CN115713691A - Pixel-level electric power popularity estimation method and device based on noctilucent remote sensing - Google Patents
Pixel-level electric power popularity estimation method and device based on noctilucent remote sensing Download PDFInfo
- Publication number
- CN115713691A CN115713691A CN202211452144.1A CN202211452144A CN115713691A CN 115713691 A CN115713691 A CN 115713691A CN 202211452144 A CN202211452144 A CN 202211452144A CN 115713691 A CN115713691 A CN 115713691A
- Authority
- CN
- China
- Prior art keywords
- electric power
- popularity
- night light
- pixel
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 59
- 238000011160 research Methods 0.000 claims abstract description 40
- 238000013136 deep learning model Methods 0.000 claims abstract description 27
- 230000036541 health Effects 0.000 claims abstract description 20
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 230000002354 daily effect Effects 0.000 claims description 34
- 238000012545 processing Methods 0.000 claims description 23
- 238000012549 training Methods 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000002790 cross-validation Methods 0.000 claims description 9
- 238000012952 Resampling Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 230000003203 everyday effect Effects 0.000 claims description 4
- 238000010200 validation analysis Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000012417 linear regression Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- DFPOZTRSOAQFIK-UHFFFAOYSA-N S,S-dimethyl-beta-propiothetin Chemical compound C[S+](C)CCC([O-])=O DFPOZTRSOAQFIK-UHFFFAOYSA-N 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 150000001412 amines Chemical class 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000019800 disodium phosphate Nutrition 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000003331 infrared imaging Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
A pixel-level electric power popularity estimation method and device based on noctilucent remote sensing comprises the steps of obtaining an annual average night light image, a daily night light image, population data of corresponding years and population and health survey electric power popularity data of a research area in research time, preprocessing the images and the data, calculating a gray level histogram of the preprocessed daily night light image and converting the gray level histogram into gray level frequency information of the noctilucent image, extracting multi-dimensional features related to electric power popularity of the annual average night light image of the large-scale area on the space from the preprocessed annual average night light image, building a deep learning model based on the features, calculating electric power popularity of all pixels, and obtaining the whole electric power popularity of the research area by using the pixel-level electric power popularity. The method combines the characteristics related to the electric power popularity rate in the long-time sequence and the large-range spatial neighborhood in the noctilucent image, and the accuracy of the provided method for estimating the electric power popularity rate is high.
Description
Technical Field
The invention belongs to the technical field of noctilucent remote sensing socioeconomic estimation, and particularly relates to a pixel-level scale power popularity estimation method and device based on noctilucent remote sensing.
Background
The popularity of electricity is very important. At present, power is in poor popularity in many regions, such as sub-saharan africa, where over 70% of the sanitation facilities do not have a reliable power supply and one quarter of the sanitation facilities do not have a power supply at all. The method can be used for timely and accurately acquiring the electric power popularity in the areas, and has important significance for developing electric power assistance plans in a targeted manner, evaluating the success of the existing electrified projects, promoting social fairness and the like.
With the continuous development of satellite remote sensing technology, noctilucent remote sensing data with different space-time resolutions provides a reliable data source for power popularity estimation. A new generation night light sensor, namely a National Polar-orbital Satellite Visible Infrared Imaging Radiometer (NPP/VIIRS), has the capability of detecting weak night light, has higher radiation resolution compared with the data of a military Meteorological Satellite planned linear scanning service System (DMSP/OLS), and can more accurately estimate the electric power popularity. Compared with the traditional field investigation method, the method for estimating the electric power popularity based on the noctilucent remote sensing data has the advantages of low cost, quick updating, good consistency and the like, and meets the requirements of electric power data users at the present stage. In recent years, with the progress of a noctilucent remote sensing processing means and the continuous abundance of data sources, the estimation of the electric power popularity by using night light data becomes a practical and effective choice.
The current research of many scholars on the estimation of electric power popularity mainly comprises: min and Gaba take Vietnam as research area to evaluate the influence of street lamps and electrified families on radiance of noctilucent remote sensing image records [1] (ii) a Min utilizes the luminous image time sequence, calculates the instability of power supply of each pixel through the statistical relationship between the variance and the luminous value, evaluates the power fluctuation and interruption of India based on the index, and finally determines the unstable area of the power system [2] (ii) a Tingzon et al combines deep learning with satellite images to provide a highly profitable estimation method that can estimate social channels such as power supplyEconomic indicators [3] Data used for it includes geographic information data of Open Street Maps (OSM) and night light satellite images; dhorn and the like are combined with annual average noctilucence data, griffinder power grid data and OSM road network data, and the electric power popularity of multiple countries is estimated through a random forest model [4] 。
By integrating with previous researches, the existing pixel-level scale electric power popularity estimation method based on noctilucent remote sensing is not mature enough, the processing of the long-time sequence of the noctilucent remote sensing images is simpler, and the long-time sequence is generally converted into indexes with single dimensionality such as mean value, standard deviation and the like and low relevance to the electric power popularity. In addition, the conventional estimation method only considers the luminous data in a small range (such as in the neighborhood of 7 × 7 pixels) near the position to be estimated, so that the characteristics contained in a nearby large-range image related to the electric power popularity are ignored.
The relevant documents are as follows:
[1]Min B,Gaba K M.Tracking electrification in Vietnam using nighttime lights[J].Remote Sensing,2014,6(10):9511-9529.
[2]Min B K,O'Keeffe Z,Zhang F.Whose power gets cutUsing high-frequency satellite images to measure power supply irregularity[J].Using High-Frequency Satellite Images to Measure Power Supply Irregularity(June 29,2017).World Bank Policy Research Working Paper,2017(8131).
[3]Tingzon I,Orden A,Sy S,et al.Mapping poverty in the Philippines using machine learning,satellite imagery,and crowd-sourced geospatial information[C]//AI for Social Good ICML 2019Workshop.2019.
[4]Dhorne M,Nicolas C,Arderne C,et al.Tracking Advances in Access to Electricity Using Satellite-Based Data and Machine Learning to Complement Surveys[J].2021.
disclosure of Invention
According to the defects of the prior art, the invention aims to provide a pixel-level power popularity estimation method and device based on noctilucent remote sensing.
In order to solve the technical problems, the invention adopts the technical scheme that:
a pixel-level power popularity estimation method based on noctilucent remote sensing comprises the following steps:
step 1, acquiring an average night light image of a research area in the research time, a night light image every day, population data of a corresponding year and population and health survey electric power popularity rate data;
step 2, preprocessing the annual average night light image and the daily night light image of the research area, wherein the preprocessing comprises geometric correction and resampling, and for the position of each population and the health survey electric power popularity rate data sample, extracting the annual average night light image and the population distribution map in the neighborhood of the position according to population data of corresponding years;
step 3, after obtaining radiance time sequence information with the dimensionality equal to the number of effective days of noctilucent data for the daily night light images preprocessed in the step 2, calculating a gray level histogram of the images, extracting gray level frequency information of the noctilucent images based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
step 4, extracting multi-dimensional characteristics related to the electric power popularity rate contained in the annual average night light image of the large-range area in space from the annual average night light image of the research area preprocessed in the step 2;
step 5, establishing a deep learning model by using the average night light image preprocessed in the step 2, deep characteristic information extracted from gray frequency information in the step 3, the multidimensional characteristic obtained in the step 4 as input data, and the electric power popularity of each position as an output value, and training;
and 6, calculating the electric power popularity rates of all pixel levels by using the deep learning model trained in the step 5, and calculating the whole electric power popularity rate of the research area by using the electric power popularity rates of the pixel levels.
Further, step 2 specifically includes:
step 2.1, realizing geographic coordinate registration among population distribution maps, annual average night light images and daily night light image data by selecting uniformly distributed control points, and unifying spatial resolution;
2.2, extracting an average night light image of the year in a 225 x 225 pixel neighborhood, a night light image of each day in a 3 x 3 pixel neighborhood of all the days of the year and a population distribution map of the pixel according to the position and the year of the population and health survey electric power popularity data;
in the step 3, the method for extracting the gray frequency information includes:
calculating the frequency of each gray level interval of radiance in the daily night light image in one year, converting radiance information into frequency information of each radiance interval of n dimensionalities, and for a certain pixel, according to the daily night light image of the current year, the gray level frequency T belonging to the ith dimensionality in one year i The calculation formula of (a) is as follows:
T i ＝NTL i /NTL V
wherein, NTL V Is the total number of days in the year that the mandatory quality indicator identifies as data valid, NTL i The number of days that the mandatory quality mark in the year is marked as effective data and the value of the data B0 wave band belongs to the ith interval, i is more than or equal to 1 and less than or equal to n;
will T 1 ,T 2 ,T 3 ,...,T n Connecting to form a new characteristic T, wherein the T represents frequency information of each radiance interval at the pixel, taking the characteristic T at each pixel in a 3 x 3 pixel neighborhood of the pixel where population and health survey electric power popularity rate data samples are located as gray frequency information, and using the information as the input of a subsequent deep learning model;
the step 4 specifically includes:
step 4.1, processing the annual average noctilucence data of 225 x 225 pixels near the estimated position through a pre-trained VGG-16 model, so that the estimated model provided by the invention fully considers the information contained in the noctilucence data in a large-scale space and outputs 1000-dimensional space-related characteristics;
step 4.2, feature selection, namely calculating correlation coefficients between 1000-dimensional features and the electric power popularity rate respectively, reserving features with the highest correlation between 10-dimensional features and the electric power popularity rate from the features, and using the reserved features to estimate the electric power popularity rate;
the step 5 specifically comprises:
step 5.1, extracting deep feature information related to the electric power popularity rate from the gray frequency information provided in the step 3 through the convolution layer I, setting the template size of the convolution layer I to be 3 x 3, setting the dimension of an output waveband to be 5, and reducing the output feature dimension to be 5 through the convolution layer 1 and the active layer;
step 5.2, connecting the output of the step 5.1 with the multidimensional characteristics provided by the step 4 in the wave band dimension, generating 15-dimensional characteristics for each position, constructing a linear regression model based on the 15-dimensional characteristics to fit the electric power popularity rate of the pixel, constructing the regression model mainly through a convolution layer II, setting the template size of the convolution layer II to be 1 x 1, and setting the output wave band dimension of the convolution layer II to be 1;
step 5.3, carrying out logarithm processing on the average annual average luminous value near the estimated position, establishing a full connection layer, modifying the estimation result of the previous layer of neural network based on the average annual average luminous value in the region, and finally estimating the electric power popularity at the corresponding position;
step 5.4, determining hyper-parameters of the model through five-fold cross validation, randomly and uniformly dividing the sample into 5 parts, training 4 parts in the 5 parts each time, and reserving the remaining 1 part for validation, wherein the determined hyper-parameters are as follows, the batch size is set to be 10, the learning rate is set to be 0.003, the weight attenuation is set to be 1E-9, an Adam optimizer is selected for optimization, and an L1 norm loss function is selected as a loss function;
in the step 6, the hyper-parameters determined in the step 5 are used, and the average absolute error E of the estimation method on the training set and the verification set is calculated according to the five-fold cross verification Absolute Is calculated as the formula:
wherein EA 1 Is the electric power prevalence given by the statistical data, EA 2 The power popularity obtained by the invention, the formula can also be used for measuring the L1 norm loss in the deep learning process, and n is the number of samples.
A pixel level electric power popularity estimation device based on noctilucence remote sensing comprises:
the data acquisition module is used for acquiring annual average night light images, daily night light images, population data of corresponding years and population and health survey electric power popularity rate data of a research area in research time;
the system comprises a preprocessing module, a data acquisition module, a data processing module and a data processing module, wherein the preprocessing module is used for preprocessing the annual average night light image and the daily night light image of a research area, the preprocessing comprises geometric correction and resampling, and for the position of each population and the position of a health survey electric power popularity rate data sample, the annual average night light image and a population distribution map in the neighborhood of the position are extracted according to population data of corresponding years;
the daily night light image processing module is used for calculating a gray level histogram of the preprocessed daily night light image after obtaining radiance time sequence information with the dimensionality equal to the number of effective days of noctilucent data, extracting gray level frequency information of the noctilucent image based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
the annual average night light image processing module is used for extracting the multidimensional characteristics related to the electric power popularity rate, which are contained in the annual average night light image of the large-range area in space, from the preprocessed annual average night light image of the research area;
the deep learning module is used for establishing a deep learning model by taking the preprocessed average night light images, deep characteristic information of gray frequency information and multidimensional characteristics as input data and electric power popularity of each position as an output value for training;
and the pixel-level electric power popularity rate acquisition module is used for calculating the electric power popularity rates of all pixel levels by using the trained deep learning model and calculating the whole electric power popularity rate of the research area by using the pixel-level electric power popularity rates.
A pixel-level electric power popularity estimation device based on noctilucent remote sensing comprises a processor and a memory, wherein the memory is used for storing a computer program capable of being run on the processor, and when the processor is used for running the computer program, any one of the steps of the pixel-level electric power popularity estimation method based on noctilucent remote sensing is executed.
A computer storage medium, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of any one of the pixel-level power popularity estimation method based on noctilucent remote sensing are realized.
Compared with the prior art, the invention has the following advantages and beneficial effects:
1. according to the pixel-level scale electric power popularity estimation method and device based on the noctilucent remote sensing, the multi-dimensional characteristics related to the electric power popularity are extracted from the noctilucent remote sensing images with a long time sequence and a large space range near the estimated position respectively, the multi-dimensional characteristics are used for estimating the electric power popularity, fluctuation information contained in the time sequence of the night light images every day and the reflection of the electric power popularity at the position to be estimated by annual average night light in a large space range are considered, and therefore the electric power popularity estimation result of the pixel-level scale is more reasonable and accurate.
2. According to the pixel-level scale electric power popularity estimation method and device based on the noctilucent remote sensing, the deep learning model is used for replacing a specific function model, estimation deviation caused by improper function model selection is avoided, and data processing and calculation are simpler.
3. According to the pixel-level scale electric power popularity estimation method and device based on the noctilucent remote sensing, the annual average noctilucent image and the daily noctilucent image time sequence are used for estimating the electric power popularity, the modeling is completely based on noctilucent data, multi-source data such as an OSM road network and the like used by scholars and having low relevance to the electric power popularity are abandoned, the established method can effectively extract the space-time characteristics related to the electric power popularity, and the potential of noctilucent remote sensing data in the electric power popularity estimation aspect is fully exerted.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention, are incorporated in and constitute a part of this application. The exemplary embodiments and descriptions of the present invention are provided to explain the present invention and not to limit the present invention. In the drawings:
FIG. 1 is a flow chart of the pixel-level scale power popularity estimation method based on noctilucent remote sensing.
FIG. 2 is a schematic diagram of a training process of the pixel-level scale power popularity estimation method based on noctilucent remote sensing;
FIG. 3 (a) is a diagram illustrating the variation of the 1 st-fold error with the period in the five-fold cross validation in the embodiment of the present invention;
FIG. 3 (b) is a diagram illustrating the variation of the error of the 2 nd fold with the period in the five-fold cross validation in the embodiment of the present invention;
FIG. 3 (c) is a diagram illustrating the variation of the 3 rd fold error with the period in the five-fold cross validation in the embodiment of the present invention;
FIG. 3 (d) is a diagram illustrating the variation of the 4 th fold error with the period in the five-fold cross validation in the embodiment of the present invention;
fig. 3 (e) shows the variation of the 5 th fold error with the period in the five-fold cross validation in the embodiment of the present invention.
Detailed Description
The technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
The embodiment of the application provides a pixel-level scale electric power popularity estimation method based on noctilucent remote sensing, and solves the problems that in the prior art, the pixel-level scale electric power popularity estimation method based on noctilucent remote sensing is not mature enough, the long-time sequence of noctilucent remote sensing images is processed simply, and the annual noctilucent images in a large range in space related to the electric power popularity and the characteristics of the long-time sequence daily noctilucent images are not fully utilized.
In order to solve the above problems, as shown in fig. 1 and 2, the technical solution in the embodiment of the present application provides a pixel-level scale power popularity estimation method based on noctilucent remote sensing, which includes the following steps:
step 1, acquiring an average night light image of a research area in the research time, a night light image every day, population data of a corresponding year and population and health survey electric power popularity rate data;
step 2, preprocessing the annual average night light image and the daily night light image of the research area, wherein the preprocessing comprises geometric correction and resampling, and for the position of each population and the health survey electric power popularity rate data sample, extracting the annual average night light image and the population distribution map in the neighborhood of the position according to population data of corresponding years;
step 3, after obtaining radiance time sequence information with the dimensionality equal to the number of effective days of noctilucent data for the daily night light images preprocessed in the step 2, calculating a gray level histogram of the images, extracting gray level frequency information of the noctilucent images based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
step 4, extracting multi-dimensional characteristics related to the electric power popularity contained in the annual average night light image of the large-range area in space from the annual average night light image of the research area preprocessed in the step 2;
step 5, establishing a deep learning model by using the average night light image preprocessed in the step 2, deep characteristic information extracted from the gray frequency information obtained in the step 3, the multidimensional characteristic obtained in the step 4 as input data, and the electric power popularity rate of each position as an output value, and training;
and 6, calculating the electric power popularity rate of the pixel-level scales at all positions by using the deep learning model trained in the step 5, and acquiring the whole electric power popularity rate of the research area by combining the electric power popularity rate of the pixel level and population distribution data.
The invention provides a pixel-level scale electric power popularity estimation method based on noctilucent remote sensing, which comprises the steps of extracting radiance time sequence information with the dimension equal to the number of effective days of noctilucent data from daily night light images in a research area in research time, calculating a gray level histogram of the radiance time sequence information, calculating gray level frequency information of the noctilucent image in the current year based on the gray level histogram, and obtaining fluctuation information of the noctilucent image time sequence; and determining estimated positions according to population and health survey data, extracting multidimensional characteristics related to the electric power popularity from the noctilucent remote sensing images in a large space range near the estimated positions, and using the characteristics for estimating the electric power popularity, wherein the characteristics can reflect the influence of the average night light in a large-range area on the space on the electric power popularity, so that the electric power popularity estimation result in a pixel-level scale is more reasonable and accurate.
According to the pixel-level scale electric power popularity estimation method based on the noctilucent remote sensing, the deep learning model is established to replace a specific function model, deviation of a calculation result caused by improper function model selection is avoided, and data processing and calculation are simpler.
According to the pixel-level scale electric power popularity estimation method based on the noctilucent remote sensing, the annual average noctilucent image and the daily noctilucent image time sequence are used for estimating the electric power popularity, modeling is completely based on noctilucent data, the established method can effectively extract space-time characteristics related to the electric power popularity, and the potential of noctilucent remote sensing data in the aspect of electric power popularity estimation is fully exerted.
In the step 1, the average night light image per year is Annual VNL V2/average-masked, and the night light image per day is VNP46A2.
The step 2 specifically comprises the following steps:
and 2.1, realizing geographic coordinate registration among population distribution maps, annual average night light images and daily night light image data by selecting uniformly distributed control points, and unifying spatial resolution.
And 2.2, extracting an average night light image of the year in a 225 x 225 pixel neighborhood, night light images of each day in a 3 x 3 pixel neighborhood of all days of the year and a population distribution map of each pixel according to the position and the year of each population and Health survey electric power popularity Data (DHS).
In step 2.1, the uniform spatial resolution is 500m.
The frequency information obtained in the step 3 is helpful for improving the precision of the existing power popularity rate estimation method. After introducing frequency information, the existing estimation method can compress the related information of the time series and reduce the complexity of an estimation model.
In the step 3, the frequencies of all gray scale intervals of the radiance of the daily night light image in one year are calculated, and the frequencies are used for replacing the original irregular night light data time sequence, so that the problem of compensation in processing the time sequence can be avoided, and the influence caused by different total effective days in one year of the night light data in different areas can be ignored. In this process, the most important step is to define the interval of the histogram.
In the step 3, the method for extracting the deep feature information of the grayscale frequency information includes:
calculating the frequency of each gray level interval of radiance in the daily night light image in one year, converting radiance information into frequency information of each radiance interval of n dimensionalities, and for a certain pixel, according to the daily night light image of the current year, the gray level frequency T belonging to the ith dimensionality in one year i The calculation formula of (a) is as follows:
T i ＝NTL i /NTL V
wherein, NTL V Is the total number of days in the year that the mandatory quality indicator identifies as data valid, NTL i The number of days that the mandatory quality mark in the year is marked as valid data and the value of the B0 wave band of the data belongs to the ith interval is more than or equal to 1 and less than or equal to n;
will T 1 ,T 2 ,T 3 ,...,T n The T represents the frequency information of each radiance interval at the pixel, and the characteristic T at each pixel in the 3 x 3 pixel neighborhood of the pixel where the population and health survey electric power popularity rate data sample is located is used as the deep characteristic credit of the gray frequency informationAnd outputting the information.
And averaging the frequency information T in the 3 x 3 pixel neighborhood of the pixel where the DHS sample is located to obtain T', calculating a correlation coefficient between the frequency information T and the electric power popularity rate of the sample, and extracting a partial feature through dimension reduction of a histogram and the correlation coefficient between the partial feature and the electric power popularity rate.
The calculation formula of the correlation coefficient X between T' and the electric power popularity rate of the sample is as follows:
in step 4, the Annual VNL V2/average-masked average night light images obtained in step 2 are subjected to extraction of the multidimensional characteristics related to the electric power popularity of the large-scale area night light images. The multi-dimensional features of the luminous images in the large-range area are helpful for estimating the electric power popularity, and the introduction of the features can further enrich the features extracted by the model on the basis of acquiring the frequency information in the step 3, so that the accuracy of the existing electric power popularity estimation method is improved.
The step 4 specifically comprises the following steps:
and 4.1, taking a pre-training model trained on a similar task as a part of the electric power popularity estimation model through transfer learning so as to extract the characteristics related to the electric power popularity in the luminous remote sensing image in a large range near the position to be estimated.
And 4.2, selecting characteristics, namely calculating correlation coefficients between the 1000-dimensional characteristics output by the 4.1 and the electric power popularity rate respectively, and reserving the characteristics with the highest correlation between the 10-dimensional characteristics and the electric power popularity rate from the characteristics for estimating the electric power popularity rate.
In step 4.1, a VGG-16 model pre-trained on an ImageNet image classification data set is selected, annual average noctilucent data of 225 × 225 pixels near an estimated position is used as input, information contained in noctilucent data in a large-range space is fully considered, and therefore power popularity related information outside a small-range neighborhood can be extracted, and 1000-dimensional space related features are finally output.
And calculating a correlation coefficient between the 10-dimensional feature Z output in the step and the electric power popularity rate X of the sample.
And (3) training a deep learning model for estimating the electric power popularity by taking the average night light image obtained in the step 2, the frequency information of each radiance interval in the small neighborhood near the corresponding DHS sample obtained in the step 3 and the 10-dimensional space-related characteristics obtained in the step 4 as input data and taking the electric power popularity at the corresponding position as an output value.
The features extracted in step 3 and step 4 have high correlation with the electric power popularity, and in order to fully utilize the features to estimate the electric power popularity, a certain model is required to be used for combining the features. The deep learning model is a model which can be effectively combined with multidimensional characteristics and is used for regression target information (the regression target in the invention is the electric power popularity rate), and appropriate model parameters can be determined through training, so that the deep learning model is combined with the steps 3 and 4 to extract the multidimensional characteristics.
The step 5 specifically comprises the following steps:
step 5.1, extracting features related to the electric power popularity rate from the deep feature information provided in the step 3 through the convolutional layer I, setting the size of a convolutional layer I template to be 3 multiplied by 3, setting the output waveband dimension to be 5, and reducing the width of the output feature dimension to be 5 through the convolutional layer 1 and the active layer;
step 5.2, connecting the output of the step 5.1 with the multidimensional characteristics provided by the step 4 in the wave band dimension, generating 15-dimensional characteristics for each position, combining the 15-dimensional characteristics to construct a linear regression model to fit the electric power popularity rate of the pixel, constructing the regression model mainly through a convolution layer II, wherein the size of a template of the convolution layer II is 1 multiplied by 1, and the output wave band dimension of the convolution layer II is 1;
step 5.3, carrying out logarithmic processing on the average annual average luminous value near the estimated position, establishing a full connection layer, modifying the estimation result of the previous layer of neural network based on the average annual average luminous value in the region, and finally estimating the electric power popularity at the corresponding position;
the input characteristics of the step comprise the average annual average luminous value processed logarithmically and the characteristics output in the step 5.2, the processed structure is a full connection layer and an activation layer, and the output result is the electric power popularity rate at the pixel.
Step 5.4, as shown in fig. 3 (a) -3 (e), determining the hyper-parameters of the model through five-fold cross validation, randomly and uniformly dividing the sample into 5 parts, using 4 parts of the samples for training each time, and reserving the remaining 1 part for validation, wherein in one embodiment of the invention, the determined hyper-parameters are as follows: the batch size was set to 10, the learning rate was set to 0.003, the weight decay was set to 1E-9, and in addition, an Adam optimizer was selected for optimization, with the L1 norm loss function as the loss function.
In step 6, the pixel level power popularity at all positions is estimated by using the deep learning model trained in step 5. Using the hyperparameters determined in step 5, the mean absolute error over the training and validation sets was calculated according to a five-fold cross-validation estimation method, and then the model was trained on the complete training set containing 6267 valid samples.
Mean absolute error E Absolute value The calculation formula of (2) is as follows:
wherein EA 1 Is the electric power prevalence given by the statistical data, EA 2 The power popularity obtained by the invention, the formula can also be used for measuring the L1 norm loss in the deep learning process, and n is the number of samples.
In specific implementation, the pixel-level power popularity estimation method based on noctilucent remote sensing can adopt software technology to realize automatic operation. According to the embodiment provided by the invention, south Africa is taken as a research area, and the change situation of the electric power popularity rate between 2013 and 2018 is detected. The annual average night light image covering the south Africa region in 2013-2018 and the daily night light image covering the south Africa region in 2013-2018 are collected, and meanwhile, the population distribution map data and the DHS power prevalence rate data in 2013-2018 are collected.
The data acquisition can be carried out in advance when the method is implemented.
Wherein the Annual average night light image is Annual VNL V2/average-masked, and the data sources of the Annual average night light image covering south Africa areas in 2013-2018 are as follows: https:// eogdata. Amines. Edu/products/vnl/, the daily night light image is VNP46A2, and the data of the daily night light image covering the south Africa region from 2013-2018 is from https:// ladssweb. Models. Eosdis. Nasa. Gov/.
According to the experience of estimating the power prevalence rate by using the luminance threshold method in the past, the radiance is low in most regions in south africa, and many regions in which it is difficult to determine whether power is supplied belong to this category. Therefore, in the step, when the histogram is constructed, the region with the radiance less than 2nW/cm/sr is focused. Furthermore, even in urban centers in south Africa, it is difficult for outdoor lighting to produce radiances in excess of 70nW/cm/sr, so step 4 merges sections with brightnesses in excess of 73.2 nW/cm/sr.
The gray scale interval of the step 4 is divided into 3 gray scale intervals which are 0-2nW/cm/sr, 2.4-5.2nW/cm/sr and 9.2-73.2nW/cm/sr respectively, in the three gray scale intervals, the interval between 0-2nW/cm/sr is 0.1nW/cm/sr, the interval between 2.4-5.2nW/cm/sr is determined as 0.4nW/cm/sr, and the interval between 9.2-73.2nW/cm/sr is determined as 4nW/cm/sr.
According to the interval, for a certain pixel, according to the day luminous image of the current year, calculating the gray frequency T belonging to the ith interval in one year i The calculation formula is as follows:
T i ＝NTL i /NTL V
wherein NTL V Is the total number of days, NTL, that the mandatory quality indicator for VNP46A2 data in the year identifies as data valid i Is the number of days in the year in which the mandatory quality flag for VNP46A2 data identifies that the data is valid and the value of the B0 band for that data belongs to the i-th interval.
Will T 1 ,T 2 ,T 3 ,...,T 47 Connected to form a new feature T representing the picture elementAnd (3) taking the frequency information of each radiance interval as the output of the step, wherein the characteristic T of each pixel in the 3 x 3 pixel neighborhood of the pixel where the DHS sample is located.
The frequency information T in the 3 × 3 neighborhood of the pixel where the DHS sample is located is averaged, and the correlation coefficient between the frequency information T and the power prevalence rate of the sample is calculated, as shown in table 3.1.
TABLE 3.1 correlation coefficient between partial features extracted by histogram dimensionality reduction and Power prevalence
The correlation coefficient between the 10-dimensional feature output in step 4 and the power prevalence rate of the sample is shown in table 4.1.
Table 4.1 correlation coefficient between partial features extracted at step 4.2 and electric power prevalence
The average absolute error of the established power popularity rate estimation method on a training set is as low as 0.130, and the average absolute error on a verification set is only 0.145.
The invention also provides a pixel-level electric power popularity estimation device based on noctilucent remote sensing, which comprises:
the data acquisition module is used for acquiring annual average night light images, daily night light images, population data of corresponding years and population and health survey electric power popularity rate data of a research area in research time;
the system comprises a preprocessing module, a data acquisition module, a data processing module and a data processing module, wherein the preprocessing module is used for preprocessing the annual average night light image and the daily night light image of a research area, the preprocessing comprises geometric correction and resampling, and for the position of each population and the position of a health survey electric power popularity rate data sample, the annual average night light image and a population distribution map in the neighborhood of the position are extracted according to population data of corresponding years;
the daily night light image processing module is used for calculating a gray level histogram of the preprocessed daily night light image after obtaining radiance time sequence information with the dimensionality equal to the number of effective days of noctilucent data, extracting gray level frequency information of the noctilucent image based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
the annual average night light image processing module is used for extracting multidimensional characteristics related to electric power popularity contained in annual average night light images of large-scale areas in space for the preprocessed annual average night light images of the research areas;
the deep learning module is used for establishing a deep learning model by taking the preprocessed average night light images, deep characteristic information of gray frequency information and multidimensional characteristics as input data and electric power popularity of each position as an output value for training;
and the pixel-level electric power popularity rate acquisition module is used for calculating the electric power popularity rates of all pixel levels by using the trained deep learning model and calculating the whole electric power popularity rate of the research area by using the pixel-level electric power popularity rates.
The invention also provides pixel-level power popularity estimation equipment based on noctilucent remote sensing, which comprises a processor and a memory, wherein the memory is used for storing a computer program capable of running on the processor, and the processor is used for executing any one of the steps of the pixel-level power popularity estimation method based on noctilucent remote sensing when running the computer program.
The memory in the embodiment of the invention is used for storing various types of data to support the operation of the pixel-level power popularity estimation equipment based on noctilucent remote sensing.
The pixel-level power popularity estimation method based on the noctilucent remote sensing disclosed by the embodiment of the invention can be applied to a processor or realized by the processor. The processor may be an integrated circuit chip having signal processing capabilities. In the implementation process, each step of the pixel-level power popularity rate estimation method based on noctilucent remote sensing can be completed through an integrated logic circuit of hardware in a processor or an instruction in a software form. The Processor may be a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. The processor may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present invention. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed by the embodiment of the invention can be directly implemented by a hardware decoding processor, or can be implemented by combining hardware and software modules in the decoding processor. The software module can be located in a storage medium, the storage medium is located in a memory, a processor reads information in the memory, and the steps of the pixel-level power popularity estimation method based on the noctilucent remote sensing provided by the embodiment of the invention are completed by combining hardware of the software module.
In an exemplary embodiment, the pixel level power popularity estimation Device based on night light remote sensing may be implemented by one or more Application Specific Integrated Circuits (ASICs), DSPs, programmable Logic Devices (PLDs), complex Programmable Logic Devices (CPLDs), FPGAs, general purpose processors, controllers, micro Controllers (MCUs), microprocessors (microprocessors), or other electronic components for performing the aforementioned methods.
It will be appreciated that the memory can be either volatile memory or nonvolatile memory, and can include both volatile and nonvolatile memory. Among them, the nonvolatile Memory may be a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a magnetic random access Memory (FRAM), a magnetic random access Memory (Flash Memory), a magnetic surface Memory, an optical Disc, or a Compact Disc Read-Only Memory (CD-ROM); the magnetic surface storage may be disk storage or tape storage. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of illustration and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), synchronous Static Random Access Memory (SSRAM), dynamic Random Access Memory (DRAM), synchronous Dynamic Random Access Memory (SDRAM), double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), enhanced Synchronous Dynamic Random Access Memory (ESDRAM), enhanced Synchronous Dynamic Random Access Memory (Enhanced Synchronous Dynamic Random Access Memory), synchronous linked Dynamic Random Access Memory (DRAM, synchronous Link Dynamic Random Access Memory), direct Memory (DRmb Random Access Memory). The described memory for embodiments of the present invention is intended to comprise, without being limited to, these and any other suitable types of memory.
The invention also provides a computer storage medium, wherein a computer program is stored in the computer storage medium, and when the computer program is executed by a processor, the steps of any one of the pixel-level power popularity estimation method based on the noctilucent remote sensing are realized.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.
Claims (5)
1. A pixel-level power popularity estimation method based on noctilucent remote sensing is characterized by comprising the following steps:
step 1, acquiring an average night light image of a research area in the research time, a night light image every day, population data of a corresponding year and population and health survey electric power popularity rate data;
step 2, preprocessing the annual average night light image and the daily night light image of the research area, wherein the preprocessing comprises geometric correction and resampling, and for the position of each population and the health survey electric power popularity rate data sample, extracting the annual average night light image and the population distribution map in the neighborhood of the position according to population data of corresponding years;
step 3, after obtaining radiance time sequence information with dimensionality equal to the number of effective days of noctilucent data for the night light images preprocessed in the step 2, calculating a gray level histogram of the night light images, extracting gray level frequency information of the noctilucent images based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
step 4, extracting multi-dimensional characteristics related to the electric power popularity contained in the annual average night light image of the large-range area in space from the annual average night light image of the research area preprocessed in the step 2;
step 5, establishing a deep learning model by using the average night light image preprocessed in the step 2, deep characteristic information extracted from gray frequency information in the step 3, the multidimensional characteristic obtained in the step 4 as input data, and the electric power popularity of each position as an output value, and training;
and 6, calculating the electric power popularity rates of all pixel levels by using the deep learning model trained in the step 5, and calculating the whole electric power popularity rate of the research area by using the electric power popularity rates of the pixel levels.
2. The pixel-level scale power prevalence estimation method based on noctilucent remote sensing according to claim 1, wherein step 2 specifically comprises:
step 2.1, realizing geographic coordinate registration among population distribution maps, annual average night light images and daily night light image data by selecting uniformly distributed control points, and unifying spatial resolution;
2.2, extracting an average night light image of the year in a 225 x 225 pixel neighborhood, night light images of each year in a 3 x 3 pixel neighborhood of all days and population distribution maps of the pixel according to the position and the year of the population and health survey electric power popularity data in the pixel neighborhood;
in the step 3, the method for extracting the gray frequency information includes:
calculating the frequency of each gray level interval of radiance in the daily night light image in one year, converting radiance information into frequency information of each radiance interval of n dimensionalities, and for a certain pixel, according to the daily night light image of the current year, the gray level frequency T belonging to the ith dimensionality in one year i The calculation formula of (a) is as follows:
T i ＝NTL i /NTL V
wherein, NTL V Is the total number of days in the year that the mandatory quality indicator identifies as data valid, NTL i The number of days that the mandatory quality mark in the year is marked as effective data and the value of the data B0 wave band belongs to the ith interval, i is more than or equal to 1 and less than or equal to n;
will T 1 ,T 2 ,T 3 ,...,T n Connecting to form a new characteristic T, wherein the T represents frequency information of each radiance interval at the pixel, taking the characteristic T at each pixel in a 3 x 3 pixel neighborhood of the pixel where population and health survey electric power popularity rate data samples are located as gray frequency information, and using the information as the input of a subsequent deep learning model;
the step 4 specifically includes:
step 4.1, processing the annual average noctilucence data of 225 x 225 pixels near the estimated position through a pre-trained VGG-16 model, so that the estimated model provided by the invention fully considers the information contained in the noctilucence data in a large-scale space and outputs 1000-dimensional space-related characteristics;
step 4.2, feature selection, namely calculating correlation coefficients between 1000-dimensional features and the electric power popularity rate respectively, reserving features with the highest correlation between 10-dimensional features and the electric power popularity rate from the features, and using the reserved features to estimate the electric power popularity rate;
the step 5 specifically includes:
step 5.1, extracting deep feature information related to the electric power popularity rate from the gray frequency information provided in the step 3 through the convolution layer I, setting the template size of the convolution layer I to be 3 x 3, setting the dimension of an output wave band to be 5, and reducing the output feature dimension to be 5 through the convolution layer 1 and the active layer;
step 5.2, connecting the output of the step 5.1 with the multidimensional characteristics provided by the step 4 in the wave band dimension, generating 15-dimensional characteristics for each position, constructing a linear regression model based on the 15-dimensional characteristics to fit the electric power popularity rate of the pixel, constructing the regression model mainly through a convolution layer II, setting the template size of the convolution layer II to be 1 x 1, and setting the output wave band dimension of the convolution layer II to be 1;
step 5.3, carrying out logarithm processing on the average annual average luminous value near the estimated position, establishing a full connection layer, modifying the estimation result of the previous layer of neural network based on the average annual average luminous value in the region, and finally estimating the electric power popularity at the corresponding position;
step 5.4, determining hyper-parameters of the model through five-fold cross validation, randomly and uniformly dividing the sample into 5 parts, training 4 parts in the 5 parts each time, and reserving the remaining 1 part for validation, wherein the determined hyper-parameters are as follows, the batch size is set to be 10, the learning rate is set to be 0.003, the weight attenuation is set to be 1E-9, an Adam optimizer is selected for optimization, and an L1 norm loss function is selected as a loss function;
in the step 6, the hyper-parameters determined in the step 5 are used, and the average absolute error E of the estimation method on the training set and the verification set is calculated according to the five-fold cross verification Absolute Is calculated as the formula:
wherein EA 1 Is the electric power prevalence given by the statistical data, EA 2 The electric power popularity rate obtained by the invention, the formula can also be used for measuring the deep learningThe L1 norm loss of the equation, n is the number of samples.
3. A pixel level electric power popularity estimation device based on noctilucent remote sensing is characterized by comprising:
the data acquisition module is used for acquiring annual average night light images, daily night light images, population data of corresponding years and population and health survey electric power popularity rate data of a research area in research time;
the system comprises a preprocessing module, a data acquisition module, a data processing module and a data processing module, wherein the preprocessing module is used for preprocessing the annual average night light image and the daily night light image of a research area, the preprocessing comprises geometric correction and resampling, and for the position of each population and the position of a health survey electric power popularity rate data sample, the annual average night light image and a population distribution map in the neighborhood of the position are extracted according to population data of corresponding years;
the daily night light image processing module is used for calculating a gray level histogram of the preprocessed daily night light image after obtaining radiance time sequence information with the dimensionality equal to the number of effective days of noctilucent data, extracting gray level frequency information of the noctilucent image based on the gray level histogram, and subsequently extracting deep layer feature information from the gray level frequency information based on a deep learning model;
the annual average night light image processing module is used for extracting multidimensional characteristics related to electric power popularity contained in annual average night light images of large-scale areas in space for the preprocessed annual average night light images of the research areas;
the deep learning module is used for establishing a deep learning model by taking the preprocessed average night light images, deep characteristic information of gray frequency information and multidimensional characteristics as input data and electric power popularity of each position as an output value for training;
and the pixel-level electric power popularity rate acquisition module is used for calculating the electric power popularity rates of all pixel levels by using the trained deep learning model and calculating the whole electric power popularity rate of the research area by using the pixel-level electric power popularity rates.
4. A pixel level electric power popularity estimation equipment based on noctilucence remote sensing is characterized in that: comprising a processor and a memory for storing a computer program capable of running on the processor, the processor being configured to perform the steps of the pixel-level power prevalence estimation method based on noctilucent remote sensing according to any of claims 1-2 when running the computer program.
5. A computer storage medium, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the pixel-level power prevalence estimation method based on noctilucent remote sensing according to any one of claims 1-2 are implemented.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211452144.1A CN115713691B (en) | 2022-11-21 | 2022-11-21 | Noctilucent remote sensing-based pixel-level power popularity rate estimation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211452144.1A CN115713691B (en) | 2022-11-21 | 2022-11-21 | Noctilucent remote sensing-based pixel-level power popularity rate estimation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115713691A true CN115713691A (en) | 2023-02-24 |
| CN115713691B CN115713691B (en) | 2024-01-30 |
Family
ID=85233907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211452144.1A Active CN115713691B (en) | 2022-11-21 | 2022-11-21 | Noctilucent remote sensing-based pixel-level power popularity rate estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115713691B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117495425A (en) * | 2023-12-29 | 2024-02-02 | 武汉大学 | Asset financial estimation method and system based on multidimensional noctilucent features |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107368922A (en) * | 2017-07-20 | 2017-11-21 | 华中师范大学 | Average Price of City Residence predictor method based on nighttime light intensity |
| CN108399399A (en) * | 2018-03-23 | 2018-08-14 | 武汉大学 | A kind of city scope extracting method based on noctilucence remote sensing image |
| CN108804394A (en) * | 2018-05-23 | 2018-11-13 | 河海大学 | A kind of construction method of city noctilucence total amount-urban population regression model |
| CN110765885A (en) * | 2019-09-29 | 2020-02-07 | 武汉大学 | City expansion detection method and device based on heterogeneous luminous remote sensing image |
| CN114004278A (en) * | 2021-09-30 | 2022-02-01 | 厦门理工学院 | Multi-source night light data correction fusion method, terminal equipment and medium |
| CN115293974A (en) * | 2022-05-30 | 2022-11-04 | 北京林业大学 | GDP estimation method based on big data combined with night light remote sensing image |
-
2022
- 2022-11-21 CN CN202211452144.1A patent/CN115713691B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107368922A (en) * | 2017-07-20 | 2017-11-21 | 华中师范大学 | Average Price of City Residence predictor method based on nighttime light intensity |
| CN108399399A (en) * | 2018-03-23 | 2018-08-14 | 武汉大学 | A kind of city scope extracting method based on noctilucence remote sensing image |
| CN108804394A (en) * | 2018-05-23 | 2018-11-13 | 河海大学 | A kind of construction method of city noctilucence total amount-urban population regression model |
| CN110765885A (en) * | 2019-09-29 | 2020-02-07 | 武汉大学 | City expansion detection method and device based on heterogeneous luminous remote sensing image |
| CN114004278A (en) * | 2021-09-30 | 2022-02-01 | 厦门理工学院 | Multi-source night light data correction fusion method, terminal equipment and medium |
| CN115293974A (en) * | 2022-05-30 | 2022-11-04 | 北京林业大学 | GDP estimation method based on big data combined with night light remote sensing image |
Non-Patent Citations (1)
| Title |
|---|
| 郑渊茂;何原荣;王晓荣;高元衡;: "夜光遥感数据应用述评与展望", 遥感信息, no. 03 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117495425A (en) * | 2023-12-29 | 2024-02-02 | 武汉大学 | Asset financial estimation method and system based on multidimensional noctilucent features |
| CN117495425B (en) * | 2023-12-29 | 2024-04-12 | 武汉大学 | Asset financial estimation method and system based on multidimensional noctilucent features |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115713691B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zheng et al. | The desaturation method of DMSP/OLS nighttime light data based on vector data: Taking the rapidly urbanized China as an example | |
| Lauko et al. | Local color and morphological image feature based vegetation identification and its application to human environment street view vegetation mapping, or how green is our county? | |
| CN115713691A (en) | Pixel-level electric power popularity estimation method and device based on noctilucent remote sensing | |
| CN114120101A (en) | Soil moisture multi-scale comprehensive sensing method | |
| CN114241326A (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
| CN116863341B (en) | Crop classification and identification method and system based on time sequence satellite remote sensing image | |
| CN113989287A (en) | Urban road remote sensing image segmentation method and device, electronic equipment and storage medium | |
| CN111696055B (en) | Method for acquiring long-time sequence noctilucent remote sensing image based on radiation correction | |
| CN114511787A (en) | Neural network-based remote sensing image ground feature information generation method and system | |
| CN116229001A (en) | Urban three-dimensional digital map generation method and system based on spatial entropy | |
| CN116611725A (en) | Land type identification method and device based on green ecological index | |
| CN113344247B (en) | Deep learning-based power facility site selection prediction method and system | |
| CN116415499A (en) | Community comfort simulation prediction method | |
| CN114580975A (en) | Method, system, device and medium for acquiring street vitality | |
| CN115909080A (en) | Land utilization data-based multi-source noctilucent remote sensing image integration method | |
| CN115641498A (en) | Medium-term rainfall forecast post-processing correction method based on space multi-scale convolutional neural network | |
| CN114581758A (en) | Plant leaf disease identification method based on EfficientNet network and deep learning | |
| CN110276755A (en) | A kind of knub position positioning system and relevant apparatus | |
| CN115690585B (en) | Method and system for extracting wheat tillering number based on digital photo | |
| CN117523417B (en) | Method and electronic equipment applied to unified right-confirming registration of natural resources | |
| CN117574787B (en) | Indoor lighting rate simulation system, method and device for indoor design | |
| CN117195016B (en) | Sewage treatment mode determining method and device, computer equipment and storage medium | |
| CN114743108B (en) | Grassland mouse condition identification and quantification method and mouse condition recorder | |
| CN117994769A (en) | Intelligent image recognition method and system for lightning arrester counter and gas pressure gauge | |
| CN117332903A (en) | Future farmland distribution prediction method and device, electronic equipment and storage medium |
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 |