the mean space of the feature space of each category of samples. ,in The number of categories in the sample;

[0042] Similarly, these samples are used to statistically analyze land cover types in other years requiring migration (f1) ’ ), shape features (f2) ’ ), population density (f3) ’ ), building density (f4) ’ ), topography and landforms (f n ’ ), calculate the target feature space T(p)={ f1(p)', f2(p)', f3(p)', f4(p)',…, fn(p)'} for a certain sample point p; use a generative adversarial neural network (GAN) to learn the target feature space T(p) and the mean space M(p) to realize the transfer of samples to other missing years.

[0043] Step S2.2: Using the migrated samples, land cover remote sensing products, and population density geographic grids as input data, train a random forest classifier to ultimately achieve continuous mapping of urban settlements, rural settlements, and other land use types for missing years in the study period.

[0044] Step S3: Using the results of urban-rural continuous spectrum mapping and mobile phone signaling data as input data, reveal the trajectory analysis of changes in settlement types and population movement between urban and rural areas, and also use machine learning methods to identify urban-rural settlement integration patterns.

[0045] The specific steps of step S3 are as follows:

[0046] Step S3.1: Use the spatial overlay method to identify the trajectory of settlement type change between urban and rural areas using the urban and rural settlement type map cube;

[0047] Step S3.2: Based on the above-calculated changes in urban and rural settlement types and population movement trajectories, two explicit indicators of urban-rural settlement integration (urban settlement expansion rate) are obtained at the prefecture-level city level. Rural settlement expansion rate ) and two implicit indicators of urban-rural integration (the number of people migrating from rural to urban areas) Number of people migrating from urban to rural areas );

[0048] Step S3.3: Based on the above four indicators, the urban-rural settlement integration model of prefecture-level city units is divided into six categories: urban-rural separation, urban-rural dualism (rural-urban development), urban-rural dualism (rural decline), urban-rural dualism (co-decline of rural and urban areas), urban-rural integration, and urban-rural fusion.

[0049] like >0, >0, = =0, then it is divided into urban and rural areas; if >0, >0, If the value is >0, it is divided into urban-rural dual structure (rural-urban development);

[0050] like <0, >0, If the value is >0, then it is classified as a dual urban-rural structure (rural decline);

[0051] like <0, <0, If the value is >0, then the urban-rural dual structure is divided (urban and rural areas decline together).

[0052] like >0, >0, This is then categorized as urban-rural integration;

[0053] like >0, >0, If so, it is classified as urban-rural integration.

[0054] S4. Using time-series change detection methods, determine the year, time-series trajectory, and frequency of occurrence of the transformation of the urban-rural settlement integration model;

[0055] S4.1. For the urban-rural settlement integration pattern map cube, use the time-series change detection method to identify the years (y1, y2, ..., yj) of the first, second, ..., jth transformations of the urban-rural settlement integration pattern for each prefecture-level city. j ), frequency of occurrence j;

[0056] S4.2, Use ArcGIS to analyze the year y. j The two phases of urban-rural settlement integration models (P) yj-1 , P yj Overlay analysis was performed to identify the temporal transformation trajectory of a certain urban-rural settlement integration pattern.

Claims

1. A spatiotemporal seamless monitoring method for the transformation of urban-rural settlement integration patterns based on the fusion of multi-source heterogeneous Earth big data, characterized in that, Includes the following steps: Step S1: Collect big data about Earth and perform data preprocessing; The specific steps of step S1 are as follows: Step S1.1: Acquire Earth big data from different data sources with different data structures, and use geographic information system software to perform data preprocessing operations such as georegistration, projection transformation, and raster clipping on the collected multi-source heterogeneous Earth big data in sequence. Step S1.2: Examine and correct the misclassified pixels in the land cover type remote sensing product obtained from the data preprocessing in step S1.1 to obtain an accurate year-by-year land cover type raster layer. Step S1.3: Clean the mobile signaling data using data mining and machine learning algorithms; Step S1.4: Upload the preprocessed Earth big data to the GEE cloud processing platform to build a basic geographic database; Step S2: Using annual land cover mapping products and population density geographic grids as input data, a machine learning algorithm with a random forest classifier is used to produce a spatiotemporally seamless urban and rural settlement type map for spatialized urban settlements, rural settlements and other land use types. The specific steps of step S2 are as follows: Step S2.1: Select samples from existing settlement mapping products for the corresponding year, and then use a generative adversarial neural network to transfer the data to other time periods to obtain long-term classification samples of urban and rural settlement types covering the entire research period. Step S2.2: Using the migrated samples, land cover remote sensing products, and population density geographic grids as input data, train a random forest classifier to achieve the subdivision of primary and secondary classes for urban settlements, rural settlements, and other land use types; Step S3: Using the results of the urban and rural settlement type map and mobile phone signaling data as input data, reveal the trajectory analysis of changes in settlement types and population movement between urban and rural areas, and also use machine learning methods to identify the urban and rural settlement integration pattern. The specific steps of step S3 are as follows: Step S3.1: Use the spatial overlay method to identify the trajectory of settlement type change between urban and rural areas using the urban and rural settlement type map cube; Step S3.2: Using prefecture-level cities as units, calculate the interannual movement trajectory indicators between urban and rural settlements in the prefecture-level city using mobile phone signaling data, and calculate the explicit and implicit indicators of urban-rural settlement integration based on the indicators. Step S3.3: Using prefecture-level cities as units, the explicit and implicit indicators of urban-rural settlement integration obtained in steps S3.1 and S3.2 are divided into thresholds to identify the urban-rural settlement integration patterns year by year. S4. Using time-series change detection methods, determine the year, time-series trajectory, and frequency of occurrence of the transformation of the urban-rural settlement integration model; S4.1 Using the time-series change detection method, identify the year and frequency of the transformation of the urban-rural settlement integration pattern in each prefecture-level city of the urban-rural settlement integration pattern cube; S4.2 Compare the urban-rural settlement integration patterns before and after the year of occurrence to identify its temporal transformation trajectory.

2. The spatiotemporal seamless monitoring method for the transformation of urban-rural settlement integration model based on the fusion of multi-source heterogeneous Earth big data as described in claim 1, characterized in that, The Earth big data in step S1.1 includes land cover remote sensing mapping products, population density geographic rasters, building distribution data, mobile phone signaling data, and administrative boundary areal vector data.

3. The spatiotemporal seamless monitoring method for the transformation of urban-rural settlement integration model based on the fusion of multi-source heterogeneous Earth big data as described in claim 1, characterized in that, The movement trajectory indicators in step S3.2 include the direction of population movement and the frequency of movement.

Images