An unmanned aerial vehicle automatic aerial survey method based on a Beidou grid code and an AI image control point

Abstract

The application discloses an unmanned aerial vehicle automatic aerial survey method based on a Beidou grid code and an AI image control point, and comprises the following steps: S1, a Beidou grid code is used to divide a grid of an unmanned aerial vehicle aerial survey area, and an image control point is arranged in the divided grid; S2, aerial survey is performed when the unmanned aerial vehicle is located on a preset route, and image data of the aerial survey area is obtained; S3, the image control point of the aerial survey area is recognized according to the image data; and S4, the Beidou grid code of the image control point in the aerial survey area is determined. The application solves the problems of non-uniform coordinates and large matching errors in traditional positioning from the root, and provides a solid foundation for subsequent integration and splicing of aerial survey data.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, specifically relating to an automatic aerial surveying method for unmanned aerial vehicles based on Beidou grid codes and AI image control points. Background Technology

[0002] In the field of UAV aerial surveying, image control points (ARPCs) are key references for aerial surveying work such as densification, mapping, and field surveying. Their accurate deployment and identification play a decisive role in the accuracy of aerial survey results.

[0003] In traditional aerial surveying operations, the placement of control points mainly relies on manual on-site selection. Operators need to bring specialized equipment to the site and determine the location of control points based on experience. This process is time-consuming and labor-intensive, especially in aerial surveying areas with complex terrain and vast areas, where the difficulty and cost of manual placement increase dramatically.

[0004] In the control point identification process, traditional methods rely on manual searching and marking of control points one by one in aerial survey images. This is not only inefficient, but also prone to errors or omissions in control point identification due to factors such as visual fatigue and subjective judgment differences among operators, which seriously restricts the overall progress and quality of aerial survey work.

[0005] In addition, traditional aerial surveying lacks efficient spatial positioning and regional division methods. When conducting aerial surveys over large areas, spatial matching and unified management of control points between aerial survey images from different segments and batches face many difficulties, making it difficult to form an accurate and coherent spatial reference system, which further affects the accuracy and usability of aerial survey results.

[0006] With the continuous development and improvement of the BeiDou Navigation Satellite System, the BeiDou grid code, with its ability to accurately and regularly divide the Earth's surface into grids, provides a new and unified spatial location coding method for aerial survey areas. This enables refined and standardized spatial division of the survey area, providing precise spatial coordinate references and positioning frameworks for ground control points (GCPs). Meanwhile, artificial intelligence (AI) technology has made breakthroughs in computer vision, image recognition, and other fields, possessing powerful pattern recognition and automatic analysis capabilities, enabling efficient and accurate identification of GCPs from massive amounts of aerial survey imagery data.

[0007] Based on the shortcomings of traditional aerial surveying methods and the advantages of BeiDou grid codes, this paper proposes to combine the spatial division capabilities of BeiDou grid codes to achieve automatic identification of control points in UAV aerial surveying. Summary of the Invention

[0008] To address the above problems, this invention proposes an automatic aerial surveying method for unmanned aerial vehicles based on BeiDou grid codes and AI image control points.

[0009] The technical solution of this invention is: an automatic aerial surveying method for unmanned aerial vehicles based on Beidou grid codes and AI image control points, comprising the following steps:

[0010] S1. Use the BeiDou grid code to divide the UAV aerial survey area into grids, and set up image control points in the divided grids;

[0011] S2. Conduct aerial surveys while the UAV is on a preset flight path to obtain image data of the survey area;

[0012] S3. Identify the image control points in the aerial survey area based on the image data;

[0013] S4. Determine the BeiDou grid code of the control point in the aerial survey area.

[0014] Furthermore, S3 includes the following sub-steps:

[0015] S31. Extract the average color value of all pixels in the image data and use it as the threshold for dividing the image data.

[0016] S32. Input the difference between the color value of each pixel and the division threshold into the sign function to obtain the calculation result, and determine the distribution coefficient of each pixel based on the calculation result;

[0017] S33. Use the exponential function to process the distribution coefficient of the pixel to obtain the local transformation value of the pixel in the image data.

[0018] S34. Calculate the image control suspicion of a pixel based on the local transformation value of the pixel in the image data;

[0019] S35. Identify control points in the aerial survey area based on the control point suspicion of all pixels.

[0020] The beneficial effects of the above-mentioned further solutions are as follows: In this invention, the segmentation threshold acts as a baseline, used to distinguish whether a pixel's color is too bright or too dark. The distribution coefficient reflects the distribution characteristics of pixels in the image. By processing the distribution coefficient of pixels using an exponential function, the local transformation value of the pixel in the image data is obtained. This local transformation value can highlight the characteristics of the pixel in its surrounding local area.

[0021] Furthermore, in S32, the distribution coefficient of the pixels The expression is:

[0022] ;

[0023] in, Represents the color value of a pixel. This indicates the threshold for dividing image data. Represents an exponential function. This represents the result of the operation on the pixel. It represents the standard deviation of the color values ​​of all pixels in the image data.

[0024] Furthermore, S33 includes the following sub-steps:

[0025] S331. Calculate the ratio between the distribution coefficient of a pixel and the mean distribution coefficient of its four neighboring pixels;

[0026] S332. Perform a logarithmic operation to the base 10 on the comparison result, then perform a square operation, and use the square result as the numerator;

[0027] S333. Squaring the standard deviation of the distribution coefficients of all pixels and using the squaring result as the denominator.

[0028] S334. Take the opposite of the ratio of the numerator to the denominator and input it into the exponential function to obtain the local transformation value of the pixel in the image data.

[0029] The beneficial effects of the above-mentioned further solutions are as follows: In this invention, a four-neighbor pixel refers to the four adjacent pixels above, below, left, and right of a given pixel. The calculated ratio reflects the degree of difference between the distribution coefficient of that pixel and the average distribution coefficient of its surrounding pixels. Logarithmic operations can scale the numerical differences to a certain extent, highlighting the different characteristics of that pixel compared to its surrounding pixels. The standard deviation reflects the dispersion of the distribution coefficient. The properties of the exponential function transform the negative number to obtain the local transformation value of the pixel in the image data. This local transformation value can better reflect the uniqueness of the pixel in a local region.

[0030] Furthermore, S34 includes the following sub-steps:

[0031] S341. Construct the overall matrix of pixels using the color values ​​of the pixels and the color values ​​of their 8 neighboring pixels;

[0032] S342. Construct several row vectors and column vectors for the image data;

[0033] S343. Calculate the image control resemblance of a pixel in the image data based on the local transformation value of the pixel, the overall matrix, the row vector of the pixel's row, and the column vector of the pixel's column.

[0034] The beneficial effects of the above-mentioned further scheme are as follows: In this invention, the overall matrix encompasses the color information of the pixel and its 8-neighborhood, extending the analysis of the pixel from a single point to a local region, which is more in line with the characteristic that control points are usually points with obvious local regional features. The row vectors and column vectors respectively reflect the color changes in the horizontal and vertical directions of the image. The color changes around control points usually have their own particularities, and these vectors can capture these particularities, providing directional features for the calculation of control suspicion. By comprehensively calculating the control suspicion from local transformation values, the overall matrix, row vectors, and column vectors, the pixel is evaluated from multiple perspectives, avoiding the one-sidedness of judging a single feature, and making the calculation of control suspicion more accurate.

[0035] Further, in S342, the color value difference between every two adjacent pixels in each row is used as an element of the row vector; and the color value difference between every two adjacent pixels in each column is used as an element of the column vector.

[0036] Furthermore, in S343, the pixel's image control accuracy in the image data... The expression is:

[0037] ;

[0038] in, This represents the local transformation value of a pixel in the image data. This represents the row vector of the row containing the pixel. This represents the column vector of the column containing the pixel. A matrix representing the total number of pixels. This represents the matrix rank operation.

[0039] The beneficial effect of the above-mentioned further scheme is that, in this invention, the characteristics of whether a pixel is a control point are described from different mathematical dimensions by using the matrix rank and 0 norm. The row and column vectors where the control point is located often have a certain degree of sparsity.

[0040] Furthermore, in S343, pixels with a control suspicion degree less than the standard deviation of all control suspicion degrees are designated as control points.

[0041] The beneficial effects of this invention are as follows: Through multi-dimensional feature analysis, including distribution coefficients, local transformation values, and control point suspicion, this invention effectively distinguishes control points from interference pixels and determines the final result based on the characteristics of the BeiDou grid code. The accurately determined control points provide a high-quality control benchmark for aerial survey data processing and facilitate the secondary application of aerial survey results. At the same time, it fundamentally solves the problems of inconsistent coordinates and large matching errors in traditional positioning, providing a solid foundation for the subsequent integration and stitching of aerial survey data. Attached Figure Description

[0042] Figure 1 This is a flowchart of an automated aerial surveying method for unmanned aerial vehicles based on BeiDou grid codes and AI image control points. Detailed Implementation

[0043] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0044] like Figure 1 As shown, this invention provides an automatic aerial surveying method for unmanned aerial vehicles (UAVs) based on BeiDou grid codes and AI image control points, including the following steps:

[0045] S1. Use the BeiDou grid code to divide the UAV aerial survey area into grids, and set up image control points in the divided grids;

[0046] S2. Conduct aerial surveys while the UAV is on a preset flight path to obtain image data of the survey area;

[0047] S3. Identify the image control points in the aerial survey area based on the image data;

[0048] S4. Determine the BeiDou grid code of the control point in the aerial survey area.

[0049] In this embodiment of the invention, S3 includes the following sub-steps:

[0050] S31. Extract the average color value of all pixels in the image data and use it as the threshold for dividing the image data.

[0051] S32. Input the difference between the color value of each pixel and the division threshold into the sign function to obtain the calculation result, and determine the distribution coefficient of each pixel based on the calculation result;

[0052] S33. Use the exponential function to process the distribution coefficient of the pixel to obtain the local transformation value of the pixel in the image data.

[0053] S34. Calculate the image control suspicion of a pixel based on the local transformation value of the pixel in the image data;

[0054] S35. Identify control points in the aerial survey area based on the control point suspicion of all pixels.

[0055] In this invention, the segmentation threshold acts as a baseline, used to distinguish whether a pixel's color is too bright or too dark. The distribution coefficient reflects the distribution characteristics of pixels in the image. By processing the pixel's distribution coefficient using an exponential function, the local transformation value of the pixel in the image data is obtained. This local transformation value can highlight the characteristics of the pixel in its surrounding local area.

[0056] In this embodiment of the invention, in S32, the distribution coefficient of the pixel points The expression is:

[0057] ;

[0058] in, Represents the color value of a pixel. This indicates the threshold for dividing image data. Represents an exponential function. This represents the result of the operation on the pixel. It represents the standard deviation of the color values ​​of all pixels in the image data.

[0059] In this embodiment of the invention, S33 includes the following sub-steps:

[0060] S331. Calculate the ratio between the distribution coefficient of a pixel and the mean distribution coefficient of its four neighboring pixels;

[0061] S332. Perform a logarithmic operation to the base 10 on the comparison result, then perform a square operation, and use the square result as the numerator;

[0062] S333. Squaring the standard deviation of the distribution coefficients of all pixels and using the squaring result as the denominator.

[0063] S334. Take the opposite of the ratio of the numerator to the denominator and input it into the exponential function to obtain the local transformation value of the pixel in the image data.

[0064] In this invention, a four-neighbor pixel refers to the four pixels adjacent to a given pixel (top, bottom, left, and right). The calculated ratio reflects the difference between the distribution coefficient of this pixel and the average distribution coefficient of its surrounding pixels. Logarithmic operations can scale the numerical difference to a certain extent, highlighting the unique characteristics of this pixel compared to its neighbors. The standard deviation reflects the dispersion of the distribution coefficient. The exponential function transforms the negative value, obtaining a local transformation value for the pixel in the image data. This local transformation value better reflects the uniqueness of the pixel in its local region.

[0065] In this embodiment of the invention, S34 includes the following sub-steps:

[0066] S341. Construct the overall matrix of pixels using the color values ​​of the pixels and the color values ​​of their 8 neighboring pixels;

[0067] S342. Construct several row vectors and column vectors for the image data;

[0068] S343. Calculate the image control resemblance of a pixel in the image data based on the local transformation value of the pixel, the overall matrix, the row vector of the pixel's row, and the column vector of the pixel's column.

[0069] In this invention, the overall matrix encompasses the color information of a pixel and its eight neighbors, extending pixel analysis from a single point to a local area. This aligns better with the characteristic that control points are typically points with distinct local features. Row and column vectors represent color variations in the horizontal and vertical directions of the image, respectively. Color variations around control points often exhibit unique characteristics, and these vectors capture these characteristics, providing directional features for control suspicion calculation. By comprehensively calculating control suspicion using local transformation values, the overall matrix, row vectors, and column vectors, pixels are evaluated from multiple perspectives, avoiding the limitations of single-feature judgments and making control suspicion calculation more accurate.

[0070] In this embodiment of the invention, S342, the color value difference between every two adjacent pixels in each row is used as an element of the row vector; the color value difference between every two adjacent pixels in each column is used as an element of the column vector.

[0071] In this embodiment of the invention, in S343, the pixel's image control accuracy in the image data is... The expression is:

[0072] ;

[0073] in, This represents the local transformation value of a pixel in the image data. This represents the row vector of the row containing the pixel. This represents the column vector of the column containing the pixel. A matrix representing the total number of pixels. This represents the matrix rank operation.

[0074] In this invention, the rank of the matrix and the zero norm are used to characterize whether a pixel is a control point from different mathematical dimensions. The row and column vectors where the control point is located often have a certain degree of sparsity.

[0075] In this embodiment of the invention, in S343, pixels with a control suspicion degree less than the standard deviation of all control suspicion degrees are designated as control points.

[0076] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. An unmanned aerial vehicle automatic aerial surveying method based on a Beidou grid code and a control point, characterized in that, The method comprises the following steps: S1, dividing the UAV aerial survey area into grids based on the Beidou grid code, arranging control points in the divided grids, and recording coordinate information; S2, aerial survey when the UAV is on the preset route to obtain image data of the aerial survey area; S3, searching and identifying the control points of the aerial survey area according to the image data; S4, determining the Beidou grid code and coordinate information of the control points in the aerial survey area to realize automatic mapping; The S3 comprises the following sub-steps: S31, extracting the average color value of all pixel points in the image data as the division threshold of the image data; S32, inputting the difference between the color value of each pixel point and the division threshold into the sign function to obtain an operation result, and determining the distribution coefficient of each pixel point according to the operation result; S33, processing the distribution coefficient of the pixel point by using the exponential function to obtain the local transformation value of the pixel point in the image data; S34, calculating the control point suspiciousness of the pixel point according to the local transformation value of the pixel point in the image data; S35, identifying the control points of the aerial survey area according to the control point suspiciousness of all pixel points; In the S32, the distribution coefficient of the pixel point The expression is: ； wherein, represents a color value of a pixel point, represents a division threshold of image data, represents an exponential function, represents an operation result of a pixel point, represents a standard deviation of color values of all pixel points in image data; The S34 comprises the following sub-steps: S341, constructing the overall matrix of the pixel point by using the color value of the pixel point and the 8-neighbor color value; S342, constructing a plurality of row vectors and column vectors of the image data; S343, calculating the control point suspiciousness of the pixel point in the image data according to the local transformation value of the pixel point, the overall matrix, the row vector of the row and the column vector of the column; In the S343, the pixel point is suspicious in the image data The expression is: ； wherein, represents a local transformation value of the pixel point in the image data, represents a row vector of the row where the pixel point is located, represents a column vector of the column where the pixel point is located, represents an overall matrix of the pixel point, represents a matrix rank operation. 2.The method of claim 1, wherein, The S33 comprises the following sub-steps: S331, calculating the ratio between the distribution coefficient of the pixel point and the average value of the distribution coefficients of the four-neighbor pixel points; S332, performing logarithmic operation on the ratio result with base 10 and then performing square operation, and taking the square result as the numerator; S333, performing square operation on the standard deviation of the distribution coefficients of all pixel points, and taking the square result as the denominator; S334, taking the reciprocal of the ratio of the numerator and the denominator, inputting it into the exponential function to obtain the local transformation value of the pixel point in the image data. 3.The method of claim 1, wherein, In the S342, the color value difference between each two adjacent pixel points in each row is taken as an element of the row vector, and the color value difference between each two adjacent pixel points in each column is taken as an element of the column vector.

4. The automatic aerial photogrammetry method based on Beidou grid code and image control points of claim 1, wherein, In the S343, the pixel point with a control point suspiciousness less than the standard deviation of all control point suspiciousness is taken as the control point.

Images