Method for realizing rural landscape management based on oblique photography and image recognition technology

Abstract

The present application relates to a kind of method for realizing rural landscape management based on oblique photography and image recognition technology, which utilizes oblique photography and image recognition technology to extract and manage the landscape of rural buildings. Specifically, the present application forms a three-dimensional real scene model and an orthographic image by oblique photography technology, then extracts the buildings by the technical method of image recognition classification supervision, and combines linear regression calculation to optimize the building range line, establishes building model and data archives. Based on this, it summarizes the landscape parameters, discriminates the landscape coordination of new buildings, and discriminates the landscape coordination of village construction changes. The present application can quickly collect and identify the rural landscape situation, realize multi-dimensional management of rural construction and the establishment of landscape control standards, build a rural landscape management system, and also can conveniently and efficiently complete the dynamic update of village landscape situation, realize the dynamic supervision of village landscape.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rural landscape management, and particularly relates to a method for realizing rural landscape management based on oblique photography and image recognition technology. BACKGROUND

[0002] Most of the rural areas in China have a long history and cultural characteristics. After the baptism of rapid industrialization and urbanization, the rural landscape in China has undergone tremendous changes. Although most counties and cities have prepared village planning, rural housing standard atlas and other guidance for village construction to guide and control the architectural style, the connection in the implementation of village construction management is still insufficient, and it is difficult to effectively implement. Therefore, it is particularly important to explore an effective management method for rural landscape.

[0003] At present, due to the scattered layout of villages, the management work is complicated and requires a large amount of manpower and material resources for investigation. Regular monitoring and management work is even more difficult, and illegal construction, false reporting and arbitrary construction often occur. Therefore, there are problems in the management of village construction, such as lack of data, weak management and difficult supervision.

[0004] At present, remote sensing technology is generally used for village planning and construction management. High-resolution remote sensing images are used to divide and plan the rural land types, but this method has the following two problems. First, the data precision of high-resolution remote sensing images is insufficient, and it is difficult to realize the management of small-size construction such as village buildings and structures. Second, high-resolution remote sensing images are two-dimensional images, and there is a lack of data on building facade style, height and material, which is not convenient for building style management. Therefore, the village construction management based on remote sensing images still has certain limitations. SUMMARY

[0005] Therefore, in order to overcome the problems of data vacancy, large amount of management and difficult supervision in rural landscape management, the present application provides a method for realizing rural landscape management based on oblique photography and image recognition technology, so as to realize efficient and scientific management of rural landscape.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is:

[0007] A method for realizing rural landscape management based on oblique photography and image recognition technology, the method comprising the following steps:

[0008] Step S1: using a UAV to take oblique photographs of the village, and setting ground RTK calibration points and check points; based on the UAV photographing data and RTK data, the village is calculated and processed to obtain a village three-dimensional real scene model and a digital orthographic image map;

[0009] Step S2: converting and correcting the three-dimensional real scene model and the orthographic image in the GIS spatial geographic information system in the spatial coordinate system, so that the coordinate systems of the two are coincident;

[0010] Step S3: performing supervised classification on the digital orthographic image in the image analysis software, performing automatic preliminary cutting on the digital orthographic image, selecting samples of buildings, structures and other construction elements in the digital orthographic image, and performing "reclassification" on the orthographic image through comprehensive identification and training of sample color, spectral band ratio, elevation, height difference, shape and aspect ratio sample feature parameters, and exporting a grid map with classification attributes;

[0011] The exported grid map with classification attributes includes buildings, vegetation, ground and water bodies with different attributes; wherein, sample points are extracted from the ground attribute grid map, and a curved surface is automatically fitted based on the sample points, that is, a digital elevation model DEM;

[0012] Step S4: based on the building attribute grid map, extracting and optimizing each building boundary line, and extracting and numbering each building based on the optimized building boundary line, and preliminarily creating a building file;

[0013] Step S5: based on the separated real scene model of each building and the optimized building boundary line, extracting building material space data and constructing a village building material space database;

[0014] Specifically, the separated three-dimensional real scene model of each building is converted into a three-dimensional space point cloud model, and point cloud analysis software is used to analyze the building height (H x ), roof slope (i x ), and facade material color (R x , G x , B x ); based on the optimized building boundary, the building length (L x ), width (W x ), and spacing (D x ) are analyzed;

[0015] The building height (H x ), roof slope (i x ), and facade material color (R x , G x , B x ) and the building length (L x ), width (W x ), and spacing (D x ) constitute the building material space data;

[0016] Step S6: Based on the village building material space database, the village landscape parameter elements are generally statistically analyzed and calculated, and the parameter mean values of the building length-width ratio, height, spacing, color, and roof slope are obtained, which are the building landscape reference parameters of the village, facilitating the preparation of rural housing construction atlas by county, city, and other regions; at the same time, the building landscape of the submitted new construction application can be automatically preliminarily judged and approved;

[0017] The method for automatically preliminarily judging and approving the building landscape of the submitted new construction is as follows: using image recognition technology to intelligently segment and recognize the images such as the building plan and the elevation effect drawing of the submitted application, extracting the length-width-height, material color, roof form, and calculating the evaluation value based on the building landscape parameters;

[0018] Step S7: Regularly shoot the village three-dimensional real scene model and orthographic image map using a drone, establish the latest village building model and plot through the above S1-S5 methods, and compare with the original village database to find changes; then calculate the evaluation value of the changed buildings based on the building landscape parameters (step S6), and evaluate the coordination of the changed plot according to the evaluation value; position and enlarge the buildings with low coordination degree, and send them to the supervision department personnel to realize the three-dimensional supervision and management of the village building landscape.

[0019] In step S4, the method for extracting and optimizing the building boundary line is as follows:

[0020] Get the plot of the grid map of the building attribute, put the plot boundary line into the XY two-dimensional coordinate system, divide it into four direction edges by center cutting, extract each edge to sample points with a density of 50%, read their X and Y values respectively, and calculate the X sum (∑X), Y value sum (∑Y), X*Y value sum (∑X*Y), X*X value sum (∑X*X), Y*Y value sum (∑Y*Y), and put them into the following two formulas to calculate the regression line slope b and the correlation factor a:

[0021]

[0022]

[0023] Thus, the regression straight line is obtained, and linear regression calculation is performed on the four edges in turn, and combined into a closed building boundary line to form the optimal building boundary line.

[0024] In step 5,

[0025] The method for analyzing the building height parameter is as follows: the extracted building point cloud model is subjected to point cloud distance calculation with the village digital elevation model DEM, since the point cloud heights on the elevation are not the same, while the roof layer point cloud is basically at the same height, so the peak value of the distance calculation result is the building roof height value (H x );

[0026] The method for analyzing the roof slope of a building is as follows: The extracted building point cloud model is used to calculate the spatial angles between points. Since the point cloud on the facade is mostly vertical (∠γ0=90°), while the point cloud on the roof is parallel to the ground (∠γ0=0°) or has a certain angle (∠γ0<60°), the peak angle excluding the 90-degree peak is the roof slope (i). x );

[0027] The method for analyzing the material color of building facades is as follows: The extracted building point cloud model is rotated along the XYZ axes in point cloud analysis software to make the building facade parallel to the spatial axis plane. The orthographic projection image of the building facade is then exported. Image recognition software is used to intelligently segment the facade, and the RGB (R) color values ​​of the facade color blocks are extracted. x G x B x ) and the area ratio of color blocks S cx ( S represents the area of ​​the building or region image. x Let x represent the area of ​​different color blocks within the region (where x represents the number of the different color blocks), and sort them according to the area ratio of the color blocks to form the main color tone (S). cx (higher values) and influence on hue (S) cx (lower value);

[0028] Based on the optimized building boundary line, the building length (L) x Width (W) x The building spacing (D) can be obtained by directly reading the data and calculating the relative spacing between the building boundaries. x ).

[0029] The method for calculating the landscape evaluation value is as follows:

[0030] Architectural style assessment value Y x =HX*40%+iX*20%+CX*15%+LWX*10%+DX*5%

[0031] in:

[0032] The highest rating is:

[0033] The roof form evaluation value is: if 10° < i x If i < 60°, then iX = 1, if i x If the angle is less than 10°, then iX = 0.

[0034] Color rating: Where n is the area S of the color block cx The sorting is performed, and the top 3 colored blocks are included in the calculation value (i.e., n = 1, 2, 3). x represents the evaluated building or area, and Rxn R value of the color block color ranked as the n th, and the G value and the B value are the same, R value of the color block color ranked as the n th, and the G value and the B value are the same,

[0035] The plane form evaluation value is:

[0036] The building spacing evaluation value is:

[0037] After the above scheme is adopted, the application constructs a relatively systematic method to facilitate the management of rural construction style, has multiple advantages such as high efficiency, scientificity, low cost and the like; can quickly collect and identify rural construction conditions, realizes multi-dimensional management of rural construction and establishment of style control standards, builds a rural construction management system, overcomes the management of the missing style dimension of the traditional construction remote sensing monitoring system, and can conveniently and efficiently complete dynamic updating of village style conditions, and realizes dynamic supervision of village style.

[0038] The application optimizes extraction of building range lines based on oblique photography and image segmentation technology, forms a regression boundary line through linear regression calculation, avoids the problem of uneven building contour range caused by data acquisition errors of the point cloud model, simplifies the formation of a graph with vector properties more in line with surveying and mapping standards, and makes image and model of the building and morphological data extraction more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 It is a method flow diagram for realizing rural construction management based on oblique photography and image recognition technology provided by the application;

[0040] Figure 2 It is a building grid map and building boundary line extracted based on supervised classification;

[0041] Figure 3 It is a building boundary line formed based on regression calculation of building boundary point clouds;

[0042] Figure 4 It is a village style color parameter value extracted based on image segmentation;

[0043] Figure 5 It is a building facade orthographic image extracted based on XYZ axis correction. DETAILED DESCRIPTION

[0044] As shown in Figure 1 The application discloses a method for realizing rural style management based on oblique photography and image recognition technology, which comprises the following steps:

[0045] Step S1: On-site survey the overall situation of the target village, plan the flight route and flight parameters of the unmanned aerial vehicle for oblique shooting according to the range of the village, and set the ground RTK calibration points and check points; based on the unmanned aerial vehicle shooting data and RTK data, the village is processed by three-dimensional real scene modeling to obtain the three-dimensional real scene model and digital orthographic image of the village.

[0046] Step S2: The three-dimensional real scene model and the orthographic image are converted and corrected in the spatial coordinate system in the GIS spatial geographic information system, so that the coordinate systems of the two are coincident.

[0047] Step S3: The digital orthographic image is classified in the image analysis software (such as eCongnition), the digital orthographic image is automatically cut, the construction elements such as buildings and structures in the digital orthographic image are selected, the digital orthographic image is reclassified by the comprehensive identification and training of the sample characteristics such as color, spectral band ratio, height, height difference, shape, and aspect ratio, and the grid map with classification attributes is exported, as shown in Figure 2

[0048] The exported grid map with classification attributes includes buildings, vegetation, ground, water body and other different attributes. The ground attribute grid map is extracted and fitted to generate a curved surface based on the sample points, that is, a digital elevation model DEM. The grid maps with the same attribute are merged to form a current land class map including cultivated land, forest land, grassland, construction land, transportation land, and water area.

[0049] Step S4: Based on the building attribute grid map, the building boundary line is extracted and optimized, and the three-dimensional real scene model and the orthographic image which have been calibrated and aligned in the GIS are cut according to the optimized building boundary line, and the buildings are extracted and numbered to preliminarily create a building file.

[0050] In this embodiment, the method of extracting and optimizing the building boundary line is as follows: obtaining the graph of the building attribute grid map, putting the graph boundary line into the XY two-dimensional coordinate system, cutting the center into four direction edges, extracting sample points of each edge at a density of 50%, respectively reading the X and Y values, and calculating the total sum of X (∑X), the total sum of Y (∑Y), the total sum of X*Y (∑X*Y), the total sum of X*X (∑X*X), the total sum of Y*Y (∑Y*Y), and then calculating the regression line slope b and the correlation factor a by using the following two formulas:

[0051]

[0052]

[0053] ​The regression straight line of the direction edge is obtained, linear regression calculation is sequentially performed on the four edges, and the closed building boundary line is combined to form the optimal building boundary line, as shown in Figure 3

[0054] Due to the modeling limitations of the oblique photography modeling technology and the error of the image segmentation technology parameter discrimination, the grid map of the building attributes extracted based on the image supervised classification method in step S3 has a certain unevenness, which is inconsistent with the building boundary line standard (smooth curve or flat straight line) in the real world and the topographic surveying and mapping industry. Therefore, the regression line of each edge is obtained by using the regression calculation method, and the building boundary line is combined.

[0055] Step S5: Based on the separated real scene model of each building and the optimized building boundary line, the building material space data is extracted, and the village building material space database is constructed.

[0056] Specifically, the separated three-dimensional real scene model of each building is converted into a three-dimensional space point cloud model, and the point cloud analysis software is used to analyze the building height (H x ), roof slope (i x ), facade material color (R x , G x , B x ), etc. Based on the optimized building boundary, the building length (L x ), width (W x ), and spacing (D x ) are analyzed. The building height (H x ), roof slope (i x ), facade material color (R x , G x , B x ), and building length (L x ), width (W x ), and spacing (D x ) constitute the building material space data.

[0057] The method for analyzing the building height parameter is: the extracted building point cloud model is subjected to point cloud distance calculation with the village digital elevation model DEM. Since the point cloud height on the facade is not uniform, and the roof layer point cloud is basically the same height, the peak value of the distance calculation result is the building roof height value (H x ).

[0058] ​The method for analyzing the roof slope of a building is as follows: The extracted building point cloud model is used to calculate the spatial angles between points. Since the point cloud on the facade is mostly vertical (∠γ0=90°), while the point cloud on the roof is parallel to the ground (∠γ0=0°) or has a certain angle (∠γ0<60°), the peak angle excluding the 90-degree peak is the roof slope (i). x ).

[0059] like Figure 5 As shown, the method for analyzing the material color of a building facade is as follows: The extracted building point cloud model is rotated along the XYZ axes in point cloud analysis software to make the building facade parallel to the spatial axis plane. The orthographic projection image of the building facade is then exported. Image recognition software (such as eCongnition) is used to intelligently segment the facade, and the RGB (R) color values ​​of the facade color blocks are extracted. x G x B x ) and the area ratio of color blocks S cx ( S represents the area of ​​the building or region image. x Let x represent the area of ​​different color blocks within the region (where x represents the number of the different color blocks), and sort them according to the area ratio of the color blocks to form the main color tone (S). cx (higher values) and influence on hue (S) cx (e.g., lower value). Figure 4 (As shown).

[0060] The analysis method for building length, width, and spacing is as follows: based on the optimized building boundary line, the building length (L) is... x Width (W) x The building spacing (D) can be obtained by directly reading the data and calculating the relative spacing between the building boundaries. x ).

[0061] Step S6: Based on the village building material space database, conduct overall statistical analysis of village style parameters, calculate the average values ​​of parameters such as building aspect ratio, height, spacing, color, and roof slope, which serve as reference parameters for the village's architectural style. This facilitates the compilation of rural housing construction atlases at the county and city levels. Simultaneously, it allows for automatic preliminary judgment and approval of architectural styles in submitted new construction applications. The method for automatic preliminary judgment and approval of submitted new building styles is as follows: Image recognition technology is used to intelligently segment and identify images such as submitted building floor plans and elevation renderings, extracting dimensions, height, material color, roof form, etc., and calculating evaluation values ​​based on architectural style parameters. The calculation method for style evaluation values ​​is as follows:

[0062] Architectural style assessment value Y x =HX*40%+iX*20%+CX*15%+LWX*10%+DX*5%

[0063] Wherein:

[0064] The high evaluation value is:

[0065] The roof form evaluation value is: if 10° < i x < 60°, iX = 1, if i x < 10°, iX = 0

[0066] The color evaluation value is: Wherein, n is the color block area S cx The order, take the top 3 color blocks in the calculation value (that is, n = 1, 2, 3), x refers to the evaluation building or area, R xn Refers to the color R value of the color block ranked n, and the G value and B value, Refers to the color average value ranked n in the style parameter, and the G value and B value;

[0067] The planar form evaluation value is:

[0068] The building spacing evaluation value is:

[0069] The regulatory personnel can decide whether to approve the construction of the building according to the style evaluation value of the newly built building, and return the design draft to the rectification scheme that is not passed.

[0070] Step S7: periodically shooting by using a drone to generate a village three-dimensional real scene model and an orthographic image, comparing the latest village building model and the plot with the original village database through the above S1-S5 methods to find changes, and then calculating the evaluation value of the changed building based on the building style parameters (synchronous step S6), and evaluating the coordination of the changed plot according to the evaluation value, for example, when the evaluation value is lower than a certain threshold, it is considered that the coordination is low. The buildings with low coordination are positioned and enlarged, and are sent to the regulatory personnel, so as to realize the three-dimensional supervision and management of the village building style.

[0071] The present application summarizes the village building style parameters, evaluates the style coordination of the change points through the parameters, filters out the changes caused by changes in residents' life or vegetation growth or facade improvement, etc., predicts the style coordination in advance, and reduces the on-site verification work of style control management. That is, the present application scores the changed buildings from the style parameters after comparing the changes, and extracts the change pictures, so as to reduce the on-site supervision workload of the staff.

[0072] In conclusion, the present application constructs a relatively systematic method to facilitate the management of rural construction style, which has multiple advantages such as high efficiency, scientificity, low cost, etc.; it can quickly collect and identify the rural construction situation, realize multi-dimensional management of rural construction and establishment of style control standards, build a rural construction management system, overcome the lack of style dimension management in the traditional construction remote sensing monitoring system, and conveniently and efficiently complete the dynamic update of village style situation, and realize dynamic supervision of village style.

[0073] The present application optimizes the extraction of building range line based on oblique photography and image segmentation technology, forms a regression boundary line through linear regression calculation mode, avoids the uneven building contour range problem caused by data acquisition error of point cloud model, simplifies the formation of a graph with vector attributes more in line with surveying and mapping standards, and makes the image and model of the building and the morphological data extraction more accurate.

[0074] The above is only an embodiment of the present application, and does not limit the technical scope of the present application in any way, so any slight modification, equivalent change and modification of the above embodiment according to the technical essence of the present application still belongs to the scope of the technical solution of the present application.

Claims

1. A method for rural landscape management based on oblique photography and image recognition technology, characterized in that: The method includes the following steps: Step S1: Use a drone to take oblique photos of the village and set ground RTK calibration points and checkpoints; perform data calculation and processing on the village based on the drone data and RTK data to obtain a 3D real-scene model and digital orthophoto map of the village. Step S2: Convert and correct the spatial coordinate system of the 3D real scene model and orthophoto map in the GIS spatial geographic information system so that the coordinate systems of the two coincide. Step S3: Supervised classification of digital orthophotos is performed in image analysis software. The digital orthophotos are automatically pre-cut. Samples of buildings and structures in the digital orthophotos are selected. The orthophotos are reclassified by comprehensively identifying and training the sample feature parameters such as color, spectral band ratio, elevation, height difference, shape, and aspect ratio, and grid map with classification attributes is exported. The exported grid map with classification attributes includes different attributes of buildings, vegetation, ground, and water bodies; among them, sampling points are extracted from the ground attribute grid map, and a surface is automatically fitted based on the sampling points to generate a digital elevation model (DEM). Step S4: Based on the grid map of building attributes, extract and optimize the boundary lines of each building, and extract and number each building according to the optimized building boundary lines of the 3D real scene model and orthophoto map that have been aligned and calibrated in the GIS, and initially create building archives; In step S4, the method for extracting and optimizing the building boundary line is as follows: Obtain the grid map of building attributes, place the boundary line of the grid map into the XY two-dimensional coordinate system, divide it into four directional edges through the center, extract sample points on each edge at a density of 50%, read their X and Y values, and calculate the sum of X values. Sum of Y values The sum of X*Y values The sum of X*X values The sum of Y*Y values Substitute the values ​​into the following two formulas to calculate the slope of the regression line, b, and the correlation factor, a: This yields the regression line of the side in that direction. Linear regression calculations are then performed on the four sides in sequence, and the results are combined to form a closed building boundary line, thus creating the optimal building boundary line. Step S5: Based on the real-world models of each separated building and the optimized building boundary lines, extract the building material space data and construct a village building material space database; Specifically, the separated 3D reality models of each building are converted into 3D spatial point cloud models, and point cloud analysis software is used to determine the building height. Roof slope Facade material and color ( ) to perform analysis; based on the optimized building boundary, the building length is... ,width ,spacing Perform analysis; The building height Roof slope Facade material and color ( and building length ,width ,spacing Constituting spatial data of building materials; Step S6: Based on the village building material space database, perform overall statistics and analysis on the village style parameter elements, calculate the average values ​​of parameters such as building length-to-width ratio, height, spacing, color, and roof slope, which are the reference parameters for the architectural style of the village, facilitating the compilation of rural housing construction atlases at the county and city levels; at the same time, automatically perform preliminary judgment and approval of the architectural style of submitted new construction applications. The method for automatically preliminarily judging and approving the submitted architectural style is as follows: using image recognition technology to intelligently segment and recognize the submitted architectural floor plan and elevation rendering, extracting length, width, height, material color, and roof form, and calculating the evaluation value based on architectural style reference parameters; Step S7: Use drones to periodically take pictures to generate a 3D real-world model and orthophoto of the village. Use the methods in S1-S5 above to establish the latest village building model and patches, and compare them with the original village database to find changes. Then, calculate the evaluation value of the buildings that have changed based on the architectural style reference parameters, and evaluate the coordination of the changed patches based on the evaluation value. Locate and enlarge the buildings with low coordination, and send them to the regulatory department personnel to realize three-dimensional supervision and management of the village's architectural style.

2. The method for rural landscape management based on oblique photography and image recognition technology according to claim 1, characterized in that: In step S5 The method for analyzing building height parameters is as follows: The extracted building point cloud model and the village digital elevation model (DEM) are used to calculate the point cloud distance. Since the point cloud heights on the facades are different, while the point cloud heights on the roof are basically the same, the peak value of the distance calculation result is the building roof height. ; The method for analyzing the roof slope of a building is as follows: The extracted building point cloud model is used to calculate the spatial angles between points. Since the point cloud on the facade is mostly vertical (∠γ0 = 90°), while the point cloud on the roof is parallel to the ground (∠γ0 = 0°) or has a certain angle (∠γ0 < 60°), the peak angle excluding the 90° peak is the roof slope. ; The method for analyzing the material color of building facades is as follows: The extracted building point cloud model is rotated along the XYZ axes in point cloud analysis software to make the building facade parallel to the spatial axis plane. The orthographic projection image of the building facade is then exported. Image recognition software is used to intelligently segment the facade, and the RGB color values ​​of the facade color blocks are extracted. ) and the proportion of color block area S represents the area of ​​the building or region in the image. Let x represent the area of ​​different color blocks within the region, and let x represent the different color block numbers. The color blocks are sorted according to their area percentage to form the hue. High-value primary hues and Lower values ​​affect hue; Based on the optimized building boundary line, the building length is... ,width The building spacing can be obtained by directly reading the data and calculating the relative spacing between the building boundaries. .

3. The method for rural landscape management based on oblique photography and image recognition technology according to claim 1, characterized in that: The method for calculating the appearance evaluation value is as follows: Architectural style assessment value in: The highest rating is: HX= ; The roof form evaluation value is: if Then iX=1, if If the angle is 10°, then iX = 0; Color rating: CX= Where n is the area of ​​the color block. The sorting is performed, and the top 3 colored blocks are included in the calculation value, i.e., n = (1, 2, 3), where x represents the evaluated building or area. The R value refers to the color value of the nth color block in the sequence; similarly, the G and B values ​​are also relevant. The color mean of the nth ranked parameter in the landscape parameters; similarly... Value and value; The planar morphology evaluation value is: LWX= ; The evaluation value for building spacing is: .

Images