Urban street land fine division method and system based on deep learning

Abstract

The invention discloses an urban street land fine division method and system based on deep learning, and belongs to the field of geographic information processing. The method comprises the following steps: firstly, collecting remote sensing image data, building space form data, business form POI data and urban land data of a case city district, extracting various kinds of information, converting the information into two-dimensional plane image format data, and carrying out rasterization and data enhancement processing; then training the land use division convolutional neural network deep learning model through a grid translation window method and a grid land use function judgment rule, performing rasterization function judgment on urban land use in a target area, dividing grid boundaries through a watershed method, and finally checking and correcting a result to obtain a target urban land use function judgment result. And obtaining a final result of land division of the target district. According to the method, the internal land of the urban land street can be finely divided, automatic division and function judgment of the urban land are realized, and digital and intelligent application conversion of urban planning work is promoted.

Description

Technical Field

[0001] The present invention belongs to the field of geographic information processing, and specifically relates to a method and system for finely dividing urban neighborhood land based on deep learning. Background Art

[0002] On the one hand, with the development of deep learning, relevant research on the automatic identification of urban land by machines has emerged. There are a lot of contents on the theme of automatic identification of land functions in current research, and different identification technologies are used to study the automatic identification of land functions from different perspectives. However, there are problems such as inaccurate function identification, unclear boundary division, and weak interactivity of identification results. On the other hand, traditional large-scale urban land identification through manual verification consumes huge manpower and material resources. In traditional methods, land information collection often relies on manpower for field research, which is time-consuming and labor-intensive, and has a long update cycle. Therefore, it is necessary to combine existing remote sensing images, architectural space forms, business POI and other multi-source data using deep learning technology to study a method for automatically and accurately determining urban neighborhood-scale land use. Summary of the invention

[0003] Purpose of the invention: In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method and system for fine division of urban neighborhood land use based on multi-source data and deep learning, so as to improve the speed of identifying the existing land use functions and the accuracy of land use function boundary division, shorten the cycle of land use function data update, and reduce manpower consumption.

[0004] Technical solution: The purpose of the present invention can be achieved through the following technical solution:

[0005] A method for finely dividing urban neighborhood land based on deep learning, comprising the following steps:

[0006] Step S1: Collect remote sensing image data, building space morphology data, business POI data and urban land use data of the case city area to build a case city database;

[0007] Step S2: extracting remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and converting various types of information into two-dimensional plane image format data, rasterizing the two-dimensional information image and performing data enhancement processing to construct a case city raster information database;

[0008] Step S3: Through the translation window method and grid land use function determination rules, the remote sensing image information, building space form information, business POI information and corresponding urban land use function information contained in each grid are subjected to deep learning training, and the training accuracy of the land use division convolutional neural network deep learning model is recorded; the model is trained multiple times with different window sizes, and the results of multiple trainings are superimposed to select the model with the highest training accuracy;

[0009] Step S4: Perform two-dimensional transformation and data enhancement on the remote sensing image data, building space morphology data, and business POI data of the target area, and perform rasterized functional determination on the urban land in the target area based on the land use division convolutional neural network deep learning model constructed in step S3; divide the grid boundaries by the watershed method to obtain the land division and functional determination results of the target area, and perform verification and correction to obtain the final result of the land division of the target area.

[0010] Furthermore, in step S2, remote sensing image information, building space form information and business POI information of each neighborhood in the case city database are extracted, and various types of information are converted into two-dimensional plane image format data, specifically including the following contents: calculating the range mask, boundary mask and distance mask of the remote sensing image data to form a two-dimensional plane image data set of remote sensing image information; calculating the spatial coordinate data, floor area data, total building area data and height data of each building unit of the building space form data, using kernel density analysis to calculate its distribution status, using the natural break point classification method to divide its distribution status, and forming a two-dimensional plane image data set of building space form information; calculating the total number data, density data, mixing degree data and nearest neighbor index data of each neighborhood POI data, using kernel density analysis to calculate its distribution status, using the natural break point classification method to divide its distribution status, and forming a two-dimensional plane image data set of business POI information.

[0011] Furthermore, in step S3, deep learning training is performed by using a translation window method, specifically including: using the ResUNet-a algorithm, establishing a plane coordinate system for the area, recording the image grid of any plot as S(x, y), where x is the horizontal coordinate and y is the vertical coordinate, and establishing a 2×2 inference window W(x, y) consisting of four adjacent grids, Then each grid S(x,y) is included in the reasoning windows W(x,y), W(x-1,y), W(x,y-1), and W(x-1,y-1), becoming the intersection of the four windows. Therefore, an inference model based on the ResUNet-a algorithm is established to perform deep learning inference on the land use function determination of S(x,y).

[0012] Furthermore, the grid land use function determination rule for grid S(x, y) in step S3 includes two steps: the first step is to parse the semantic features of each grid remote sensing image based on the translation window method, and obtain the preliminary determination result of the grid attribution by clustering the image semantic features; the second step is to first connect the grid missing spatial morphology and POI data with the nearest similar grid based on the preliminary determination result of the grid attribution, and give it the spatial morphology and POI data of the similar grid; based on the grid function determination rule system of spatial morphology and POI data shown in Table 1, the significance of each spatial morphology characteristic index of the grid is calculated by comparing the actual land use nature corresponding to each grid, and the grid is divided in combination with the determination rule and the urban land classification and planning and construction land standards;

[0013] Table 1. The grid function determination rule system of spatial morphology and POI data

[0014]

[0015]

[0016] The image semantic feature clustering specifically refers to the ResUNet-a algorithm model judging the grid based on the multi-dimensional features of the image constituent elements, such as type, shape, area ratio, relative position, and direction;

[0017] The grid classification specifically refers to classifying the grid into three categories, namely, R, B, and G, based only on feature differences, which is not strictly corresponding to the three types of grids in the second step;

[0018] The significance of the spatial morphological characteristic index specifically refers to the degree of influence of each spatial morphological characteristic index on the grid determination result.

[0019] Furthermore, in step S3, the model is trained multiple times with different window sizes, and the results of multiple trainings are superimposed to select the model with the highest training accuracy. This means that multiple windows of various specifications such as 2×2, 3×3, 4×4, 5×5, 6×6… are used to infer the function of S(x, y) until the inference window area reaches 25% of the overall plot area, and the training result with the highest accuracy is selected.

[0020] Furthermore, in step S4, the watershed method is used to divide the grid boundary, specifically, the discontinuous grid is determined as the preset foreground, the remaining grid is determined as the preset background, the preset foreground is expanded, the image is processed using the distance transformation function cv2.distanceTransform(), and the result is segmented based on the similarity of grid features to obtain a determined foreground area and an unknown foreground area, the preset foreground and the determined foreground are used as the foreground, the unknown foreground area and other remaining grids are used as the background, and then the foreground and the background are retained or deleted in combination with the inference rules shown in Table 2 to complete the division of the land boundary;

[0021] Table 2 Grid boundary division rule system

[0022]

[0023] The expansion operation specifically refers to connecting discontinuous grids to form a preset region with discontinuous grid features, and then judging whether the preset region has significant differences from the background.

[0024] Furthermore, in step S4, for each land use unit, the floor area ratio data, building density number, and average building height data of its building space form information are calculated respectively, and the density data and mixing degree data of the business POI information are calculated. According to the urban land classification and planning and construction land standards, a parameterized rule system is constructed to verify and automatically correct the function judgment results.

[0025] Furthermore, the method also includes step S5: establishing a holographic projection sand table, outputting the land division results of the target area, and interactively displaying various types of information of the target area.

[0026] A system for fine-grained urban neighborhood land division based on deep learning, including:

[0027] The data collection and database construction module is used to collect remote sensing image data, building space morphology data, business POI data and urban land use data of the case city area to build the case city database;

[0028] The multi-source data two-dimensional conversion and data enhancement processing module is used to extract the remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and convert various types of information into two-dimensional plane image format data, rasterize the two-dimensional information image and perform data enhancement processing to build the case city raster information database;

[0029] The deep learning model construction and database training module is used to conduct deep learning training on the remote sensing image information, building space form information, business POI information and corresponding urban land function information contained in each grid through the translation window method and grid land function determination rules, and record the training accuracy of the land division convolutional neural network deep learning model; conduct multiple rounds of training on the model with different window sizes, superimpose multiple training results, and select the model with the highest training accuracy;

[0030] The target area land division and correction module is used to perform two-dimensional conversion and data enhancement processing on the remote sensing image data, building space morphology data, and business POI data of the target area, and to perform rasterized functional judgment on the urban land in the target area based on the trained land division convolutional neural network deep learning model; the grid boundaries are divided by the watershed method to obtain the land division and functional judgment results of the target area, which are then checked and corrected to obtain the final result of the land division of the target area.

[0031] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the method for fine division of urban neighborhood land based on deep learning.

[0032] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0033] 1. The present invention uses the translation window method and grid land use function determination rules to conduct deep learning training on the remote sensing image information, building space morphology information, business POI information and corresponding urban land use function information contained in each grid. By performing multiple rounds of training on the model with different window sizes, the model with the highest training accuracy is selected, thereby improving the accuracy of land use function identification and the accuracy of grid division.

[0034] 2. The present invention adopts the watershed method to divide the grid boundaries, and verifies and corrects the functional determination results after the land use division, which can effectively improve the division speed and accuracy, improve the accuracy of land boundary demarcation, improve the credibility of the division results, and greatly reduce the workload of manual verification.

[0035] 3. The present invention further constructs a parameterized rule system for each land unit. After obtaining the functions of each land unit, it adds the plot ratio data, building density, and average building height data of the building space form information for each land unit, and calculates the density data and mixed degree data of the business POI information, and further refines the accuracy of the recognition results. Compared with the traditional recognition method, this method effectively increases the accuracy of the recognition results. By limiting the threshold of the plot attribute data, it reduces the possibility of misidentification of some special plots and effectively guarantees the credibility of the results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a method flow chart of an embodiment of the present invention;

[0037] Figure 2 is a schematic diagram of the gridding of neighborhood land in an embodiment of the present invention;

[0038] Figure 3 Schematic diagram of the grid translation window method for neighborhood land use in an embodiment of the present invention;

[0039] Figure 4 Schematic diagram of land grid boundary division using the watershed method in an embodiment of the present invention;

[0040] Figure 5 This is a visual display example in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] like Figure 1 As shown, a method for finely dividing urban neighborhood land based on deep learning disclosed in an embodiment of the present invention includes the following steps:

[0043] Step S1, data collection and database construction, collect remote sensing image data, building space morphology data, business POI (Point of Interest) data and urban land use data of the case city area, and build a case city database.

[0044] In some embodiments, spatial calibration and error correction are also performed on various types of collected data, including the following: all remote sensing image data, building space morphology data, business POI data, and urban land data are imported into the same GIS database, and all data are unified into the projection coordinates WGS1984UTM ZONE 51N. Among them, the business POI data is cleaned to remove irrelevant data and data with missing values, and spatial registration is performed on data with inaccurate coordinates.

[0045] Step S2, two-dimensional conversion and data enhancement processing of multivariate data, extracting remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and converting various types of information into two-dimensional plane image format data, further rasterizing the two-dimensional information image, and performing data enhancement processing on each neighborhood data to construct a case city raster information database.

[0046] like Figure 2 As shown, in this embodiment, the grid size is 5m*5m. The remote sensing image information, building space form information and business POI information of each neighborhood in the case city database are calculated, and various types of information are converted into two-dimensional plane image format data, specifically including the following contents: calculating the range mask, boundary mask and distance mask of the remote sensing image information to form a two-dimensional plane image data set of the remote sensing image information; calculating the spatial coordinate data, floor area data, total building area data and height data of each building unit of the building space form data, using kernel density analysis in the ArcGIS platform to calculate its distribution status, using the natural break point classification method to divide its distribution status, and forming a two-dimensional plane image data set of the building space form information; calculating the total number data, density data, mixed degree data and nearest neighbor index data of each neighborhood POI data of the business POI, using kernel density analysis in the ArcGIS platform to calculate its distribution status, using the natural break point classification method to divide its distribution status, and forming a two-dimensional plane image data set of the business POI information.

[0047] The data enhancement processing of each neighborhood data specifically includes: geometric transformation and color space transformation of the images in the case city raster information database to increase data diversity and improve the generalization ability of the model, optimize scale adaptability, and avoid overfitting. Specifically, the geometric transformation includes horizontal and vertical flipping and perspective transformation of the image, and the color space transformation includes adjusting the brightness of the image, increasing or decreasing the saturation of the image, adding Gaussian noise and color shift.

[0048] Step S3, construction of deep learning model of convolutional neural network for land use division and database training, based on the two-dimensional information database of the case city established in step S2, establish an initial grid for the neighborhood information of the case city in the ArcGIS platform, and use the translation window method and grid land use function determination rules to conduct deep learning training on the remote sensing image information, building space form information, business POI information and corresponding urban land use function information contained in each grid, and record the training accuracy of the deep learning model of convolutional neural network for land use division. Perform multiple rounds of training on the model with different window sizes, superimpose multiple training results, and select the model with the highest training accuracy.

[0049] In this embodiment, the deep learning training by the translation window method specifically includes: using the ResUNet-a algorithm to establish a plane coordinate system for the area, recording the image grid of any plot as S(x, y), where x is the horizontal coordinate and y is the vertical coordinate, and establishing a 2×2 inference window W(x, y) consisting of four adjacent grids, Then each grid S(x,y) is included in the inference windows W(x,y), W(x-1,y), W(x,y-1), and W(x-1,y-1), becoming the intersection of the four windows. Therefore, an inference model [(W(x,y)),...,(W(x-1,y-1))] based on the ResUNet-a algorithm is established to perform deep learning inference on the land use function determination of S(x,y).

[0050] The grid land use function judgment rule for S(x,y) includes two steps: the first step is to analyze the semantic features of each grid remote sensing image based on the translation window method, and obtain the preliminary judgment result of the grid attribution by clustering the image semantic features. The second step is to first connect the grid missing spatial morphology and POI data with the nearest grid of the same type based on the preliminary judgment result of the grid attribution, and give it the spatial morphology and POI data of the same type. Further, based on the grid function judgment rule system based on spatial morphology and POI data shown in Table 1, the significance of each grid spatial morphology characteristic index is calculated by comparing the actual land use nature corresponding to each grid, and the grid function is divided in combination with the judgment rule and the urban land classification and planning and construction land standard (GB50137-2011), where R grid represents residential function, B grid represents commercial function, and G grid represents green space function.

[0051] Table 1. The grid function determination rule system of spatial morphology and POI data

[0052]

[0053] The image semantic feature clustering specifically refers to the ResUNet-a algorithm model judging the grid based on the multi-dimensional features of the image constituent elements, such as type, shape, area ratio, relative position, and direction.

[0054] The grid classification specifically refers to classifying the grid into three categories, namely, R, B, and G, based only on feature differences, which is not strictly corresponding to the three types of grids in the second step.

[0055] The significance of the spatial morphological characteristic index specifically refers to the degree of influence of each spatial morphological characteristic index on the grid determination result.

[0056] By comparing the actual land use properties corresponding to each grid, the significance of each spatial morphological characteristic index of the grid is calculated, and the grid is further judged according to the rules in the table.

[0057] The model is trained multiple times with different window sizes, multiple training results are superimposed, and the highest model training accuracy is selected. This means that multiple windows of various specifications such as 2×2, 3×3, 4×4, 5×5, 6×6, etc. are used to infer the function of S(x, y) until the inference window area reaches 25% of the overall land area, and the training result with the highest accuracy is selected. Figure 3 shown.

[0058] Step S4: Refined land use division and parameterized rule correction of the target area. The remote sensing image data, building space morphology data, and business POI data of the target area are converted into two dimensions and data enhanced by the method described in step S2. Based on the land use division convolutional neural network deep learning model constructed in step S3, the urban land in the target area is subjected to rasterized function determination. The grid boundary is further divided by the watershed method to obtain the land use division and function determination results of the target area, and the results are checked and corrected to obtain the final result of the land use division of the target area.

[0059] In this embodiment, the watershed method is used to divide the grid boundary, specifically, the non-continuous grid is determined as the preset foreground, the remaining grid is determined as the preset background, the preset foreground is expanded, the image is processed using the distance transformation function cv2.distanceTransform(), and the result is segmented based on the similarity of grid features to obtain a determined foreground area and an unknown foreground area, the preset foreground and the determined foreground are used as the foreground, the unknown foreground area and other remaining grids are used as the background, and then the foreground and the background are retained or deleted in combination with the inference rules shown in Table 2 to complete the division of the land boundary. The schematic diagram of the watershed method for dividing the land grid boundary is shown in FIG. Figure 4 As shown, (a) is an example of a discontinuous grid before correction, and (b) is an example of a grid boundary after correction.

[0060] Table 2 Grid boundary division rule system

[0061]

[0062] The expansion operation specifically refers to connecting discontinuous grids to form a preset region with discontinuous grid features, and then judging whether the preset region has significant differences from the background.

[0063] In some embodiments, a parameterized rule system is also constructed for each land unit, specifically, for each land unit, the volume ratio data, building density, and average building height data of its building space form information are calculated, and the density data and mixed degree data of the business POI information are calculated. According to the urban land classification and planning and construction land standard (GB50137-2011), a parameterized rule system is constructed to check and automatically correct the function determination results. Exemplarily, the constructed parameterized rule system is shown in Table 3:

[0064] Table 3 Parameterized rule system for each land use unit

[0065]

[0066] In some embodiments, it also includes: step S5, result output and interactive display, establishing a holographic projection sand table, outputting the land division results of the target area, interactively displaying various types of information of the target area, and the required equipment includes a three-dimensional physical sand table, a digital holographic projector and a gesture recognizer.

[0067] The holographic projection sand table is established to output the land division results of the target area and interactively display various information of the target area. Specifically, a three-dimensional digital sand table is constructed using a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, and a sand table, and a gesture recognizer is used to display the land use information of the target area in a virtual scene. Figure 5 Shown is an example of an exported visualization.

[0068] The method for fine division of urban neighborhood land based on deep learning described in the embodiment of the present invention is of great significance in quickly and accurately identifying the functions of urban neighborhood land and automatically determining them. It solves many problems such as the difficulty of traditional land identification, unclear boundary division, large workload of manual identification, and difficulty in determining machine identification boundaries, and realizes the fine division of urban neighborhood land and automatic verification of the results.

[0069] The technical solution of the present invention will be described in detail below by taking the detailed division of urban neighborhood land in a certain area of ​​Nanjing as an example.

[0070] (1) Taking a certain area in Nanjing as the target area, remote sensing image data, building space morphology data, business POI data and urban land use data of the area are collected to build a database of a certain area in Nanjing, including:

[0071] (1.1) Import remote sensing image data, building space form data, business POI data and urban land data into the same GIS database, and unify the coordinates of all data into the projection coordinates WGS1984UTM ZONE 51N. The business POI data is the SHP file of each POI point; the urban land data includes the CAD / SHP file of the current closed plot in the area and the CAD / SHP file of the current building in the plot.

[0072] (1.2) Clean the business POI data, remove irrelevant data and data with missing values, and perform spatial registration on data with inaccurate coordinates.

[0073] (2) The remote sensing image information, architectural space form information and business POI information of each neighborhood in a database of a certain area in Nanjing are converted into two-dimensional plane image format data, including the following contents: calculating the range mask, boundary mask and distance mask of the remote sensing image information to form a two-dimensional plane image data set of the remote sensing image information; calculating the spatial coordinate data, floor area data, total building area data and height data of each building unit of the architectural space form information, using kernel density analysis on the ArcGIS platform to calculate their distribution status, using the natural break point classification method to divide their distribution status, and forming a two-dimensional plane image data set of the architectural space form information; calculating the total number of POI data, density data, mixed degree data and nearest neighbor index data of each neighborhood of the business POI information, using kernel density analysis on the ArcGIS platform to calculate their distribution status, using the natural break point classification method to divide their distribution status, and forming a two-dimensional plane image data set of the business POI information.

[0074] The obtained two-dimensional plane image dataset is first subjected to geometric transformation, including horizontal and vertical flipping and perspective transformation; secondly, the color space transformation is performed, including adjusting the image brightness, increasing or decreasing the image saturation, adding Gaussian noise and color shift, so as to increase data diversity and improve the model generalization ability, optimize scale adaptability, and avoid overfitting.

[0075] (3) Construction of deep learning model of convolutional neural network for land use division and database training. Based on the database of a certain area in Nanjing, a grid with an initial size of 5m*5m was established for the neighborhood information in the area in the ArcGIS platform. Through the translation window method and grid land use function determination rules, the remote sensing image information, building space form information, business POI information and corresponding urban land use function information contained in each grid were trained for deep learning, and the training accuracy of the deep learning model of convolutional neural network for land use division was recorded. Multiple rounds of training with different window sizes were performed on the function of S(x,y) using multiple windows of various specifications such as 2×2, 3×3, 4×4, 5×5, 6×6, etc., until the reasoning window area reached 25% of the overall plot area. The training results were superimposed and the highest model training accuracy was selected.

[0076] (4) Refined division of land use in the target area and modification of parameterized rules. The watershed method is used to divide the grid boundary. Specifically, the discontinuous grid is determined as the preset foreground, and the remaining grid is determined as the preset background. The preset foreground is expanded, and the distance transformation function cv2.distanceTransform() is used to process the image. The result is segmented based on the similarity of grid features to obtain the determined foreground area and the unknown foreground area. The preset foreground and the determined foreground are used as the foreground, and the unknown foreground area and other remaining grids are used as the background. Then, the foreground and the background are retained or deleted in combination with the inference rules shown in Table 2 to complete the division of the land boundary.

[0077] A parameterized rule system is constructed for each land use unit. Specifically, for each land use unit, the floor area ratio data, building density, and average building height data of its building space morphology information are calculated, and the density data and mixed degree data of the business POI information are calculated. According to the urban land classification and planning and construction land standard (GB50137-2011), the rule system is constructed as shown in Table 3.

[0078] (5) Result output and interactive display: Establish a holographic projection sand table, output the land use division results of the area, and interactively display various information of the target area. The required equipment includes a three-dimensional physical sand table, a digital holographic projector, and a gesture recognizer.

[0079] Among them, a holographic projection sand table is established to output the land division results of the target area, and interactively display various information of the target area. Specifically, a three-dimensional digital sand table is constructed using a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, and a sand table, and a gesture recognizer is used to display the land use information of the target area in a virtual scene.

[0080] Based on the same inventive concept, an embodiment of the present invention discloses a system for finely dividing urban neighborhood land based on deep learning, including:

[0081] The data collection and database construction module is used to collect remote sensing image data, building space morphology data, business POI data and urban land use data of the case city area to build the case city database;

[0082] The multi-source data two-dimensional conversion and data enhancement processing module is used to extract the remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and convert various types of information into two-dimensional plane image format data, rasterize the two-dimensional information image and perform data enhancement processing to build the case city raster information database;

[0083] The deep learning model construction and database training module is used to conduct deep learning training on the remote sensing image information, building space form information, business POI information and corresponding urban land function information contained in each grid through the translation window method and grid land function determination rules, and record the training accuracy of the land division convolutional neural network deep learning model; conduct multiple rounds of training on the model with different window sizes, superimpose multiple training results, and select the model with the highest training accuracy;

[0084] The target area land division and correction module is used to perform two-dimensional conversion and data enhancement processing on the remote sensing image data, building space morphology data, and business POI data of the target area, and to perform rasterized functional judgment on the urban land in the target area based on the trained land division convolutional neural network deep learning model; the grid boundaries are divided by the watershed method to obtain the land division and functional judgment results of the target area, which are then checked and corrected to obtain the final result of the land division of the target area.

[0085] An embodiment of the present invention also discloses a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for fine division of urban neighborhood land based on deep learning.

Claims

1. A method for fine division of urban neighborhood land based on deep learning, characterized in that: The following steps are involved: Step S1: Collect remote sensing image data, building space morphology data, business POI data and urban land use data of the case city area to build a case city database; Step S2: extracting remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and converting various types of information into two-dimensional plane image format data, rasterizing the two-dimensional information image and performing data enhancement processing to construct a case city raster information database; Step S3: Through the translation window method and grid land use function determination rules, the remote sensing image information, building space form information, business POI information and corresponding urban land use function information contained in each grid are subjected to deep learning training, and the training accuracy of the land use division convolutional neural network deep learning model is recorded; the model is trained multiple times with different window sizes, and the results of multiple trainings are superimposed to select the model with the highest training accuracy; Step S4: Perform two-dimensional transformation and data enhancement on the remote sensing image data, building space morphology data, and business POI data of the target area, and perform rasterized functional determination on the urban land in the target area based on the land use division convolutional neural network deep learning model constructed in step S3; divide the grid boundaries by the watershed method to obtain the land division and functional determination results of the target area, and perform verification and correction to obtain the final result of the land division of the target area.

2. According to the method of claim 1, the method is characterized by: In step S2, remote sensing image information, building space form information and business POI information of each neighborhood in the case city database are extracted, and various types of information are converted into two-dimensional plane image format data, which specifically includes the following contents: Calculate the range mask, boundary mask and distance mask of the remote sensing image data to form a two-dimensional plane image data set of remote sensing image information; The spatial coordinate data, floor area data, total building area data and height data of each building unit in the architectural space morphology data are calculated, and their distribution status is calculated using kernel density analysis. The distribution status is divided using the natural break point classification method to form a two-dimensional plane image data set of architectural space morphology information; the total number data, density data, mixing degree data and nearest neighbor index data of each neighborhood POI data of business format are calculated, and their distribution status is calculated using kernel density analysis. The distribution status is divided using the natural break point classification method to form a two-dimensional plane image data set of business format POI information.

3. According to the method of fine division of urban neighborhood land based on deep learning in claim 1, it is characterized in that: In step S3, deep learning training is performed by using the translation window method, which specifically includes: using the ResUNet-a algorithm to establish a plane coordinate system for the area, recording the image grid of any plot as S(x, y), where x is the horizontal coordinate and y is the vertical coordinate, and establishing a 2×2 inference window W(x, y) consisting of four adjacent grids. Then each grid S(x,y) is included in the reasoning windows W(x,y), W(x-1,y), W(x,y-1), and W(x-1,y-1), becoming the intersection of the four windows. Therefore, an inference model based on the ResUNet-a algorithm is established to perform deep learning inference on the land use function determination of S(x,y).

4. According to the method of fine division of urban neighborhood land based on deep learning in claim 3, it is characterized in that: The grid land use function determination rule for grid S(x, y) in step S3 includes two steps: the first step is to parse the semantic features of each grid remote sensing image based on the translation window method, and obtain the preliminary determination result of the grid attribution by clustering the image semantic features; the second step is to first connect the grid missing spatial morphology and POI data with the nearest similar grid based on the preliminary determination result of the grid attribution, and give it the spatial morphology and POI data of the similar grid; based on the grid function determination rule system of spatial morphology and POI data shown in Table 1, the significance of each spatial morphology characteristic index of the grid is calculated by comparing the actual land use nature corresponding to each grid, and the grid is divided in combination with the determination rules and the urban land classification and planning and construction land standards; Table 1. The grid function determination rule system of spatial morphology and POI data The image semantic feature clustering specifically refers to the ResUNet-a algorithm model judging the grid based on the multi-dimensional features of the image constituent elements, such as type, shape, area ratio, relative position, and direction; The grid classification specifically refers to classifying the grid into three categories, namely, R, B, and G, based only on feature differences, which is not strictly corresponding to the three types of grids in the second step; The significance of the spatial morphological characteristic index specifically refers to the degree of influence of each spatial morphological characteristic index on the grid determination result.

5. According to the method of fine division of urban neighborhood land based on deep learning in claim 1, it is characterized in that: In step S3, the model is trained for multiple rounds with different window sizes, and the results of multiple trainings are superimposed to select the model with the highest training accuracy. This means that 2×2, 3×3, 4×4, 5×5, 6×6… multiple windows of various specifications are used to infer the function of S(x, y) until the inference window area reaches 25% of the overall plot area, and the training result with the highest accuracy is selected.

6. The method for finely dividing urban neighborhood land based on deep learning according to claim 1 is characterized in that: In step S4, the watershed method is used to divide the grid boundary, specifically, the discontinuous grid is determined as the preset foreground, the remaining grid is determined as the preset background, the preset foreground is expanded, the image is processed using the distance transformation function cv2.distanceTransform(), and the result is segmented based on the similarity of grid features to obtain the determined foreground area and the unknown foreground area, the preset foreground and the determined foreground are used as the foreground, the unknown foreground area and other remaining grids are used as the background, and then the foreground and the background are retained or deleted in combination with the inference rules shown in Table 2 to complete the division of the land boundary; Table 2 Grid boundary division rule system The expansion operation specifically refers to connecting discontinuous grids to form a preset region with discontinuous grid features, and then judging whether the preset region has significant differences from the background.

7. The method for finely dividing urban neighborhood land based on deep learning according to claim 1 is characterized in that: In step S4, for each land use unit, the floor area ratio data, building density number, and average building height data of its building space form information are calculated respectively, and the density data and mixing degree data of the business POI information are calculated. According to the urban land classification and planning and construction land standards, a parameterized rule system is constructed to verify and automatically correct the function judgment results.

8. The method for finely dividing urban neighborhood land based on deep learning according to claim 1 is characterized in that: The method also includes step S5: establishing a holographic projection sand table, outputting the land division result of the target area, and interactively displaying various information of the target area.

9. A system for finely dividing urban neighborhood land based on deep learning, characterized in that: include: The data collection and database construction module is used to collect remote sensing image data, building space morphology data, business POI data and urban land use data of the case city area to build the case city database; The multi-source data two-dimensional conversion and data enhancement processing module is used to extract the remote sensing image information, building space form information and business POI information of each neighborhood in the case city database, and convert various types of information into two-dimensional plane image format data, rasterize the two-dimensional information image and perform data enhancement processing to build the case city raster information database; The deep learning model construction and database training module is used to conduct deep learning training on the remote sensing image information, building space form information, business POI information and corresponding urban land function information contained in each grid through the translation window method and grid land function determination rules, and record the training accuracy of the land division convolutional neural network deep learning model; conduct multiple rounds of training on the model with different window sizes, superimpose multiple training results, and select the model with the highest training accuracy; The target area land division and correction module is used to perform two-dimensional conversion and data enhancement processing on the remote sensing image data, building space morphology data, and business POI data of the target area, and to perform rasterized functional judgment on the urban land in the target area based on the trained land division convolutional neural network deep learning model; the grid boundaries are divided by the watershed method to obtain the land division and functional judgment results of the target area, which are then checked and corrected to obtain the final result of the land division of the target area.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by the processor, the steps of the method for fine division of urban neighborhood land based on deep learning according to any one of claims 1-8 are implemented.

Images