A Dual-Point Location Method for Multi-Perspective Oblique Images

Abstract

The present invention relates to a multi-view oblique image double-point positioning method, which mainly includes the following steps: 1) Using an unmanned aerial vehicle (UAV) to obtain at least six images that can simultaneously observe two object points; 2) Obtaining the image positions through the GPS of the UAV; 3) Obtaining the image point coordinates of the two object points in the UAV images; 4) Substituting the image positions, image point coordinates and camera internal parameters into the double-point positioning algorithm to obtain the object point coordinates. This method avoids the serious influence of the IMU angle on the positioning accuracy and the influence of inaccurate exterior orientation elements obtained by aerial triangulation of images under uncontrolled conditions on the positioning accuracy, and effectively improves the positioning accuracy of object points.

Description

Technical Field

[0001] The present invention relates to a method for object point positioning, and particularly to a dual-point positioning method for multi-view oblique images. Background Art

[0002] In the field of mapping, it is widely involved in solving the object coordinates of ground points. The determination of object coordinates plays an important role in aspects such as natural resource management and monitoring, road monitoring, and topographic map updating. Existing methods for solving object point coordinates include the two-image forward intersection method and the multi-image forward intersection method.

[0003] The two-image forward intersection method uses two images for forward intersection and adopts an adjustment method to solve the object point coordinates. However, the accuracy of the obtained ground object point coordinates is still not very high. The main reasons are that the image quality in the two-image forward intersection is not high, resulting in errors in image point coordinates, and at the same time, the size of the intersection angle will also affect the accuracy of object coordinates.

[0004] The multi-image forward intersection method, that is, the multi-baseline photogrammetric forward intersection method, can generally be divided into the bundle adjustment forward intersection method and the linear forward intersection method. The equation of the linear forward intersection method is a linear equation, which does not require linearization and does not need to provide initial values, and the solution speed is relatively fast. However, the accuracy of the obtained object points is not high. The bundle adjustment forward intersection method uses the object point coordinates obtained by the two-image forward intersection as the initial values, takes the collinearity equation as the basic equation and linearizes it, and uses the least squares iteration to solve and finally obtain high-precision object measurement coordinates. However, the positioning accuracy seriously depends on the accuracy of the exterior orientation elements, and the accuracy of this method is difficult to guarantee without control conditions. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a dual-point positioning method for multi-view oblique images. Through this method, the object point coordinates can be solved without the angle of an inertial measurement unit (IMU) and without control conditions to obtain high-precision object point coordinates.

[0006] To achieve the above object, the present invention relates to a dual-point positioning method for multi-view oblique images. It mainly includes the following steps: 1) Using a drone to obtain at least 6 images that can simultaneously observe two object points; 2) Obtaining the image positions through the GPS of the drone; 3) Obtaining the image point coordinates of the two object points in the drone images; 4) Substituting the image positions, image point coordinates, and camera internal parameters into the dual-point positioning algorithm to obtain the object point coordinates.

[0007] The present invention proposes a new method for dual-point positioning of multi-view oblique images, avoiding the serious influence of the IMU angle on the positioning accuracy and the influence of inaccurate exterior orientation elements obtained by aerial triangulation of images without control conditions on the positioning accuracy, and effectively improving the object point positioning accuracy.

[0008] Specifically, the method of the present invention includes the following steps:

[0009] (1) Use a drone to obtain at least 6 images that can simultaneously observe two object points.

[0010] (2) Obtain the image positions through the GPS of the drone.

[0011] (3) Obtain the image point coordinates of the two object points in the drone images.

[0012] (4) Substitute the image positions, image point coordinates, and camera internal parameters into the double-point positioning algorithm to obtain the object point coordinates.

[0013] In the present invention, the images obtained by the drone should simultaneously observe two object points and the number of images should be no less than 6.

[0014] In the present invention, the image positions recorded when the drone obtains images, the corresponding image point coordinates of the object points in the images, and the camera internal parameters are substituted into the double-point positioning algorithm to obtain the object point coordinates. The calculation formula used in the double-point positioning algorithm is:

[0015]

[0016] In the formula, O is the projection center of the image, a and b are the projection points of the object points on the image, and ∠AOB is the observation angle.

[0017]

[0018] In the formula, O is the projection center of the image, A and B are the two object points observed by the image, and ∠AOB is the observation angle.

[0019] The observation angle ∠AOB is first obtained through the image point coordinates and image positions using formula (1); substituting into formula (2), the object points A and B are unknown, and there are a total of 6 unknowns. The coordinates of the object points A and B are jointly solved through at least 6 observation images.

[0020] In the formula proposed by the present invention, the observation angle can be obtained through the image point coordinates and image positions. The object points A and B are unknown, and there are a total of 6 unknowns. The coordinates of the object points A and B are jointly solved through at least 6 observation images. In the method of the present invention, the IMU angle is not involved in the calculation, avoiding the serious influence of the IMU angle on the positioning accuracy; there is no need to perform aerial triangulation to solve the exterior orientation elements, avoiding the influence of inaccurate exterior orientation elements obtained by aerial triangulation of images under uncontrolled conditions on the positioning accuracy; the object point coordinates can be obtained more simply and efficiently, and it is applicable to more situations at the same time. Description of the Drawings

[0021] Figure 1It is a flow chart for dual - point positioning of multi - perspective oblique images;

[0022] Figure 2 It is a schematic diagram of the principle of dual - point positioning. Specific implementation manner

[0023] This embodiment is a method for dual - point positioning of multi - perspective oblique images based on the dual - point positioning algorithm under a simulation experiment.

[0024] The simulation experiment means that taking the ground point coordinates (0, 0, 0) as the center point, setting the altitude of the UAV to 100 meters and the radius from the center point to 100 meters, and the UAV orbits around the center point for shooting. The set target ground point coordinates are point 1 (37.8937381963, 0, 0) and point 2 (-37.8937381963, 0, 0). The imaging size of the UAV is 4000×4000, the focal length is 3464.10162, and the offset of the principal point of the image is 2000, where the unit is pixel. The tilt angle of the UAV is the same as the photography angle, taking 45 degrees as the standard simulation experiment, and a set of UAV image data is obtained through simulation. Here, the number of UAV images is set to 8. The attitude and position of the UAV are known. By constructing the projection matrix P, the corresponding relationship between the ground point and the image point can be obtained, and the image point coordinates corresponding to the ground point can be acquired. The image positions of the 8 images, the image point coordinates corresponding to the ground points, and the internal parameters of the camera are known.

[0025] The specific implementation process of this embodiment is as follows:

[0026] Step 1, obtain UAV images, camera internal parameters, and image positions

[0027] Use the UAV to orbit and shoot two object - space points, set the altitude of the UAV to 100 meters and the radius from the center point to 100 meters, and the tilt angle of the UAV is 45 degrees, obtaining 8 images containing the two object - space points. The UAV is equipped with GPS to obtain the image positions. The imaging size of the UAV is 4000×4000, and the camera internal parameters are obtained by consulting the UAV parameters, with the focal length being 3464.10162 and the offset of the principal point of the image being 2000, where the unit is pixel. The image positions are shown in Table 1:

[0028] Table 1: Image positions

[0029] Photo 1 Photo 2 Photo 3 Photo 4 Photo 5 Photo 6 Photo 7 Photo 8 X 0.000000 70.710678 100.000000 70.710678 0.000000 -70.710678 -100.000000 -70.710678 Y -100.000000 -70.710678 0.000000 70.710678 100.000000 70.710678 0.000000 -70.710678 Z 100.000000 100.000000 100.000000 100.000000 100.000000 100.000000 100.000000 100.000000

[0030] Step 2, obtain image point coordinates

[0031] In the image, the method of manual selection can be used to determine the image point coordinates corresponding to the two object - space points. The image point coordinates corresponding to the two object - space points are shown in Table 2.

[0032] Table 2: Image point coordinates

[0033] Photo 1 Photo 2 Photo 3 Photo 4 Photo 5 Photo 6 Photo 7 Photo 8 <![CDATA[x1]]> 2928.202798 2757.874427 2000.000001 1242.125575 1071.797202 1421.205302 2000.000001 2578.794699 <![CDATA[y1]]> 1999.999997 2535.898143 2809.763278 2535.898146 2000 1590.730355 1448.208682 1590.730357 <![CDATA[x2]]> 1071.797202 1421.205301 1999.999999 2578.794698 2928.202798 2757.874425 1999.999999 1242.125573 <![CDATA[y2]]> 1999.999997 1590.730357 1448.208682 1590.730355 1999.999994 2535.898146 2809.763278 2535.898143

[0034] Step 3, object-space point positioning by double-point positioning algorithm

[0035] 3.1 The projection point of object-space point 1 on the image is a, and the projection coordinates of object-space point 2 on the image are b. The known focal length of the interior orientation elements of the image is 3464.10162 pixels, and the offset of the principal point of the image is 2000 pixels. Substituting the image point coordinates, image position, and camera internal parameters into the observation angle formula (1), the value of the observation angle ∠AOB can be obtained. The observation angle formula (1) is as follows:

[0036]

[0037] In the formula, O is the projection center of the image, and a and b are the projection points of object-space points 1 and 2 on the image, is the vector expression of the light ray Oa in the image space coordinate system, is the vector expression of the light ray Ob in the image space coordinate system, and ∠AOB is the observation angle.

[0038]

[0039]

[0040] In the formula, x1 and y1 are the image point coordinates of the projection point of object-space point 1 on the image, and x2 and y2 are the image point coordinates of the projection point of object-space point 2 on the image; C X 、C Y are the offsets of the principal point of the image in the camera internal parameters; -f is the camera focal length.

[0041] 3.2 The vector expression of the light ray OA in the image space coordinate system is (A - O), and the vector expression of the light ray OB in the image space coordinate system is (B - O). Then, an unknown equation can be constructed according to the observation angle formula (2). The object-space points A and B are unknown, and there are a total of 6 unknowns. Each image obtains an observation angle ∠AOB through formula (1), and the 6 obtained angles are used to construct an equation to jointly solve the coordinates of the object-space points A and B. The observation angle formula (2) is as follows:

[0042]

[0043] In the formula, O is the projection center of the image, A and B are two object-space points observed by the image, and ∠AOB is the observation angle.

[0044] The specific process of solving the equation is to use the Google open-source Ceres Solver for calculation, and the selected initial iteration coordinates are (30, 0, 0) and (-30, 0, 0).

[0045] Table 3: The two object-space points finally calculated

[0046] X Y Z Object point 1 37.893721970099 0.000000000024 -0.000007417518 Object point 2 -37.893721970099 -0.000000000025 -0.000007417518

[0047] The final calculation result is shown in Table 3.

Claims

1. A dual-point positioning method for multi-view oblique images, characterized by the following steps: (1) Using a drone to obtain at least 6 images that can simultaneously observe two object points; (2) Obtaining the image positions through the GPS of the drone; (3) Obtaining the image point coordinates of the two object points in the drone images; (4) Substituting the image positions, image point coordinates, and camera internal parameters into the dual-point positioning algorithm to obtain the object point coordinates; The calculation formula of the double-point positioning algorithm is as follows: (1) where O is the projection center of the image, and a and b are the image point coordinates of the object point on the image, is the observation angle; （2） where O is the projection center of the image, and A and B are two object points observed in the image, is the observation angle; Observation angle It is obtained by using formula (1) with the image point coordinates and the image position first; substituting into formula (2), the object points A and B are unknown, and there are a total of 6 unknowns. The coordinates of the object points A and B are jointly solved by at least 6 observation images.

Images