Deep learning and remote sensing image-based suburb population distribution calculation method and system

Abstract

The invention discloses a suburb population distribution calculation method and system based on deep learning and remote sensing images. The method comprises the following steps: firstly, acquiring remote sensing satellite image data, and converting latitude and longitude coordinates into computable index coordinates; secondly, acquiring house boundary coordinate points under each picture, and constructing a model training sample; carrying out model training by adopting a deep learning framework; then, house boundary coordinate point identification is carried out on an unmarked area; carrying out picture splicing and coordinate conversion on the recognition result, converting the recognition result into a geographic information system format, and obtaining the space distribution and projection area of the rural area houses; and finally, based on the total population distribution in a certain village and town in the statistical yearbook, taking the projection area of the residence in the space as a distribution influence factor, and calculating according to a proportion to obtain relatively accurate space population distribution. Under the condition of lack of mobile phone signaling data, population distribution information is obtained through remote sensing satellite images and deep learning, and the problem that the data basis of suburbs and rural areas is weak is effectively solved.

Description

Technical Field

[0001] The present invention relates to a method and system for calculating the population distribution in suburban areas, specifically a method and system for calculating the population distribution in suburban areas based on deep learning and remote sensing images. Background Art

[0002] The spatial distribution of population activities has an important impact on traffic planning. The urban area has always been the focus of attention of planning, construction, and operation departments. Therefore, both the types of data bases and the data quality are relatively high, and the analysis of the spatial distribution of the population is relatively accurate. However, for the far suburbs or rural areas of the city, the data base is weak. How to effectively analyze the spatial distribution of rural population is crucial for studying suburban travel.

[0003] In the past, the spatial distribution of rural population was not taken seriously, and the technical materials were relatively weak. The main analysis tools and methods were to expand and adjust the population through mobile phone signaling data and district and county statistical data, so as to obtain the population distribution in rural areas. However, the distribution of communication base stations in rural areas is relatively sparse, and the population distribution is relatively wide. Especially when there is a lack of signaling data, there is a lack of effective means to obtain the spatial distribution of the population in rural areas. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a method and system for calculating the population distribution in suburban areas based on deep learning and remote sensing images.

[0005] In a first aspect, the present invention provides a method for calculating the population distribution in suburban areas based on deep learning and remote sensing satellite images, including the following steps:

[0006] Obtain the remote sensing satellite image data of the research area, and perform spatial encoding to convert the longitude and latitude coordinates into computable index coordinates;

[0007] Select small-scale remote sensing images, obtain the house boundary coordinate points under each picture, and construct model training samples;

[0008] Adopt a deep learning framework to train the model and obtain model training parameters;

[0009] For the unlabeled area, based on the trained deep learning model, identify the house boundary coordinate points;

[0010] Perform image stitching and coordinate conversion on the recognition results, convert them into the geographic information system format, and obtain the spatial distribution and projected area of houses in rural areas;

[0011] Based on the total population distribution in a certain village or town in the statistical yearbook, use the projected area of the residence in space as a distribution influence factor, and calculate the relatively accurate spatial population distribution according to the proportion.

[0012] Second aspect, the present invention provides a suburban population distribution calculation system based on deep learning and remote sensing satellite images, including:

[0013] A remote sensing satellite image acquisition module, configured to acquire remote sensing satellite image data of a research area, perform spatial encoding, and convert longitude and latitude coordinates into computable index coordinates;

[0014] A sample construction module, configured to select small-scale remote sensing images, obtain the house boundary coordinate points under each picture, and construct model training samples;

[0015] A model training module, configured to perform model training using a deep learning framework to obtain model training parameters;

[0016] An object recognition module, configured to recognize house boundary coordinate points for unlabeled areas based on the trained deep learning model;

[0017] An image processing module, configured to splice the recognition results and perform coordinate conversion, convert them into the format of a geographic information system, and obtain the spatial distribution and projected area of houses in rural areas;

[0018] A population distribution calculation module, configured to calculate the relatively accurate spatial population distribution based on the total population distribution of a certain village or town in the statistical yearbook, using the projected area of the residence in space as a distribution influencing factor.

[0019] Third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is executed by a processor, it implements the suburban population distribution calculation method based on deep learning and remote sensing satellite images.

[0020] Fourth aspect, the present invention provides a computer program product, characterized in that the program product includes computer-executable instructions for implementing the suburban population distribution calculation method based on deep learning and remote sensing satellite images.

[0021] Advantages of the present invention:

[0022] Improve the accuracy and efficiency of data acquisition: Utilize the deep learning model to automatically identify and extract building information in remote sensing satellite images, reducing the errors and workload of traditional manual recognition.

[0023] Solve the problem of weak data foundation: In the absence of mobile phone signaling data, obtain population distribution information through remote sensing satellite images and deep learning technology, effectively solving the problem of weak data foundation in suburban and rural areas and filling the gap in population distribution research in these areas.

[0024] Achieve refined management of spatial distribution: By taking the projected area of houses as an influencing factor for population distribution, refined management of population distribution is achieved. The spatial resolution of population distribution calculation is improved, which helps to more accurately plan transportation and layout public service facilities. Description of the Drawings

[0025] Figure 1 This is a diagram of an embodiment of the present application. Detailed Implementation Manner

[0026] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0027] In the description of the embodiments of the present application, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design solution described as "for example" or "for instance" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of words such as "for example" or "for instance" is intended to present relevant concepts in a specific manner.

[0028] As Figure 1 shown, the embodiments of the present application propose a method based on remote sensing satellite images and a deep learning framework to obtain the spatial distribution of the population in rural areas in the absence of signaling data. Specifically as follows:

[0029] Step 1: First, download the remote sensing satellite images in the research area and perform spatial encoding to convert the longitude and latitude coordinates into computable index coordinates.

[0030] Step 2: Then, select small-scale remote sensing images, obtain the house boundary coordinate points under each image, and construct model training samples.

[0031] Step 3: Select a deep learning framework based on YOLO V8 for model training to obtain model training parameters.

[0032] Step 4: Select unlabeled areas and identify the house boundary coordinate points based on the trained deep learning model.

[0033] Step 5: Image stitching and coordinate conversion, converting to the Geographic Information System (GIS) format to obtain the spatial distribution and projected area of houses in rural areas under the GIS space.

[0034] Step 6: Based on the total population distribution in a certain village and town in the statistical yearbook, considering that the building categories in rural areas are relatively simple and the residents are mainly distributed in self-built houses, the projected area of the house in space is used as the distribution influence factor, and the relatively accurate spatial population distribution is calculated proportionally.

[0035] On the basis of the above embodiments, as an alternative embodiment, step 1 specifically includes: Based on the boundary of the research area, frame the satellite map of the research area, select the zoom level, download the tile data to the local, and determine the relationship between the longitude and latitude information and the pixel coordinates of the picture.

[0036] Further: Determine the longitude and latitude boundaries of the research scope, the minimum longitude , the maximum longitude , the minimum latitude and the maximum latitude . Based on the longitude and latitude information and the zoom level Zoom, calculate the row number and column number of the tile data:

[0037] =

[0038]

[0039] where: long is the longitude, lat is the latitude; Row is the row number; Col is the column number; is the floor function; Z is the abbreviation of the zoom level Zoom.

[0040] Download the tile data based on the row number, column number of the tile data and the zoom level (the zoom level is recommended to be greater than 17).

[0041] On the basis of the above embodiments, as an alternative embodiment, step 2 specifically includes: Construct a model training sample set, select the tile data, and obtain the boundary vertex coordinates of the house projection in the image through the labeling software.

[0042] Further: Set the label options, such as house, factory, farmland, etc.; corresponding to the values 0, 1, 2 respectively.

[0043] Taking the upper left corner coordinate point of the image as the origin (0, 0), obtain the vertex coordinates of different type labels in the picture, for example, mark the house coordinate points {( , ), ( , ),( , ),( , )}.

[0044] By batch-labeling images, obtain the coordinate point sets of different objects in different images, one-to-one correspondence, to prepare for model training.

[0045] Based on the above embodiments, as an alternative embodiment, step 3 specifically includes: adopting the relatively mature deep learning framework model YOLO v8 to construct a rural area building recognition model under remote sensing satellite maps, and performing model training and verification based on the sample set constructed in step 2.

[0046] Furthermore: The deep learning framework of YOLO v8 is open-sourced by Ultralytics and supports a full range of vision AI tasks, including classification, detection, segmentation, pose estimation, and tracking.

[0047] Based on the above embodiments, as an alternative embodiment, step 4 specifically includes: based on the trained rural area building recognition model, perform building type recognition on the tile data to be processed in the research area, and obtain the boundary coordinates of different buildings.

[0048] Based on the above embodiments, as an alternative embodiment, step 5 specifically includes: merging the tile data, converting the detected building boundary coordinates into longitude and latitude coordinates, and obtaining the spatial area under different targets.

[0049] Furthermore: Based on the tile row number row, column number col, tile zoom level z, and the coordinates to be converted ( , ), ( , ), ([[]]END]] , ), ( , ), the conversion formula to longitude and latitude coordinates is as follows:

[0050] 5-1. Calculate the geographical range of the tile:

[0051]

[0052]

[0053]

[0054]

[0055] Where: 256 is the width of the pixel, arctan is the arctangent function, sinh is the hyperbolic sine function, is the west boundary longitude of the tile, is the east boundary longitude of the tile, is the north boundary latitude of the tile, Is the southern boundary latitude of the tile.

[0056] 5-2. Calculate the geographic coordinates of the pixel points

[0057] Based on the calculated tile geographic range values, convert the pixel point coordinates into longitude and latitude coordinates (lon 1 , lat 1 ), (lon 2 , lat 2 ), (lon 3 , lat 3 ), (lon 4 , lat 4 ), and the specific calculation is as follows:

[0058]

[0059]

[0060] 5-3. Convert longitude and latitude coordinates to planar coordinates

[0061] The longitude and latitude of the four obtained points are (lon 1 , lat 1 ), (lon 2 , lat 2 ), (lon 3 , lat 3 ), (lon 4 , lat 4 ), and convert them to planar coordinates ( , ), ( , ), ( , ), ( , )

[0062] Select a reference point (such as the first point) and calculate the eastward (x) and northward (y) displacements of each point (unit: meter):

[0063]

[0064]

[0065]

[0066]

[0067] Among them, 111194.9 meters is the average distance of the Earth per degree, R ≈ 6371000 meters, is the difference in longitude between the i-th point and the reference point, is the difference in latitude between the i-th point and the reference point, is the displacement of the i-th point relative to the reference point in the eastward direction, is the displacement of the i-th point relative to the reference point in the northward direction.

[0068] 5-4. The building area is calculated as follows:

[0069]

[0070] Based on the above embodiments, as an alternative embodiment, step 6 specifically includes: dividing traffic zones of appropriate sizes based on the research scope and needs, and marking the streets or villages where the zones are located. Through spatial operations, obtain the grids under the projection of the residential area, obtain the total population under the streets in the yearbook data, and use the residential area under the street as the weight to obtain the total population distribution in different regions. The specific calculation formula is as follows:

[0071]

[0072] Where: is the total population of traffic zone i; is the projected area of the residence in the traffic zone; is the total projected area of the residence in the street; is the total population of the street.

[0073] Based on the same concept as the above embodiments, the embodiment of the present application also provides a suburban population distribution calculation system based on deep learning and remote sensing satellite images, including:

[0074] A remote sensing satellite image acquisition module, configured to acquire remote sensing satellite image data of the research area, perform spatial encoding, and convert longitude and latitude coordinates into computable index coordinates;

[0075] A sample construction module, configured to select small-scale remote sensing images, obtain the house boundary coordinate points under each picture, and construct model training samples;

[0076] A model training module, configured to perform model training using a deep learning framework to obtain model training parameters;

[0077] A target recognition module, configured to recognize house boundary coordinate points for unlabeled areas based on the trained deep learning model;

[0078] An image processing module, configured to splice the recognition results and perform coordinate conversion into a geographical information system format to obtain the spatial distribution and projected area of houses in rural areas;

[0079] A population distribution calculation module, which is used to calculate a relatively accurate spatial population distribution based on the total population distribution of a certain village or town in the statistical yearbook, taking the projected area of the residence in space as the distribution influencing factor and calculating it proportionally.

[0080] It should be noted that when the system provided in the above embodiment realizes its functions, only the division of the above-mentioned functional modules is used for illustration. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiment, which will not be elaborated here.

[0081] Based on the same concept as the above embodiment, the embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the method for calculating the suburban population distribution based on deep learning and remote sensing satellite images as described above.

[0082] Based on the same concept as the above embodiment, the embodiment of the present application also provides a computer program product, which includes computer-executable instructions for realizing the method for calculating the suburban population distribution based on deep learning and remote sensing satellite images.

[0083] In this embodiment, those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware of the terminal device through a program. The program can be stored in a computer-readable storage medium, and the storage medium can include: a flash drive, a read-only memory, a random access memory, a magnetic disk or an optical disc, etc.

[0084] If the integrated unit in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in the above computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in the storage medium and includes several instructions for causing one or more computer devices (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention.

[0085] The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art of the embodiments of the present invention, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for calculating suburban population distribution based on deep learning and remote sensing satellite images, characterized in that: The following steps are involved: Obtain remote sensing satellite image data of the study area, perform spatial encoding, and convert longitude and latitude coordinates into computable index coordinates; Select small-scale remote sensing images, obtain the house boundary coordinates in each image, and construct model training samples; Use deep learning framework to train the model and obtain model training parameters; For unlabeled areas, the house boundary coordinate points are identified based on the trained deep learning model; The recognition results are stitched together and the coordinates are converted into a geographic information system format to obtain the spatial distribution and projection area of ​​houses in rural areas; Based on the total population distribution of a certain village or town in the statistical yearbook, the projected area of ​​the residence in space is used as the distribution influencing factor, and a relatively accurate spatial population distribution is obtained by proportional calculation.

2. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 1 is characterized in that: The step of obtaining remote sensing satellite image data of the study area comprises: Determine the longitude and latitude boundaries of the research area, including the minimum longitude, maximum longitude, minimum latitude, and maximum latitude; Calculate the row number and column number of the tile data based on the latitude and longitude information and the zoom level; According to the row number, column number and zoom level, the tile data is downloaded to the local computer, and the relationship between the longitude and latitude information and the coordinates of the pixel points of the image is determined.

3. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 1 or 2, characterized in that: The step of constructing a model training sample comprises: Set label options, including residential, factory, and farmland, and assign corresponding numerical labels to each; Through the marking software, the coordinates of the boundary vertices of the house projected in the image are obtained, and the coordinate point of the upper left corner of the image is taken as the origin to record the vertex coordinates of different types of labels.

4. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 3 is characterized in that: In the step of using a deep learning framework to perform model training, the deep learning framework is YOLO V8.

5. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 3 is characterized in that: The step of performing image stitching and coordinate conversion on the recognition results comprises: Merge tile data; Calculate the geographic extent of the tile based on its row number, column number, and zoom level; Obtaining the geographic coordinates of any pixel in the image from the geographic range of the tile; The geographic coordinates are converted into plane coordinates.

6. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 1, characterized in that: The step of calculating the relatively accurate spatial population distribution in proportion comprises: Divide traffic areas into appropriately sized areas and mark the streets or villages where the areas are located; Through spatial operations, the grid under the projection of the residential area is obtained; Get the total population of the street in the yearbook data, use the residential area of ​​the street as the weight, and calculate the total population distribution in different areas in proportion.

7. The method for calculating suburban population distribution based on deep learning and remote sensing satellite images according to claim 2, characterized in that: The zoom level is greater than level 17 to ensure the resolution and detail information of the image.

8. A suburban population distribution calculation system based on deep learning and remote sensing satellite images, characterized in that: include: The remote sensing satellite image acquisition module is used to acquire remote sensing satellite image data of the study area, perform spatial encoding, and convert longitude and latitude coordinates into computable index coordinates; The sample construction module is used to select small-scale remote sensing images, obtain the house boundary coordinates under each image, and construct model training samples; Model training module, used to train models using a deep learning framework and obtain model training parameters; The target recognition module is used to identify the house boundary coordinate points in unlabeled areas based on the trained deep learning model; The image processing module is used to stitch the images and convert the coordinates of the recognition results into a geographic information system format to obtain the spatial distribution and projection area of ​​houses in rural areas; The population distribution calculation module is used to calculate the total population distribution of a village or town in the statistical yearbook, take the projected area of ​​the house in space as the distribution influencing factor, and calculate the relatively accurate spatial population distribution in proportion.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the suburban population distribution calculation method based on deep learning and remote sensing satellite images as described in any one of claims 1 to 7.

10. A computer program product, characterized in that The program product includes computer executable instructions for implementing the suburban population distribution calculation method based on deep learning and remote sensing satellite images as described in any one of claims 1 to 7.

Images