Resident welfare space simulation method fusing remote sensing and perception investigation

Abstract

The invention discloses a resident welfare space simulation method fusing remote sensing and perceptual survey, which comprises the following steps of: constructing a resident welfare characteristic index system by utilizing multi-source satellite remote sensing data, constructing a resident economic welfare grid sample with the spatial resolution of 30m in combination with home-entry survey data, and introducing a deep neural network machine learning algorithm to simulate the resident welfare space. A resident welfare space evaluation intelligent model fusing remote sensing and sensing survey data is constructed, and village and town scale space welfare simulation is achieved. According to the method, advantage complementation of resident perception quantitative information and geographical environment feature information can be realized, the resident welfare spatial distribution features can be evaluated more comprehensively and accurately, and a scientific basis is provided for policy making and social development.

Description

Technical Field

[0001] The present invention belongs to the application of multi-source remote sensing data in the field of resident well-being space assessment, and specifically relates to a resident well-being space simulation method integrating remote sensing and perception survey. Background Art

[0002] Residents’ well-being is an important indicator for measuring social development and people’s quality of life. Its indicator system is complex and diverse, including objective economic, environmental, and social factors as well as subjective perceptions of well-being and happiness. Therefore, the assessment of residents’ well-being is crucial.

[0003] Due to the differences in the indicator systems and evaluation methods used by various countries to study resident well-being, the level of detail of basic data and information varies from country to country, and there are differences in the measurement indicators in various fields. There is no fully recognized indicator system. With the intensification of global economic inequality and other issues, the assessment of resident well-being that relies on statistical data and survey reports has limitations in spatial coverage and real-time performance, making it difficult to evaluate economic well-being on a large scale.

[0004] Satellite remote sensing is an important means of obtaining the characteristics of residents' welfare. It has the characteristics of large range, objectivity and accuracy. How to integrate it with household survey data, by extracting geographic information characteristics related to residents' welfare, reflecting the economic development level, ecological environment quality and infrastructure conditions of the residents in the region, is an effective means to achieve rapid and accurate spatial assessment of the economic welfare of residents in a large area, and then reflect the spatial distribution characteristics of residents' welfare, providing a scientific basis for policy formulation and social development. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the spatial simulation method of residents' well-being that integrates remote sensing and perception surveys provided by the present invention can achieve the complementary advantages of residents' perception survey information and geographic environment characteristic information, and more comprehensively and accurately evaluate the spatial distribution characteristics of residents' well-being, providing a scientific basis for policy formulation and social development.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a resident welfare space simulation method integrating remote sensing and perception survey, comprising the following steps:

[0007] S1. Collect multi-source data of residents’ welfare spatial areas and pre-process them;

[0008] S2, extract geographic information features related to residents' well-being from the preprocessed data and unify them to the same resolution grid;

[0009] The geographic information features include night light intensity, vegetation index, impervious surface index, road density, urban area ratio, farmland area ratio, elevation fluctuation index, housing concentration index and population dispersion index;

[0010] S3. Construct a multi-dimensional indicator system for comprehensive resident welfare, obtain its indicator data, and construct a sample of resident welfare spatial assessment based on the divided grids;

[0011] S4. Construct a deep learning neural network, use the extracted geographic information features as input features, use the resident welfare spatial assessment samples in the corresponding area as labels, train the deep learning neural network, and obtain a resident welfare assessment model;

[0012] S5. Use the geographic information characteristics of the area to be evaluated as the input of the resident welfare evaluation model to predict the evaluation results of the resident welfare space.

[0013] Further, in step S1, the multi-source data includes multi-source optical and radar data, night light data, located point of interest data, land use data and statistical yearbook data;

[0014] The pre-processing comprises:

[0015] Preprocessing of multi-source optical and radar data includes radiometric calibration, atmospheric correction, band registration, multi-band image synthesis and image fusion of optical remote sensing data, as well as orbit correction, image clipping, filtering, radiometric calibration, terrain correction, geocoding and image mosaicking of radar data;

[0016] The preprocessing of night light data includes radiation calibration, atmospheric correction, geometric correction, and data clipping and splicing;

[0017] The preprocessing of the located POI data includes duplicate data removal, data cleaning, coordinate system unification and data classification.

[0018] Further, in step S2, the night light intensity includes the total value of night light intensity, the mean value of night light intensity and the standard deviation of night light intensity;

[0019] The vegetation index includes the normalized difference vegetation index and the enhanced vegetation index;

[0020] The impervious surface index includes a normalized difference impervious surface index and a ratio-type building index.

[0021] Furthermore, the total nighttime light intensity TNL is: Among them, DN i represents the i-th gray level of all valid pixels in the night light data after preprocessing, C irepresents the number of pixels corresponding to the gray level, and I represents the maximum gray level;

[0022] The nighttime light intensity mean NLM is:

[0023] The nighttime light intensity standard deviation NSTD is:

[0024] The normalized vegetation index NVDI is expressed as: NVDI=(NIR-R) / (NIR+R), wherein NIR represents the reflectivity of the near infrared band, and R represents the reflectivity of the red band;

[0025] The enhanced vegetation index EVI is: EVI = 2.5 × (NIR-R) / (NIR + 6R-7.5B + ​​1), where B represents the reflectivity of the blue band;

[0026] The normalized difference impervious surface index NDISI is: Among them, SWIR represents the reflectivity of the short-wave infrared band, NIR represents the reflectivity of the near-infrared band, Green represents the reflectivity of the green band, and MIR1 represents the reflectivity of the mid-infrared band 1;

[0027] The ratio-based building index RBI is: Among them, NIR represents the reflectivity of the near-infrared band, and Red represents the reflectivity of the red band;

[0028] The road density index RDI is: RDI = total road length / area of ​​study area;

[0029] The urban area ratio is: urban area ratio = urban area / study area area;

[0030] The farmland area ratio is: farmland area ratio = farmland area / study area area;

[0031] The elevation fluctuation index σ is: H avg =∑H t / N, where H i represents the elevation value of the t-th pixel, T represents the total number of pixels, H avg Indicates the average elevation value;

[0032] The housing aggregation index HAI is: HAI = ∑(n i ) / G, where n i represents the number of houses within the radius r of each house, and G represents the total number of houses in the study area;

[0033] The population dispersion index PDI is: Pk =P ak / P total , Among them, P k represents the population proportion of each administrative unit, represents the average population proportion, P ak represents the population of the kth administrative unit, P total represents the total population of the study area, and n represents the number of administrative units in the study area.

[0034] Furthermore, step S3 includes the following steps:

[0035] S31. Construct a multidimensional indicator system;

[0036] The multidimensional indicator system includes primary indicators and secondary indicators under them, and the primary indicators include economic factor indicators, environmental factor indicators and welfare perception indicators;

[0037] S32. According to the constructed multi-dimensional indicator system, quantitative indicator data of perception of different objects are collected through household surveys, and the data are divided into grids;

[0038] S33, selecting a grid containing an indicator item exceeding a set threshold as a sample grid;

[0039] S34. Based on the indicator data in the sample grid, quantitative statistics are conducted on the welfare satisfaction of the respondents in each assessment area, and the average welfare of the residents is calculated;

[0040] S35. Calculate the average welfare of residents in all sample grids, and then construct a spatial assessment sample of resident welfare.

[0041] Furthermore, in step S32, the multidimensional indicator system forms a questionnaire based on scale analysis according to the determined secondary indicators, and collects quantitative indicator data perceived by different subjects through a questionnaire household survey;

[0042] In the questionnaire, each indicator is provided with options corresponding to different weights, and all options are weighted and counted to form a perceptual quantitative survey result.

[0043] Furthermore, the step S4 includes the following sub-steps:

[0044] S41. Construct a feature set for evaluating residents’ well-being based on the extracted geographic information features;

[0045] S42. According to the location of the grid where the resident welfare assessment sample is located, determine its relevant geographic information features, build a corresponding relationship between the sample and the geographic information features, and form a training data set;

[0046] S43. Use the geospatial information features in the training set as input features and the corresponding resident welfare assessment samples as labels to train the deep learning neural network and obtain a resident welfare assessment model.

[0047] Further, in step S4, the resident welfare assessment model includes an input layer, a hidden layer and an output layer connected in sequence;

[0048] The number of neurons in the input layer corresponds to the number of geographic information features, and the function f(x,w) of the predicted value of the resident welfare space output by the output layer is expressed as:

[0049]

[0050] In the formula, P represents the number of geographic information features, x pq represents the value of the p-th type of geographic information feature of the q-th sample, x pq represents the corresponding weight parameter, b q Indicates the offset.

[0051] Furthermore, in step S4, the deep learning neural network is trained using a mean square error loss function and an error back propagation algorithm, and the mean square error loss function is optimized using a batch stochastic gradient descent method.

[0052] Furthermore, the step S4 further includes:

[0053] The mean absolute error and precision were used as evaluation indicators, and the validation sample was used to verify the accuracy of the resident welfare assessment model.

[0054] The beneficial effects of the present invention are:

[0055] (1) By integrating multiple geographical factors that affect residents' well-being, the simulation results can show the spatial differences in residents' well-being at the grid scale, breaking through the bottleneck problem that large-scale economic well-being is difficult to simulate and evaluate spatially;

[0056] (2) Multi-source remote sensing data and geospatial data are used to extract residents’ welfare-related features, with diverse feature types and high extraction accuracy;

[0057] (3) A spatial assessment model of residents’ well-being is constructed through a deep neural network algorithm, which has high spatial applicability and can deeply explore the complex correlation mechanism between multiple features and residents’ well-being. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A flow chart of the method for spatial simulation of resident welfare by integrating remote sensing and perception survey provided by the present invention. DETAILED DESCRIPTION

[0059] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0060] The embodiment of the present invention provides a method for simulating resident welfare space by integrating remote sensing and perception survey, such as Figure 1 As shown, the following steps are included:

[0061] S1. Collect multi-source data of residents’ welfare spatial areas and pre-process them;

[0062] S2, extract geographic information features related to residents' well-being from the preprocessed data and unify them to the same resolution grid;

[0063] Among them, geographic information features include night light intensity, vegetation index, impervious surface index, road density, urban area ratio, farmland area ratio, elevation fluctuation index, housing concentration index and population dispersion index;

[0064] S3. Construct a multi-dimensional indicator system for comprehensive resident welfare, obtain its indicator data, and construct a sample of resident welfare spatial assessment based on the divided grids;

[0065] S4. Construct a deep learning neural network, use the extracted geographic information features as input features, use the resident welfare spatial assessment samples in the corresponding area as labels, train the deep learning neural network, and obtain a resident welfare assessment model;

[0066] S5. Use the geographic information characteristics of the area to be evaluated as the input of the resident welfare evaluation model to predict the evaluation results of the resident welfare space.

[0067] In step S1 of the embodiment of the present invention, the multi-source data includes multi-source optical and radar data, night light data, located point of interest data, land use data, and statistical yearbook data.

[0068] The preprocessing of the above data includes:

[0069] Preprocessing of multi-source optical and radar data includes radiometric calibration, atmospheric correction, band registration, multi-band image synthesis and image fusion of optical remote sensing data, as well as orbit correction, image clipping, filtering, radiometric calibration, terrain correction, geocoding and image mosaicking of radar data;

[0070] The preprocessing of night light data includes radiation calibration, atmospheric correction, geometric correction, and data clipping and splicing;

[0071] The preprocessing of the located POI data includes duplicate data removal, data cleaning, coordinate system unification and data classification.

[0072] Specifically, in this embodiment, for multi-source optical and radar data:

[0073] Obtain optical remote sensing data from Landsat 8, Gaofen-1, Gaofen-6 and Sentinel-2, radar remote sensing data from Gaofen-3 and Sentinel-1, and night light remote sensing data from SDGSAT-1 in the study area. The optical remote sensing data are all cloud-free, ice-free and snow-free data.

[0074] Among them, the optical remote sensing data of Landsat 8, Gaofen-1, Gaofen-6 and Sentinel-2 include 15-meter multispectral images, 2-meter panchromatic\8-meter multispectral, 10-meter, 20-meter, 60-meter and other multispectral images; after the images are acquired, they are pre-processed by radiation calibration, atmospheric correction, geometric correction, band registration, multi-band image synthesis, image fusion and so on.

[0075] The Gaofen-3 and Sentinel-1 satellites are equipped with a C-band synthetic aperture radar with multiple imaging modes and a maximum resolution of 1 meter. After the image is acquired, orbit correction, image cropping, filtering, radiation calibration, terrain correction, geocoding, image mosaicking and other pre-processing are carried out.

[0076] The preprocessed radar images, optical images and digital elevations are unified to a resolution of 30m through image registration and resampling.

[0077] In this implementation, for night light data:

[0078] The SDGSAT-1 is equipped with a low-light and multi-spectral imager to collect nighttime light images, which has one panchromatic band and three color bands, including nighttime light images with a resolution of 30 meters. After the images are acquired, they are pre-processed by radiation calibration, atmospheric correction, geometric correction, data cropping and splicing.

[0079] In this embodiment, for the located point of interest data:

[0080] POI (point of interest) data is collected through online map data platforms and mobile device positioning, including data information on various commercial facilities, public service facilities, industrial areas and township service points in towns and villages; the acquired data is pre-processed by removing duplicate data, data cleaning, coordinate system one, data classification statistics, etc.

[0081] In this embodiment, for land use data:

[0082] The Landsat series satellites and Sentinel-2 satellites were selected mainly through multispectral images. Landsat satellites carry sensors such as Multispectral Scanner (MSS), Thematic Mapper (TM), Enhanced Thematic Mapper (ETM+) and Land Imager (OLI). Sentinel-2 satellites carry Multispectral Imager (MSI), covering 13 spectral bands, which can provide information extraction on vegetation coverage, impervious surfaces, roads, farmland, etc.

[0083] In this embodiment, for statistical yearbook data:

[0084] The China County Statistical Yearbook (Township Volume) is selected, which contains information such as administrative area, registered population, number of village committees, number of industrial enterprises, number of supermarkets, etc.

[0085] In step S2 of the embodiment of the present invention, the night light intensity includes the total value of the night light intensity, the mean value of the night light intensity and the standard deviation of the night light intensity; specifically, based on the SDGSAT-1 night light data obtained by the above preprocessing, remote sensing extraction of night light intensity characteristics is carried out.

[0086] Among them, the total value of night light intensity represents the total level of night light in a region. The night light intensity values ​​of all valid pixels of the preprocessed data of SDGSAT-1 night light remote sensing in the research geographical range are accumulated. Based on this, the total value of night light intensity TNL is:

[0087] The night light intensity mean NLM represents the average nighttime light level in a region. Combined with the calculation result of the total night light intensity, the total night light intensity is divided by the total number of valid pixels to obtain the night light intensity mean NLM:

[0088] The night light intensity standard deviation NTSD represents the difference in the spatial distribution of night light. The square of the difference between the night light intensity value of each pixel and the mean night light intensity is calculated. The square of the difference of all pixels is accumulated and then divided by the total number of valid pixels to obtain the variance. The square root of the variance is taken to obtain the night light intensity standard deviation NSTD:

[0089] Among them, DN i represents the i-th gray level of all valid pixels in the night light data after preprocessing, C i represents the number of pixels corresponding to the gray level, and I represents the maximum gray level.

[0090] In step S2 of this embodiment, the vegetation index is extracted based on the pre-processed Landsat series satellite data containing visible light, near infrared and other bands, and Sentinel-2 satellite data, mainly including the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI).

[0091] Among them, the Normalized Difference Vegetation Index (NDVI) is calculated by dividing the difference between the near-infrared band reflectance (NIR) and the red band reflectance (R) by the sum of the two. The characteristics of vegetation's high reflection in the near-infrared band and strong absorption in the red band are used to highlight the difference between vegetation in these two bands, so as to effectively reflect the growth status and coverage of vegetation; based on this, the Normalized Difference Vegetation Index NVDI is expressed as: NVDI = (NIR-R) / (NIR+R), where NIR represents the reflectance of the near-infrared band and R represents the reflectance of the red band.

[0092] In addition to considering the near-infrared and red bands, the enhanced vegetation index (EVI) also introduces the blue band reflectance (B). EVI has a larger dynamic range in areas with high and low vegetation coverage, and can better reflect the subtle changes in vegetation; based on this, the enhanced vegetation index EVI is: EVI = 2.5 × (NIR-R) / (NIR + 6R-7.5B + ​​1), where B represents the reflectance of the blue band.

[0093] In step S2 of this embodiment, for the impervious surface index feature extraction:

[0094] According to the above-obtained data, high-resolution satellite images and Landsat satellite data in visible and near-infrared bands are extracted, mainly including the Normalized Difference Impervious Surface Index (NDISI) and the Ratio-based Building Index (RBI).

[0095] Among them, the Normalized Difference Impervious Surface Index (NDISI) calculates the reflectivity of the shortwave infrared band (SWIR), near infrared band (NIR), green band (Green) and mid-infrared band 1 (MIR1), cleverly utilizing the reflection differences of different bands on different objects to highlight the characteristics of the impervious surface, thereby achieving effective extraction of the impervious surface. Based on this, the Normalized Difference Impervious Surface Index NDISI is: Among them, SWIR represents the reflectivity of the shortwave infrared band, NIR represents the reflectivity of the near infrared band, Green represents the reflectivity of the green band, and MIR1 represents the reflectivity of the mid-infrared band 1.

[0096] The Ratio Building Index (RBI) uses the reflectivity ratio of the near infrared band (NIR) to the red band (Red), minus 1, to highlight the characteristics of the building, and then indirectly reflect the condition of the impermeable surface; based on this, the Ratio Building Index RBI is: Among them, NIR represents the reflectivity of the near-infrared band, and Red represents the reflectivity of the red band.

[0097] In step S2 of this embodiment, based on the extracted land use data, the road density index, the urban area ratio and the agricultural area ratio are extracted respectively.

[0098] Among them, the road density index (RDI) collects high-resolution remote sensing images and related geographic information data and performs preprocessing, then uses image classification, edge detection or object-oriented methods to extract road information, and obtains the length ratio of roads per unit area, which can intuitively show the density of roads in a region; based on this, the road density index RDI is: RDI = total road length / area of ​​study area.

[0099] Urban area ratio, the area of ​​the extracted urban areas was measured, and the urban area ratio was obtained as follows: urban area ratio = urban area / area of ​​the study area.

[0100] The farmland area ratio was obtained by identifying the farmland area based on the preprocessed Landsat series and Sentinel-2 satellite images obtained above, and the farmland area ratio was obtained as follows: farmland area ratio = farmland area / study area.

[0101] In step S2 of this embodiment, the elevation fluctuation index is usually used to describe the degree of fluctuation of the terrain in a region. After preprocessing the DEM data, the terrain features are extracted. The elevation values ​​of all pixels are accumulated and summed, and then divided by the total number of pixels. The result is the average elevation value. For the elevation value Hi of each pixel, first calculate the difference between it and the average elevation value H_avg, and then square this difference. Finally, take the square root of the difference, and the result is the standard deviation. Based on this, the elevation fluctuation index σ is obtained as: H avg =∑H t / N, where H i represents the elevation value of the t-th pixel, T represents the total number of pixels, H avg Indicates the average elevation value; the larger the standard deviation, the greater the terrain fluctuation, and the smaller the standard deviation, the relatively flat terrain.

[0102] In step S2 of this embodiment, for the housing density index, the search radius r is determined based on factors such as the size of the study area and the distribution characteristics of the houses. For each house, the number of houses ni within the search radius r is calculated. The number of houses of all houses within the search radius is accumulated and summed, and then divided by the total number of houses N to obtain a density-based housing aggregation index. The more houses there are, the denser the houses in the area are in the local area, and the higher the aggregation degree. Based on this, the housing aggregation index HAI is: HAI = ∑(n i ) / G, where n i represents the number of houses within the radius r of each house, and G represents the total number of houses in the study area.

[0103] In step S2 of this embodiment, the population dispersion index is calculated based on the collected population census data of the study area and the geographic spatial data of the study area, including administrative division boundaries, road networks, building distribution, etc. The population dispersion features are extracted based on administrative division units.

[0104] There are n administrative units in the study area, and the population of the i-th administrative unit is P ai , the total population of the study area is P total First, calculate the population proportion p of each administrative unit i , then calculate the average population proportion p, and finally calculate the population dispersion PDI. Based on this, the population dispersion index PDI is: P k =P ak / P total , Among them, P k represents the population proportion of each administrative unit, represents the average population proportion, P ak represents the population of the kth administrative unit, P total represents the total population of the study area, and n represents the number of administrative units in the study area.

[0105] Step S3 of the embodiment of the present invention includes the following steps:

[0106] S31. Construct a multidimensional indicator system;

[0107] Among them, the multidimensional indicator system includes the first-level indicators and their subordinate second-level indicators. The first-level indicators include economic factor indicators, environmental factor indicators and welfare perception indicators;

[0108] S32. According to the constructed multi-dimensional indicator system, quantitative indicator data of perception of different objects are collected through household surveys, and the data are divided into grids;

[0109] S33, selecting a grid containing an indicator item exceeding a set threshold as a sample grid;

[0110] S34. Based on the indicator data in the sample grid, quantitative statistics are conducted on the welfare satisfaction of the respondents in each assessment area, and the average welfare of the residents is calculated;

[0111] S35. Calculate the average welfare of residents in all sample grids, and then construct a spatial assessment sample of resident welfare.

[0112] In step S32 of the present embodiment, the multidimensional indicator system forms a questionnaire based on scale analysis according to the determined secondary indicators, and collects quantitative indicator data perceived by different objects through door-to-door questionnaire surveys; in the questionnaire, each indicator is provided with options corresponding to different weights, and weighted statistics are performed on all options to form a quantitative perception survey result.

[0113] In an optional embodiment of the present invention, in the questionnaire, 21 questions in 8 parts, including basic information, income, work, housing, health, education, quality of life, and happiness perception, are set to conduct a comprehensive and systematic quantitative survey on residents' well-being perception and obtain different indicator data. Furthermore, the answers to different questions are analyzed based on a four-point Likert scale, and corresponding weights are set for different answers, and then weighted statistics are performed on the answers to different questions to form a quantitative survey result of perception.

[0114] In step S33 of this embodiment, the data from the reset resident welfare questionnaire survey are divided into 30m×30m grids, and grids containing more than 5 survey data items are selected as sample grids. Based on the questionnaire survey data, no less than 5,000 resident welfare assessment samples are formed.

[0115] In step S34 of this embodiment, the average welfare of residents is obtained by using the statistical average method:

[0116]

[0117] In step S35 of this embodiment, the average welfare value of residents in the grid is statistically calculated, thereby constructing about 5,000 resident welfare assessment samples, which are expressed as:

[0118] Y={y i ,i=1,2,…,n,n=5000}

[0119] Where Y is the sample set of residents’ welfare assessment, y i is the average resident welfare of the i-th sample, and n is the number of samples for resident welfare assessment.

[0120] Step S4 of the embodiment of the present invention includes the following sub-steps:

[0121] S41. Construct a feature set for evaluating residents’ well-being based on the extracted geographic information features;

[0122] S42. According to the location of the grid where the resident welfare assessment sample is located, determine its relevant geographic information features, build a corresponding relationship between the sample and the geographic information features, and form a training data set;

[0123] S43. Use the geospatial information features in the training set as input features and the corresponding resident welfare assessment samples as labels to train the deep learning neural network and obtain a resident welfare assessment model.

[0124] In step S41 of this embodiment, the feature set for constructing the resident welfare assessment is expressed as:

[0125] H=S(V 1 ,V 2 ,V 3 ,V 4 ,V 5 ,V 6 ,V 7 ,V 8 ,V 9 )

[0126] Where S is the set feature, and the characteristic parameters are night light intensity (V1), vegetation index (V2), impervious surface index (V3), road density (V4), urban area ratio (V5), farmland area ratio (V6), elevation fluctuation (V7), housing concentration (V8), and population dispersion (V9).

[0127] In step S42 of this embodiment, approximately 5,000 sample network locations for resident welfare assessment are constructed, and geographic information feature parameters related to the economic welfare of residents in a 30m×30m grid are extracted, that is, each sample has 9 feature parameters, which serve as a training data set for the deep learning neural network.

[0128] Furthermore, in order to eliminate the dimensional effect of each indicator and unify the variation range of each indicator, all indicator information and reset costs are normalized and the value range is between 0 and 1.

[0129] In step S4 of the embodiment of the present invention, the resident welfare assessment model includes an input layer, a hidden layer, and an output layer that are connected in sequence.

[0130] Specifically, the number of neurons in the input layer corresponds to the number of geographic information features. According to the feature set constructed above, the model in this embodiment has 9 neurons in the input layer, with multiple hidden layers in the middle. Finally, the function f(x,w) of the predicted value of the resident welfare space output by the output layer is expressed as:

[0131]

[0132] In the formula, P represents the number of geographic information features, x oq represents the value of the p-th type of geographic information feature of the q-th sample, x pq represents the corresponding weight parameter, b q Indicates the offset.

[0133] In step S4 of the embodiment of the present invention, the deep learning neural network is trained using the mean square error loss function and the error back propagation algorithm so that the difference between the predicted value and the true value is as small as possible, and the parameters of the model are obtained; wherein the loss function is expressed as:

[0134]

[0135] In the formula, n is the number of samples, f(x j ,w) is the predicted value of the jth sample, y j is the true value of residents’ welfare from the jth sample household survey.

[0136] The model in this embodiment uses batch stochastic gradient descent (SGD) to optimize the loss function. The SGD function updates the weight parameters and offset parameters of the feature after each iteration to minimize the loss function value, where the updated weights and offsets are expressed as:

[0137]

[0138] Where η is the learning rate, is the gradient of the loss function in the direction of weight W, is the gradient of the loss function in the direction of offset B.

[0139] In this embodiment, the deep learning neural network is set with 4 hidden layers, and the learning rate η is set to 10-4. The training sample capacity will have a certain impact on the model accuracy. This scheme intends to select 80% of the samples (4000) as model training samples by random method, and the remaining 20% ​​(1000) as model verification samples.

[0140] After the above model training process, step S4 of this embodiment further includes:

[0141] The mean absolute error and accuracy are used as evaluation indicators, and the validation sample is used to verify the accuracy of the resident welfare assessment model. The mean absolute error (MAE) and accuracy (ACC) are expressed as:

[0142]

[0143] Among them, Y ′ is the resident welfare predicted by the model, Y is the actual resident welfare, and n is the number of samples, n=1000.

[0144] In the present invention, based on the above extracted geographic information features related to residents' economic well-being, a feature set {F i , i=1,2,…,9}, including night light intensity, vegetation index, impervious surface index, road density, urban area ratio, farmland area ratio, elevation fluctuation, house concentration, and population dispersion, as the input layer of the model. The trained and optimized model is used for spatial simulation calculation to obtain the spatial simulation result of residents' welfare. The result is a spatially continuous grid distribution with a spatial resolution of 30m×30m, which can show the heterogeneity of the spatial distribution of residents' welfare within the administrative boundary.

[0145] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0146] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for simulating the spatial well-being of residents by integrating remote sensing and perception survey, characterized in that: The following steps are involved: S1. Collect multi-source data of residents’ welfare spatial areas and pre-process them; S2, extract geographic information features related to residents' well-being from the preprocessed data and unify them to the same resolution grid; The geographic information features include night light intensity, vegetation index, impervious surface index, road density, urban area ratio, farmland area ratio, elevation fluctuation index, housing concentration index and population dispersion index; S3. Construct a multi-dimensional indicator system for comprehensive resident welfare, obtain its indicator data, and construct a sample of resident welfare spatial assessment based on the divided grids; S4. Construct a deep learning neural network, use the extracted geographic information features as input features, use the resident welfare spatial assessment samples in the corresponding area as labels, train the deep learning neural network, and obtain a resident welfare assessment model; S5. Use the geographic information characteristics of the area to be evaluated as the input of the resident welfare assessment model to predict the simulation results of the resident welfare space.

2. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 1 is characterized in that: In step S1, the multi-source data includes multi-source optical and radar data, night light data, located point of interest data, land use data and statistical yearbook data; The pre-processing comprises: Preprocessing of multi-source optical and radar data includes radiometric calibration, atmospheric correction, band registration, multi-band image synthesis and image fusion of optical remote sensing data, as well as orbit correction, image clipping, filtering, radiometric calibration, terrain correction, geocoding and image mosaicking of radar data; The preprocessing of night light data includes radiation calibration, atmospheric correction, geometric correction, and data clipping and splicing; The preprocessing of the located POI data includes duplicate data removal, data cleaning, coordinate system unification and data classification.

3. The method for simulating resident welfare space by integrating remote sensing and perception survey according to claim 1, characterized in that: In step S2, the night light intensity includes the total value of night light intensity, the mean value of night light intensity and the standard deviation of night light intensity; The vegetation index includes the normalized difference vegetation index and the enhanced vegetation index; The impervious surface index includes a normalized difference impervious surface index and a ratio-type building index.

4. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 3 is characterized in that: The total night light intensity TNL is: Among them, DN i represents the i-th gray level of all valid pixels in the night light data after preprocessing, C i represents the number of pixels corresponding to the gray level, and I represents the maximum gray level; The night light intensity mean NLM is: The nighttime light intensity standard deviation NSTD is: The normalized vegetation index NVDI is expressed as: NVDI=(NIR-R) / (NIR+R), wherein NIR represents the reflectivity of the near infrared band, and R represents the reflectivity of the red band; The enhanced vegetation index EVI is: EVI = 2.5 × (NIR-R) / (NIR + 6R-7.5B + ​​1), where B represents the reflectivity of the blue band; The normalized difference impervious surface index NDISI is: Among them, SWIR represents the reflectivity of the short-wave infrared band, NIR represents the reflectivity of the near-infrared band, Green represents the reflectivity of the green band, and MIR1 represents the reflectivity of the mid-infrared band 1; The ratio-based building index RBI is: Among them, NIR represents the reflectivity of the near-infrared band, and Red represents the reflectivity of the red band; The road density index RDI is: RDI = total road length / area of ​​study area; The urban area ratio is: urban area ratio = urban area / study area area; The farmland area ratio is: farmland area ratio = farmland area / study area area; The elevation fluctuation index σ is: H avg =∑H t / N, where H i represents the elevation value of the t-th pixel, T represents the total number of pixels, and H avg Indicates the average elevation value; The housing aggregation index HAI is: HAI = ∑(n i ) / G, where n i represents the number of houses within the radius r of each house, and G represents the total number of houses in the study area; The population dispersion index PDI is: P k =P ak / P total , Among them, P k represents the population proportion of each administrative unit, represents the average population proportion, P ak represents the population of the kth administrative unit, P total represents the total population of the study area, and n represents the number of administrative units in the study area.

5. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 1, characterized in that: The step S3 comprises the following steps: S31. Construct a multidimensional indicator system; The multidimensional indicator system includes primary indicators and secondary indicators under them, and the primary indicators include economic factor indicators, environmental factor indicators and welfare perception indicators; S32. According to the constructed multi-dimensional indicator system, quantitative indicator data of perception of different objects are collected through household surveys, and the data are divided into grids; S33, selecting a grid containing an indicator item exceeding a set threshold as a sample grid; S34. Based on the indicator data in the sample grid, quantitative statistics are conducted on the welfare satisfaction of the respondents in each assessment area, and the average welfare of the residents is calculated; S35. Calculate the average welfare of residents in all sample grids, and then construct a spatial assessment sample of resident welfare.

6. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 5 is characterized in that: In the step S32, the multidimensional indicator system forms a questionnaire based on scale analysis according to the determined secondary indicators, and collects quantitative indicator data perceived by different subjects through a questionnaire household survey; In the questionnaire, each indicator is provided with options corresponding to different weights, and all options are weighted and counted to form a perceptual quantitative survey result.

7. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 1, characterized in that: The step S4 comprises the following sub-steps: S41. Construct a feature set for evaluating residents’ well-being based on the extracted geographic information features; S42. According to the location of the grid where the resident welfare assessment sample is located, determine its relevant geographic information features, build a corresponding relationship between the sample and the geographic information features, and form a training data set; S43. Use the geospatial information features in the training set as input features and the corresponding resident welfare assessment samples as labels to train the deep learning neural network and obtain a resident welfare assessment model.

8. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 1, characterized in that: In step S4, the resident welfare assessment model includes an input layer, a hidden layer, and an output layer connected in sequence; The number of neurons in the input layer corresponds to the number of geographic information features, and the function f(x,w) of the predicted value of the resident welfare space output by the output layer is expressed as: In the formula, P represents the number of geographic information features, x pq represents the value of the p-th type of geographic information feature of the q-th sample, x pq represents the corresponding weight parameter, b q Indicates the offset.

9. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 7, characterized in that: In step S4, the deep learning neural network is trained using a mean square error loss function and an error back propagation algorithm, and the mean square error loss function is optimized using a batch stochastic gradient descent method.

10. The method for simulating the resident welfare space by integrating remote sensing and perception survey according to claim 1, characterized in that: The step S4 further comprises: The mean absolute error and precision were used as evaluation indicators, and the validation sample was used to verify the accuracy of the resident welfare assessment model.

Images