Road disease point stake number positioning method and system based on unmanned aerial vehicle

Abstract

The invention discloses a road disease point stake mark positioning method and system based on an unmanned aerial vehicle. The method comprises the following steps: constructing a road center line based on an orthoimage and a design file of a road; constructing a plurality of vertical offset lines along the center line of the road by taking each hundred stake mark point as an initial reference point, and distributing stake mark data of each vertical offset line; constructing rectangular buffer areas by taking each vertical offset line as a central axis, and combining all the rectangular buffer areas into a strip-shaped buffer area surface; performing grid division on the strip-shaped buffer area surface according to the resolution of the orthoimage, and distributing stake number data of each grid; aligning each grid of the strip-shaped buffer area surface with the orthoimage pixel by pixel, and fusing stake number data of each grid with image data of the corresponding pixel to generate corresponding fused grid data; and in combination with the image data of the road disease point and the fused raster data of each raster, searching the raster closest to the road disease point by adopting a KD tree, and accurately positioning the stake number of the road disease point according to the corresponding stake number data.

Description

Technical Field

[0001] The present invention relates to the technical field of determining the orientation of an object through image analysis, and in particular to a method and system for locating the pile number of a road defect point based on an unmanned aerial vehicle. Background Art

[0002] With the continuous expansion of highway traffic networks and the increase in service life, highway slope diseases have become an important factor affecting road safety and operational efficiency. Traditional disease detection methods mostly rely on manual inspections, which are not only inefficient but also difficult to fully cover. The development of modern computer vision and drone technology has provided a new way to automatically identify road diseases. However, the current system focuses on the collection and identification of disease images, and lacks the ability to efficiently convert the identification results into the pile number information required for actual maintenance work, resulting in inaccurate positioning of the disease location, affecting subsequent maintenance efficiency and cost control. Summary of the invention

[0003] In view of the shortcomings of the existing technology, the present invention proposes a method and system for locating the pile number of road damage points based on drones, which can accurately locate the pile number information of the damage location. The specific technical solution is as follows:

[0004] In a first aspect, a method for positioning a road damage point pile number based on an unmanned aerial vehicle is provided. In a first achievable manner of the first aspect, the method includes:

[0005] Construct road centerlines based on orthophotos and design files of the road;

[0006] Taking each whole-hundred pile number point on the road as the initial reference point, constructing a plurality of vertical offset lines along the center line of the road, and allocating pile number data of each vertical offset line according to the corresponding whole-hundred pile number;

[0007] Construct corresponding rectangular buffer zones with each vertical offset line as the central axis, and merge all rectangular buffer zones into a continuous strip buffer zone surface;

[0008] Divide the strip buffer surface into grids according to the resolution of the orthophoto, and allocate stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer;

[0009] Aligning each grid of the strip buffer surface with the orthophoto pixel by pixel, and fusing the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data;

[0010] Combining the image data of the road damage point and the fused grid data of each grid, the KD tree is used to search for the grid closest to the road damage point, and the pile number of the road damage point is determined according to its corresponding pile number data.

[0011] In combination with the first implementable manner of the first aspect, in a second implementable manner of the first aspect, constructing a road centerline based on an orthophoto of the road and a design file includes:

[0012] Taking the orthophoto as the base layer, the design file is superimposed and registered with the orthophoto through a plurality of registration points in the orthophoto, and the road centerline is constructed based on the registered road registration data.

[0013] In combination with the first implementable manner of the first aspect, in a third implementable manner of the first aspect, taking each hundredth pile number point on the road as an initial reference point includes:

[0014] The position coordinates of each hundred-pile point are corrected based on the actual measured data on site.

[0015] In combination with the first implementable manner of the first aspect, in a fourth implementable manner of the first aspect, a plurality of vertical offset lines are constructed along the center line of the road, including:

[0016] Taking each of the said whole pile number points as initial reference points, draw a perpendicular line perpendicular to the said road center line, and determine the intersection point of each perpendicular line with the road center line;

[0017] A plurality of incremental marking points are evenly arranged on the center line of the road between two adjacent intersections;

[0018] Create vertical offset lines perpendicular to the center line of the road with each incremental identification point as the reference point. The vertical offset lines cover the width of the road.

[0019] In combination with the first implementable manner of the first aspect, in a fifth implementable manner of the first aspect, merging all rectangular buffers into a continuous strip buffer surface includes:

[0020] The gap areas and overlapping areas between the rectangular buffers are detected and processed by automatically merging the gap areas and overlapping areas.

[0021] In combination with the first implementable manner of the first aspect, in a sixth implementable manner of the first aspect, generating corresponding fused raster data includes:

[0022] The stake number data of the grid is fused with the image data of the pixels corresponding to the grid through band combination.

[0023] In combination with the first implementable manner of the first aspect, a seventh implementable manner of the first aspect further includes:

[0024] Constructing a training set based on the fused grid data corresponding to each grid in the strip buffer surface;

[0025] Taking the image data as input and the stake number data as output, the constructed data regression model is trained through the training set;

[0026] When the KD tree fails to find the grid closest to the road damage point, the pile number of the road damage point is inferred by a trained data regression model based on the image data of the road damage point.

[0027] In combination with the seventh implementable manner of the first aspect, in an eighth implementable manner of the first aspect, the data regression model is a random forest regression model.

[0028] In combination with the seventh implementable manner of the first aspect, in a ninth implementable manner of the first aspect, the data regression model is constructed by combining a random forest regression model and an XGBoost model.

[0029] In the second aspect, a road damage point pile number positioning system is provided, comprising:

[0030] a centerline construction module configured to construct a road centerline based on an orthophoto and a design file of the road;

[0031] An offset line construction module is configured to construct a plurality of vertical offset lines along the center line of the road using each integer pile number point on the road as an initial reference point, and to allocate pile number data of each vertical offset line according to the corresponding integer pile number;

[0032] A buffer surface construction module is configured to construct corresponding rectangular buffers with each vertical offset line as a central axis, and merge all rectangular buffers into a continuous strip buffer surface;

[0033] A grid division module is configured to divide the strip buffer surface into grids according to the resolution of the orthophoto, and to allocate stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer;

[0034] A data fusion module is configured to align each grid of the strip buffer surface with the orthophoto pixel by pixel, and fuse the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data;

[0035] The pile number positioning module is configured to combine the image data of the road damage point and the fused grid data of each grid, use the KD tree to search the grid closest to the road damage point, and determine the pile number of the road damage point according to its corresponding pile number data.

[0036] Beneficial effects: The method and system for locating the pile number of road damage points based on drones of the present invention can obtain accurate and reliable fused grid data for the predefined buffer zone by fusing the road orthophoto obtained by field measurement with the road design data, providing a basis for accurately locating the pile number information of the damage points. Based on the obtained fused grid data and damage image data, the pile number of the damage point within the predefined buffer zone can be accurately located by using KD tree search, thereby reducing the time spent on locating road damage points, achieving faster data-driven intervention, and minimizing maintenance time. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the specific implementation of the present invention, the following will briefly introduce the drawings required for use in the specific implementation. In all the drawings, each element or part is not necessarily drawn according to the actual scale.

[0038] Figure 1 A flow chart of a method for locating a road damage point pile number based on an unmanned aerial vehicle according to an embodiment of the present invention;

[0039] Figure 2 A schematic diagram of a strip buffer surface provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following embodiments of the technical solution of the present invention are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and are therefore only used as examples, and cannot be used to limit the protection scope of the present invention.

[0041] like Figure 1 The flowchart of the method for positioning the pile number of road damage points based on UAV is shown, and the positioning method includes:

[0042] Step 1: Construct the road centerline based on the orthophoto image and design file of the road;

[0043] Step 2, taking each whole-hundred pile number point on the road as the initial reference point, constructing a plurality of vertical offset lines along the center line of the road, and allocating pile number data of each vertical offset line according to the corresponding whole-hundred pile number;

[0044] Step 3: construct corresponding rectangular buffer zones with each vertical offset line as the central axis, and merge all rectangular buffer zones into a continuous strip buffer zone surface;

[0045] Step 4: dividing the strip buffer surface into grids according to the resolution of the orthophoto, and allocating the stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer;

[0046] Step 5, aligning each grid of the strip buffer surface with the orthophoto pixel by pixel, and fusing the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data;

[0047] Step 6: Combine the image data of the road damage point and the fused grid data of each grid, use the KD tree to search for the grid closest to the road damage point, and determine the pile number of the road damage point based on its corresponding pile number data.

[0048] Specifically, first, the road centerline can be created by combining the road orthophoto image measured on site by drone and the road design file. Then, using each whole pile point as the initial reference point, multiple vertical offset lines perpendicular to the road centerline are constructed along the road centerline, and the pile information of the reference whole pile point is assigned to each vertical offset line.

[0049] After that, each vertical offset line is used as the central axis to create a corresponding rectangular buffer of equal size, and all the created rectangular buffers are merged into a strip buffer surface distributed along the road surface. Then, the strip buffer surface is grid-divided according to the image resolution of the orthophoto, and the stake information of the vertical offset line corresponding to the rectangular buffer where the grid is located is assigned to each grid. After that, all grids are aligned with each pixel of the orthophoto one by one through the set capture grid, so as to minimize the misalignment during data fusion and ensure the accuracy of the fused grid data obtained by fusing the stake data of each grid with the image data corresponding to the aligned pixels.

[0050] Finally, the existing image disease recognition method is used to identify the disease points in the orthophoto, and based on the image data of the disease point and the fused grid data of each grid, the KD tree is used to search the grid closest to the road disease point, and the pile number of the road disease point is determined according to its corresponding pile number data, thereby achieving high-precision positioning of the pile number to which the disease point belongs at the pixel level, thereby reducing the time spent on locating road disease points, achieving faster data-driven intervention, and minimizing maintenance time.

[0051] In this embodiment, optionally, in step 1, constructing a road centerline based on an orthophoto of the road and a design file includes:

[0052] Taking the orthophoto as the base layer, the design file is superimposed and registered with the orthophoto through a plurality of registration points in the orthophoto, and the road centerline is constructed based on the registered road registration data.

[0053] Specifically, when creating a road centerline, you can first import the orthophoto of the road as the main reference layer into the GIS software, and at the same time import the design file of the road and all the pile points into the GIS software, and overlay the orthophoto of the road and the pile points of the design file in the GIS software. Then, using the orthophoto as the base layer, the orthophoto, design file and pile points are aligned by presetting multiple registration points in the orthophoto to ensure the accuracy of the subsequent creation of the road centerline and vertical offset line.

[0054] After the layers are aligned, the road centerline can be created based on the aligned data. In this embodiment, the road centerline can be manually drawn using GIS software based on the aligned data. The deviation between the road centerline in the orthophoto and the road centerline in the design file can also be compared. If the deviation is within one pixel, the road centerline in the design file can be directly extracted and used, reducing the time to create the road centerline and improving efficiency.

[0055] In this embodiment, the registration points can be selected from important road features such as bends and intersections as the registration points, and these registration points need to be evenly distributed along the road as much as possible. At the same time, the acquired orthophoto can be appropriately cropped to exclude irrelevant areas far away from the road and slope.

[0056] In this embodiment, optionally, taking each hundred-pile point on the road as the initial reference point includes: correcting the position coordinates of each hundred-pile point according to the field measured data.

[0057] It should be understood that there may be differences between the pile numbers in the road design file and the actual road operation period. Therefore, the pile number information of each pile number can be corrected according to the accurate position of the pile number measured by high-precision RTK equipment on site to ensure that the vertical offset line created later is consistent with the actual situation of the road and ensure the accuracy of the pile number positioning of the disease point.

[0058] In this embodiment, optionally, in step 2, a plurality of vertical offset lines are constructed along the road center line, including:

[0059] Taking each of the said whole pile number points as initial reference points, draw a perpendicular line perpendicular to the said road center line, and determine the intersection point of each perpendicular line with the road center line;

[0060] A plurality of incremental marking points are evenly arranged on the road center line between two adjacent intersections;

[0061] Create vertical offset lines perpendicular to the center line of the road with each incremental identification point as the reference point. The vertical offset lines cover the width of the road.

[0062] Specifically, if Figure 2 As shown in the figure, when creating a vertical offset line, first use each hundredth pile point as the initial reference point, draw a vertical line perpendicular to the road centerline, and use GIS software to determine the intersection between the vertical line and the road centerline as the vertical mark point. These vertical mark points will mark the exact hundred-meter interval on the road centerline, and the error will not be transmitted, thereby preventing the error transmission typical of long linear features of the road, thereby maintaining the accuracy of the road section pile number. And these intersections inherit the pile number information of the corresponding pile number, which includes prefix letters, kilometers, and meters, such as K001+900.

[0063] Then, 99 incremental identification points can be created on the center line of the road between two adjacent vertical marking points, and each incremental identification point represents an increment of 1 meter. Then, these 99 incremental identification points are assigned in order along the increasing direction of the road stake number value, and are assigned the smaller prefix letter of the two vertical marking points and the kilometer value. For example, the 99 incremental identification points generated between the two vertical marking points K001+900 and K002+000 are K001+901, K001+902...K001+999.

[0064] Finally, each incremental identification point is used as a reference point to create a vertical offset line perpendicular to the center line of the road. The two ends of the vertical offset line extend to both sides of the road to form a vertical line segment covering the width of the road. After the vertical offset line is created, the attribute value of each vertical offset line can be determined according to the assignment of the incremental identification point and the driving direction. The attribute value includes the pile number information and the driving direction, such as K001+901 for the upward direction.

[0065] In this embodiment, optionally, in step 3, all rectangular buffers are merged into a continuous strip buffer surface, including:

[0066] The gap areas and overlapping areas between the rectangular buffers are detected and processed by automatically merging the gap areas and overlapping areas.

[0067] Specifically, after creating all vertical offset lines, you can use each vertical offset line as the center axis to create a rectangular buffer zone that is perpendicular to the center line of the road and of the same size. The length of the rectangular buffer zone can be set to the road width, and the width can be set to 1 meter. Finally, merge all the created rectangular buffer zones to generate the entire strip buffer zone surface.

[0068] During the creation process, two adjacent rectangular buffers may have overlapping and gap parts due to road curvature or misalignment. Therefore, the topology tools in the GIS software can be used to find the gap and overlapping areas between adjacent rectangular buffers, and compare the width of the gap or overlapping area with the set tolerance threshold range. If it exceeds the tolerance threshold range, the gap and overlapping areas are processed using the joint automatic merge to form a strip buffer surface with no redundant overlap and high consistency with the road route to ensure the accuracy of the subsequent fusion of raster data.

[0069] In this embodiment, optionally, in step 5, generating corresponding fused raster data includes:

[0070] The stake number data of the grid is fused with the image data of the pixels corresponding to the grid through band combination.

[0071] Specifically, after constructing the strip buffer area surface, the strip buffer area surface can be converted into a strip buffer raster map according to the resolution of the orthophoto of the road, so as to realize the pixel-level fusion of image data and pile number data later, thereby improving the accuracy of the fused raster data.

[0072] After converting the strip cache area surface into a strip buffer raster map, you can set a snap grid in it, and use the snap grid to align each raster in the strip buffer raster map with each pixel in the orthophoto, so as to minimize the misalignment when fusing data and further improve the accuracy of the fused raster data. After aligning the strip buffer raster map with the orthophoto pixel by pixel, you can use the band combination function in the GIS software to fuse the stake information of the raster in the strip buffer raster map with the image data of the aligned pixels in the orthophoto, and generate accurate and reliable fused raster data, which provides a basis for accurately locating the stake signal of the disease point.

[0073] The fused raster data includes not only the raster's geospatial information and R, G, B information, but also the pile number information. The existing disease image recognition and positioning method can be used to locate the geospatial coordinates of the road disease point in the orthophoto. Based on the geospatial coordinates, the KD tree search algorithm can be used to search for the grid closest to the road disease point from a large number of grids, and the pile number of the road disease point can be accurately located based on the pile number information in the fused raster data of the grid. This achieves high-precision positioning of the pile number to which the disease point belongs, thereby reducing the time spent on locating the road disease point, achieving faster data-driven intervention, and minimizing maintenance time.

[0074] In this embodiment, optionally, it further includes:

[0075] Constructing a training set based on the fused grid data corresponding to each grid in the strip buffer surface;

[0076] Taking the image data as input and the stake number data as output, the constructed data regression model is trained through the training set;

[0077] When the KD tree fails to find the grid closest to the road damage point, the pile number of the road damage point is inferred by a trained data regression model based on the image data of the road damage point.

[0078] It should be understood that due to the limited scope of the predefined strip buffer surface, the geographic space of the road damage points identified by the orthophoto may exceed the scope of the strip buffer surface. For example, the identified damage points are not on the road surface, but on the roadbed slope, or the searched grid does not have the corresponding pile number information.

[0079] To this end, a training set can be constructed based on the fused grid data corresponding to each grid in the strip buffer surface, and the constructed data regression model can be trained through the training set. The data regression model uses geographic spatial information and R, G, B information as input and pile number information as output. During the training process, the random search cross-validation method can be used to adjust the model parameters so that the model's determination coefficient exceeds 0.9 and the mean absolute error is less than 1 meter.

[0080] When the distance between the grid closest to the road damage point and the road damage point found by the KD tree search exceeds a set threshold, the pile number of the road damage point can be inferred by the trained data regression model based on the image data of the road damage point. In this embodiment, the threshold can be set to 1 pixel.

[0081] In this embodiment, the data regression model is optionally a random forest regression model. Specifically, the random forest regression model is relatively balanced between interpretability, accuracy and computational efficiency. Therefore, in this embodiment, the random forest regression model is selected to calculate the pile number information of the road damage point.

[0082] In this embodiment, optionally, the data regression model is constructed by combining a random forest regression model and an XGBoost model.

[0083] Specifically, in order to improve the overall accuracy of the model, the random forest regression model and the XGBoost model can be combined to construct the data regression model, thereby utilizing the respective advantages of the two models to improve the accuracy and stability of the calculation of the pile number information of the road damage points.

[0084] Specifically, the stacking ensemble method can be used to combine the random forest regression model and the XGBoost model in a hierarchical manner, and the predictions of one model are input as initial features into the other model. Alternatively, the prediction results of the random forest regression model and XGBoost are mixed, and their predictions are averaged or weighted according to the performance of the two models.

[0085] A road damage point pile number positioning system, comprising:

[0086] a centerline construction module configured to construct a road centerline based on an orthophoto and a design file of the road;

[0087] An offset line construction module is configured to construct a plurality of vertical offset lines along the center line of the road using each integer pile number point on the road as an initial reference point, and to allocate pile number data of each vertical offset line according to the corresponding integer pile number;

[0088] A buffer surface construction module is configured to construct corresponding rectangular buffers with each vertical offset line as a central axis, and merge all rectangular buffers into a continuous strip buffer surface;

[0089] A grid division module is configured to divide the strip buffer surface into grids according to the resolution of the orthophoto, and to allocate stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer;

[0090] A data fusion module is configured to align each grid of the strip buffer surface with the orthophoto pixel by pixel, and fuse the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data;

[0091] The pile number positioning module is configured to combine the image data of the road damage point and the fused grid data of each grid, use the KD tree to search the grid closest to the road damage point, and determine the pile number of the road damage point according to its corresponding pile number data.

[0092] Specifically, the positioning system includes a centerline construction module, an offset line construction module, a buffer surface construction module, a grid division module, a data fusion module and a pile number positioning module. Among them, the centerline construction module can create a road centerline by combining the design documents of the road and the measured orthophoto. The offset line construction module can use each integer pile number point as the initial reference point, construct multiple vertical offset lines perpendicular to the road centerline along the created road centerline, and assign the pile number information of the reference integer pile number point to each vertical offset line.

[0093] The buffer surface construction module can create corresponding rectangular buffers of equal size with each vertical offset line as the central axis, and merge all the created rectangular buffers into an overall strip buffer surface. The grid division module can divide the strip buffer surface into grids according to the image resolution of the orthophoto, and assign the pile number information of the whole pile number point referenced by the vertical offset line corresponding to the rectangular buffer where the grid is located to each grid. The data fusion module can align all grids with each pixel of the orthophoto one by one through the set capture grid, thereby minimizing the misalignment during data fusion and ensuring the accuracy of the fused grid data after the pile number data of each grid is fused with the image data corresponding to the aligned pixels.

[0094] The pile number positioning module can use the existing image disease recognition method to identify the disease points in the orthophoto, and based on the image data of the disease point and the fused grid data of each grid, use the KD tree to search for the grid closest to the road disease point, and determine the pile number of the road disease point according to its corresponding pile number data, thereby achieving high-precision positioning of the pile number to which the disease point belongs, thereby reducing the time spent on locating road disease points, achieving faster data-driven intervention, and minimizing maintenance time.

[0095] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A method for locating the pile number of road damage points based on drones, characterized in that: include: Construct road centerlines based on orthophotos and design files of the road; Taking each whole-hundred pile number point on the road as the initial reference point, constructing a plurality of vertical offset lines along the center line of the road, and allocating pile number data of each vertical offset line according to the corresponding whole-hundred pile number; Construct corresponding rectangular buffer zones with each vertical offset line as the central axis, and merge all rectangular buffer zones into a continuous strip buffer zone surface; Divide the strip buffer surface into grids according to the resolution of the orthophoto, and allocate stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer; Aligning each grid of the strip buffer surface with the orthophoto pixel by pixel, and fusing the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data; Combining the image data of the road damage point and the fused grid data of each grid, the KD tree is used to search for the grid closest to the road damage point, and the pile number of the road damage point is determined according to its corresponding pile number data.

2. The method for locating road damage points based on drones according to claim 1 is characterized in that: Construct road centerlines based on orthophotos and design files of the road, including: Taking the orthophoto as the base layer, the design file is superimposed and registered with the orthophoto through multiple registration points in the orthophoto, and the road centerline is constructed based on the registered road registration data.

3. The method for locating road damage points based on drones according to claim 1 is characterized in that: The initial reference points are the points with the whole number of piles on the road, including: The position coordinates of each hundred-pile point are corrected based on the actual measured data on site.

4. The method for locating road damage points based on drones according to claim 1 is characterized in that: A plurality of vertical offset lines are constructed along the center line of the road, including: Taking each of the said whole pile number points as initial reference points, draw a perpendicular line perpendicular to the said road center line, and determine the intersection point of each perpendicular line with the road center line; A plurality of incremental marking points are evenly arranged on the road center line between two adjacent intersections; Create vertical offset lines perpendicular to the center line of the road with each incremental identification point as the reference point. The vertical offset lines cover the width of the road.

5. The method for locating road damage points based on drones according to claim 1 is characterized in that: Merge all rectangular buffers into a continuous strip of buffer polygons, including: The gap areas and overlapping areas between the rectangular buffers are detected and processed by automatically merging the gap areas and overlapping areas.

6. The method for locating road damage points based on drones according to claim 1 is characterized in that: Generate the corresponding fused raster data, including: The stake number data of the grid is fused with the image data of the pixels corresponding to the grid through band combination.

7. The method for locating road damage points based on drones according to claim 1 is characterized in that: Also includes: Constructing a training set based on the fused grid data corresponding to each grid in the strip buffer surface; Taking the image data as input and the stake number data as output, the constructed data regression model is trained through the training set; When the KD tree fails to find the grid closest to the road damage point, the pile number of the road damage point is inferred by a trained data regression model based on the image data of the road damage point.

8. The method for locating road damage points based on drones according to claim 7 is characterized in that: The data regression model is a random forest regression model.

9. The method for locating road damage points based on drones according to claim 7 is characterized in that: The data regression model is constructed by combining the random forest regression model and the XGBoost model.

10. A road damage point pile number positioning system, characterized in that: include: a centerline construction module configured to construct a road centerline based on an orthophoto and a design file of the road; An offset line construction module is configured to construct a plurality of vertical offset lines along the center line of the road using each integer pile number point on the road as an initial reference point, and to allocate pile number data of each vertical offset line according to the corresponding integer pile number; A buffer surface construction module is configured to construct corresponding rectangular buffers with each vertical offset line as a central axis, and merge all rectangular buffers into a continuous strip buffer surface; A grid division module is configured to divide the strip buffer surface into grids according to the resolution of the orthophoto, and to allocate stake number data of each grid according to the vertical offset line corresponding to the rectangular buffer; A data fusion module is configured to align each grid of the strip buffer surface with the orthophoto pixel by pixel, and fuse the stake number data of each grid with the image data of the corresponding pixel to generate corresponding fused grid data; The pile number positioning module is configured to combine the image data of the road damage point and the fused grid data of each grid, use the KD tree to search the grid closest to the road damage point, and determine the pile number of the road damage point according to its corresponding pile number data.

Images