CN112070891A - Image area network adjustment method and system with digital ground model as three-dimensional control - Google Patents

Abstract

The invention provides a digital ground model as a three-dimensionally controlled image area network adjustment method and system, including sorting images according to the connection relationship of remote sensing images, establishing a coordinate observation equation of connecting points and image points; establishing a space curved surface based on the digital ground model As the observation equation of 3D control; calculate the directional derivative of the space surface according to the discrete points, count the mean and standard deviation of the distance observations, and calculate the 3D control observations of the space surface according to the accuracy of the distance observations and the influence of the gross error on the adjustment and positioning results. Taking the coordinate observation equation of the connected point and the point image and the spatial distance observation equation as the joint observation equation of the remote sensing image block adjustment, solve the remote sensing image block adjustment with the spatial surface corresponding to the digital ground model as the three-dimensional control information, and obtain the remote sensing image block adjustment. Image positioning parameters and object coordinates of connection points. The invention not only improves the elevation positioning accuracy of remote sensing images, but also effectively improves the plane positioning accuracy in the absence of ground control points.

Description

Image block adjustment method and system using digital ground model as 3D control

technical field

The invention belongs to the technical field of remote sensing, and relates to a three-dimensional image area network adjustment and positioning method using a space curved surface as three-dimensional control information, which is suitable for high-precision three-dimensional positioning of three-dimensional images without ground control points.

Background technique

High-precision geometric positioning of remote sensing stereo images requires reliable control information. Conventional methods generally use uniformly distributed obvious object points as adjustment control points to invert the geometric relationship between the image space and the ground space coordinate system and eliminate the image imaging process. Various systematic errors in the corresponding scale map surveying and mapping can be achieved. Therefore, obtaining enough ground control points is one of the important prerequisites to ensure the accuracy and reliability of the geometric positioning of the satellite remote sensing image regional network adjustment. In actual production, due to various conditions, GCP measurement is often time-consuming and labor-intensive. For difficult areas or large-scale global mapping, it is more difficult to obtain GCPs. Make full use of existing geographic information data resources, and adopt high-precision geometric positioning technology with sparse control or even no ground control to reduce the demand for ground control points. , an effective way to improve production efficiency, is also a long-term research focus and goal in the field of photogrammetry and remote sensing.

The systematic errors contained in the positioning parameters of remote sensing images may not necessarily cancel each other out in the regional network adjustment, and may even be accumulated systematically. With the scaling of the orbit direction, this error will accumulate systematically with the expansion of the area, resulting in a decrease in the absolute positioning accuracy of the uncontrolled free network adjustment. In order to get rid of the dependence on ground control points and realize the fusion application of multi-source geographic information big data, the researchers put forward the concept and technology of "cloud control", which can automatically extract data from image, vector and LIDAR point cloud data with geospatial information. Obtain generalized geometric control information.

Related prior art can be found in:

CN103823981B, a satellite image regional network adjustment method assisted by digital elevation model

CN110006408B, LIDAR data "cloud control" photogrammetry method

Zhang Zuxun, Tao Pengjie. Talking about "cloud-controlled" photogrammetry in the era of big data, Journal of Surveying and Mapping, 2017, pp. 1238-1248.

Digital elevation model (DEM) generally expresses three-dimensional terrain information in the form of regular grid, which can be used as control information to improve the accuracy of uncontrolled positioning of satellite images. The current technology can be divided into two categories: one is to use DEM as elevation control in the image block adjustment, to improve the image intersection conditions and to correct the systematic errors in the corresponding imaging model of the image, so as to improve the image elevation positioning accuracy, but the plane positioning The improvement of the accuracy also needs to rely on other control information; another type of registration of the DSM or DEM generated by image matching and the reference DEM, calculating the spatial similarity transformation parameters between the two, and correcting the satellite image positioning parameters. However, this kind of method requires automatic matching of stereo satellite images to generate DEM in advance, which increases the computational time and application complexity and limits the usage conditions of satellite image data.

By matching two sets of three-dimensional discrete point sets to solve the corresponding coordinate transformation parameters, the most classic and widely used method is called the ICP (Iterative Closest Point) method. ”, and then determine the transformation parameters between the two sets of spatial point sets through the solution, the biggest disadvantage of this method is the large amount of calculation of the nearest neighbor search in the iterative process.

SUMMARY OF THE INVENTION

The purpose of the present invention is to introduce the observation equation of the distance between the connection point and the space curved surface, and use the space curved surface of the corresponding target of the stereo image as the three-dimensional control information in the block adjustment, so as to satisfy the connection point and Corresponding to the registration constraints of the curved surface, the high-precision three-dimensional positioning of the stereo image elevation and plane is realized.

In order to achieve this purpose, the technical solution provided by the present invention provides a digital ground model as a three-dimensional control image area network adjustment method, which is characterized in that: it includes the following steps:

Step 1, sorting the images according to the connection relationship of the remote sensing images;

Step 2, establishing the coordinate observation equation of the connection point image point according to the remote sensing image imaging model;

Step 3, based on the digital ground model, establish a space curved surface as an observation equation for three-dimensional control, and obtain a space distance observation equation. The implementation method is as follows:

Take the digital ground model as a surface in the three-dimensional space, define the distance from the point in the three-dimensional space to the surface, and obtain the observation equation that uses the space surface as the three-dimensional control information, and then list the corresponding distance observation equations for all connection points to obtain the space distance observation equation;

Step 4: Calculate the directional derivative of the space surface according to the discrete points, including approximating the part of the spatial surface corresponding to the digital elevation model as the plane determined by the three vertices of the triangle, and the corresponding directional derivative and distance are calculated by the corresponding plane equation, thus Calculate and form the coefficient matrix in the spatial distance observation equation point by point;

Step 5: Calculate the mean and standard deviation of the distance observations. The implementation is as follows:

According to the accuracy of the digital ground model and the resolution of the remote sensing image, the accuracy of the distance between the connection point and the space surface is estimated, and the statistical histogram of the distance between the connection point and the corresponding space surface of the digital ground model is calculated based on this interval, and the effective connection within the distance range is calculated based on the statistical histogram. The mean and standard deviation of the distance from the point to the space surface;

Step 6, according to the result obtained in step 5, for the accuracy of the distance observation value and the influence of gross error on the adjustment positioning result, calculate the weight of the three-dimensional control observation value of the space surface;

Step 7: Use the coordinate observation equation of the connection points established in step 2 and the spatial distance observation equation established in step 3 as the joint observation equation of the remote sensing image regional network adjustment, and use the least squares principle to form a normal equation to solve the digital ground model. The remote sensing image regional network adjustment corresponding to the spatial curved surface as the three-dimensional control information, the remote sensing image positioning parameters and the object coordinate of the connection point are obtained.

Moreover, in step 1, the images are quickly sorted according to the connection points between the remote sensing images, and the implementation method is as follows:

First, establish a list of images associated with each scene image according to the connection points, calculate the number of images associated with each scene image, as the total number of connections of the images, and select the image with the largest total number of connections as the first scene image;

Then traverse the unsorted images in the sorted image association list, calculate the number of common images associated with the image and the sorted images as the number of common connections, and select the image with the largest number of common connections or the largest total number of connections as the next scene image;

Repeat the search and calculation until all images are sorted.

Moreover, in step 2, according to the image point coordinates of the connection point and the remote sensing image imaging model, the multi-slice forward intersection method is used to calculate the initial value of the object coordinate of the connection point, thereby further obtaining the observation equation of the image point coordinates of the connection point, and establishing the first group of observations Equation, matrix expression is as follows:

V ₁ =B ₁₁ X ₁ +B ₁₂ X ₂ -l ₁ , P ₁

Among them, V ₁ is the residual vector of image point coordinates, X ₁ is the correction vector of the positioning parameter of the stereo image, X ₂ is the correction vector of the object coordinate of the connection point, and B ₁₁ and B ₁₂ are respectively X ₁ , B 12 in the first group of observation equations. The coefficient matrix corresponding to X ₂ is calculated from the first-order derivative corresponding to the imaging model, l ₁ is the constant term vector, and P ₁ is the corresponding weight matrix;

Let the coordinates of all image points be independent observations, and the corresponding weight matrix P ₁ is a diagonal matrix; the weight p _i corresponding to the coordinate observations of each connection point is calculated as follows,

Among them, σ ₀ is the error in the unit weight, m is the number of observations corresponding to the connection point, and v _xi and v _yi are the residuals of the image point coordinates in the x and y directions.

Moreover, the implementation of step 3 is as follows,

Select the three-dimensional Cartesian coordinate system as the object space coordinate system, and the corresponding three coordinate axes are X, Y, Z respectively; take the digital ground model S as a surface in the three-dimensional space, and express it as the equation S(X, Y, Z) =0, the distance d from the point P in the three-dimensional space to the surface is defined as,

Among them, X _p , Y _p , Z _p are the spatial coordinates of the point P, S(X _p , Y _p , Z _p ) is the value of the surface equation corresponding to the point P, S _x , S _y , S _z are the surface equation relative to The first-order directional derivatives of X, Y, Z;

From this, the observation equation that uses the space curved surface as the three-dimensional control information is derived as follows:

Among them, ΔX, ΔY, ΔZ are the corrections of the object coordinate of the connection point, p _d is the weight of the distance observation value, and v _d is the observation residual error;

List the corresponding distance observation equations for all connection points, and obtain the spatial distance observation equations, as the second set of observation equations, the matrix form is,

V ₂ =B ₂₂ X ₂ -l ₂ ,P ₂

Among them, V ₂ is the residual vector of the distance observation value, X ₂ is the correction vector of the object coordinate of the connection point, and B ₂₂ is the coefficient matrix corresponding to X ₂ in the second group of observation equations, which is calculated by the first-order directional derivative of the space surface equation. , l ₂ is the constant term vector, P ₂ is the corresponding weight matrix;

Moreover, in step 5, if the statistical histogram has obvious peaks, the distance range corresponding to the maximum peak value is calculated, and if there is no obvious peak value, the range corresponding to the connection points with relatively concentrated distances is calculated.

Moreover, in step 6, the following formula is used to calculate the weight p _d of the distance observation value,

Among them, σ ₀ is the error in the unit weight, exp( ) is the exponential function of the natural number e; α is a positive number less than 1, _d is the distance calculated in step 3, and ed and σ _d are the distance observations, respectively The mean and standard deviation of ;

Assuming that the distance from each connection point to the digital ground model is an independent observation value, the weight matrix P ₂ is a diagonal matrix, and its diagonal elements are calculated point by point according to the above formula.

Moreover, in step 7, the point-by-point elimination method is used to eliminate the unknown X ₂ , and a reduced equation that only includes the correction vector X ₁ of the remote sensing image positioning parameter is formed; the reduced equation is solved and the positioning parameters of the stereo image are corrected, using The multi-slice forward intersection method updates the object coordinate of the connection point and realizes the intersection of two sets of unknowns.

The present invention also provides a digital ground model as a three-dimensionally controlled image area network adjustment system for realizing the above-mentioned digital ground model as a three-dimensionally controlled image area network adjustment method.

Compared with the prior art, the advantages and beneficial effects of the present invention:

In the present invention, the digital elevation model or the space curved surface is directly used as the three-dimensional control information in the adjustment process of the remote sensing image area network, and the open source geographic information data is fully utilized to effectively eliminate the system offset in the parameters of the remote sensing image imaging model, and realize the remote sensing image without High-precision 3D positioning in the case of ground control points. Taking the digital ground model as an example of a 3D control application, it can not only improve the elevation positioning accuracy, but also effectively improve the plane positioning accuracy in the absence of ground control points.

Description of drawings

FIG. 1 is a flow chart of calculation of a regional network adjustment of a digital ground model as a three-dimensional control remote sensing image according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of calculating a distance from an arbitrary point in a three-dimensional space to a space curved surface according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of a change curve of a weight function coefficient of an observation value of a distance between a connection point and a space curved surface according to an embodiment of the present invention.

Detailed ways

The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

The present invention notices that there is no direct correspondence between the point of the same name (connection point) between the stereoscopic images and the grid point of the space surface, but the connection point and the space surface grid are both descriptions of the corresponding target, and the difference between the two The registration problem should be understood as the matching of the three-dimensional space point set and the space surface. Based on such a basic fact, the present invention takes the spatial curved surface of the corresponding target of the stereo image as the three-dimensional control information in the block adjustment by introducing the corresponding observation equation, and satisfies the connection point and the corresponding connection point at the same time during the block adjustment process of the stereo remote sensing image. The registration constraint of the curved surface realizes the high-precision 3D positioning of the stereo remote sensing image.

Embodiments of the present invention provide a remote sensing image regional network adjustment method using a digital ground model as three-dimensional control information. First, a remote sensing image association list is established based on connection points, and further images are quickly sorted to reduce the normal equation bandwidth. The distance between the connection point in the three-dimensional object space and the corresponding space surface of the digital ground model is taken as the observed value of the three-dimensional control, and the directional derivative of the surface is calculated by local fitting, thereby establishing the corresponding three-dimensional control observation equation. The mean and variance of the distance observations are counted, and the corresponding weight function of the distance observations is constructed, which effectively eliminates the influence of gross errors and block adjustment while taking into account the errors of distance observations. It can not only improve the elevation positioning accuracy of remote sensing images, but also effectively improve the plane positioning accuracy without ground control points.

Referring to FIG. 1 , a method for positioning a stereoscopic image using a space curved surface as three-dimensional control information provided by an embodiment of the present invention includes the following steps:

Step 1: Sort the images according to the connection relationship of the remote sensing images.

When the distribution of remote sensing images does not have the characteristics of regular flight zones, the traditional minimum bandwidth calculation of photogrammetric block adjustment and image sorting methods are not suitable for large-scale remote sensing image block adjustment. In order to effectively reduce the bandwidth of the normal equation and improve the efficiency of the adjustment solution, in principle, it is necessary to arrange the related images together as much as possible.

The present invention further proposes to quickly sort the images according to the connection points between the remote sensing images. The specific technical solution is as follows: first, establish a list of images associated with each scene image according to the connection points, calculate the number of images associated with each scene image, take it as the total number of connections of the images, and select the image with the largest total number of connections as the first scene image ; Then traverse the unsorted images in the sorted image association list, calculate the number of common images associated with the image and the sorted images as the number of common connections, and select the image with the largest number of common connections or the largest total number of connections as the next scene image; Repeat the search and calculation until all images are sorted.

Step 2, establishing the coordinate observation equation of the connection point image point according to the remote sensing image imaging model.

According to the image point coordinates of the connection point and the remote sensing image imaging model, the multi-slice forward intersection method is used to calculate the initial value of the object coordinate of the connection point, and the observation equation of the image point coordinate of the connection point is further obtained. The expression is as follows:

V ₁ =B ₁₁ X ₁ +B ₁₂ X ₂ -l ₁ , P ₁ (1)

Among them, V ₁ is the residual vector of image point coordinates, X ₁ is the correction vector of the positioning parameter of the stereo image, X ₂ is the correction vector of the object coordinate of the connection point, and B ₁₁ and B ₁₂ are respectively X ₁ , B 12 in the first group of observation equations. The coefficient matrix corresponding to X ₂ is calculated from the first derivative corresponding to the imaging model, l ₁ is the constant term vector, and P ₁ is the corresponding weight matrix.

The present invention assumes that all image point coordinates are independent observation values, and the corresponding weight matrix P ₁ is a diagonal matrix. Further, the weight p _i corresponding to the coordinate observation value of each connection point is calculated as follows:

Among them, σ ₀ is the error in the unit weight, m is the number of observations corresponding to the connection point, and v _xi and v _yi are the residuals of the image point coordinates in the x and y directions.

In step 3, a space curved surface is established as an observation equation for three-dimensional control, and a space distance observation equation is obtained.

Referring to FIG. 2 , in the embodiment, a three-dimensional rectangular coordinate system is selected as the object space coordinate system, and the corresponding three coordinate axes are X, Y, and Z respectively. The digital ground model S is a curved surface in the three-dimensional space, and its general mathematical form can be expressed as the equation S(X, Y, Z)=0, and the distance d from the point P in the three-dimensional space to the curved surface is preferably defined as,

Among them, X _p , Y _p , Z _p are the spatial coordinates of the point P, S(X _p , Y _p , Z _p ) is the value of the surface equation corresponding to the point P, S _x , S _y , S _z are the surface equations relative to the The first directional derivatives of X, Y, Z. During specific implementation, other distance definition methods may also be used.

From this, the observation equation that takes the space curved surface as the three-dimensional control information can be derived as follows:

Among them, ΔX, ΔY, ΔZ are the corrections of the object coordinate of the connection point, p _d is the weight of the distance observation value, and v _d is the observation residual error.

List the corresponding distance observation equations for all connection points, that is, the spatial distance observation equations. At this time, the second set of observation equations is obtained, and its matrix form is,

V ₂ =B ₂₂ X ₂ -l ₂ ,P ₂ (5)

Among them, V ₂ is the residual vector of the distance observation value, X ₂ is the correction vector of the object coordinate of the connection point, and B ₂₂ is the coefficient matrix corresponding to X ₂ in the second group of observation equations, which is calculated by the first-order directional derivative of the space surface equation. , l ₂ is the constant term vector, and P ₂ is the corresponding weight matrix.

Step 4: Calculate the directional derivative of the space surface according to the discrete points.

In practical applications, it is difficult to obtain the complete mathematical expression of the spatial surface equation corresponding to the digital ground model, and a local fitting method is generally used for approximation. When the expression form of the space surface is a three-dimensional grid, a local approximate surface can be obtained by bilinear fitting or bicubic fitting, and its directional derivative and distance can be calculated accordingly. Further, the spatial surface corresponding to the digital ground model can be expressed by a spatial triangulation network composed of a set of three-dimensional discrete points. The directional derivative and distance of , can be calculated by the corresponding plane equation, so that the coefficient matrix B ₂₂ in the observation equation (5) can be calculated point by point and formed.

Step 5: Count the mean and standard deviation of the distance observations.

According to the accuracy of the digital ground model and the resolution of the remote sensing image, the accuracy of the distance between the connection point and the space surface is estimated, and the statistical histogram of the distance between the connection point and the corresponding space surface of the digital ground model is calculated based on this interval. If the histogram has obvious peaks, calculate the distance range corresponding to the largest peak; if there is no obvious peak, delete a certain number (such as 10%) of relatively discrete distance values at both ends of the histogram, and obtain relatively concentrated connection points corresponding to distance range. On this basis, the mean value _{ed and the standard deviation σ d of} the distance between the effective connection point and the space surface within the distance range are calculated statistically, that is, the mean value and standard deviation of the distance observations.

Step 6, according to the result obtained in step 5, for the accuracy of the distance observation value and the influence of gross error on the adjustment and positioning result, calculate the weight of the three-dimensional control observation value of the space surface.

There may be errors between the corresponding space surface of the digital ground model expressed by grid or discrete point set and the real terrain or target surface. Due to different data acquisition methods, the connection points obtained from the digital ground model corresponding to the space surface and remote sensing images may be incomplete. Consistent. Therefore, the weight of the distance observation value should be determined according to its statistical accuracy, and at the same time, the influence of "gross error" on the adjustment and positioning results should be effectively eliminated. In this embodiment of the present invention, the following formula is preferably used to calculate the weight p _d of the distance observation value:

Among them, σ ₀ is the error in the unit weight, and exp(·) is the exponential function of the natural number e. α is a positive number less than 1, d is the distance calculated in step 3, and _{ed and σ d are} the mean and standard deviation of the distance observations, respectively.

Referring to Figure 3, selecting α=0.03, the change curve of the weight coefficient function with the ratio of (-ed )/σ _d can be calculated _. When the ratio is less than 1, the weight coefficient is approximately equal to 1, and when the ratio is greater than 3.3, the weight coefficient is close to 0. It can be understood that the weight function constructed in this way can simultaneously take into account the accuracy of the distance observations and eliminate the influence of gross errors on the adjustment and positioning results.

During specific implementation, other methods may also be used to realize the calculation of the weight of the distance observation value.

The present invention assumes that the distance between each connection point and the digital ground model is an independent observation value, the weight matrix P2 corresponding to the observation equation ( ₅ ) is a diagonal matrix, and its diagonal elements can be calculated point by point according to the formula (6). .

Step 7, form a reduced normal equation and solve it.

Taking the coordinate observation equation (1) of the connection points established in step 2 and the spatial distance observation equation (5) established in step 3 as the joint observation equation of the regional network adjustment of remote sensing images, the least squares principle is used to form the normal equation and solve The digital ground model corresponds to the spatial curved surface as the three-dimensional control information of the remote sensing image regional network adjustment, and the remote sensing image positioning parameters and the object coordinates of the connection point can be obtained. Further, considering that the number of correction items of the object coordinate of the connection point is generally large, in order to improve the calculation efficiency, the embodiment of the present invention adopts the point-by-point elimination method to eliminate the unknown number X ₂ , and the composition only includes the correction vector X ₁ of the remote sensing image positioning parameter. The reduction equation of . Solve the reduced equation and correct the positioning parameters of the stereo image, and use the multi-slice forward intersection method to update the object coordinate of the connection point to realize the intersection of two sets of unknowns.

During specific implementation, the method proposed by the technical solution of the present invention can be automatically run by those skilled in the art using computer software technology. The system device for running the method is, for example, a computer-readable storage medium storing a computer program corresponding to the technical solution of the present invention, and a computer that runs the corresponding computer program. The computer equipment of the program should also be within the protection scope of the present invention.

Embodiments of the present invention further provide a digital ground model as a three-dimensionally controlled image area network adjustment system, which is used to implement the above-mentioned digital ground model as a three-dimensionally controlled image area network adjustment method. The system implementation can be easily implemented with reference to the above method, which is not repeated in the present invention.

In some possible embodiments, a digital ground model is provided as a three-dimensionally controlled image area network adjustment system, including a processor and a memory, the memory is used to store program instructions, and the processor is used to call the stored instructions in the processor to execute the above The described digital ground model is used as a three-dimensionally controlled image area network adjustment method.

In some possible embodiments, a digital ground model is provided as a three-dimensionally controlled image area network adjustment system, including a readable storage medium on which a computer program is stored, and when the computer program is executed, A digital ground model as described above is implemented as an image block adjustment method for 3D control.

After the preferred embodiments are described in detail, those skilled in the art can clearly understand that various changes and modifications can be made without departing from the scope and spirit of the above-mentioned patent application. Any simple modifications, equivalent changes and modifications belong to the scope of the technical solutions of the present invention. And the present invention is not limited by the example implementations exemplified in the description.

Claims (8)

1. a digital ground model as the image area network adjustment method of three-dimensional control, is characterized in that: comprise the steps, Step 1, sorting the images according to the connection relationship of the remote sensing images; Step 2, establishing the coordinate observation equation of the connection point image point according to the remote sensing image imaging model; Step 3, based on the digital ground model, establish a space curved surface as an observation equation for three-dimensional control, and obtain a space distance observation equation. The implementation method is as follows: Take the digital ground model as a surface in the three-dimensional space, define the distance from the point in the three-dimensional space to the surface, and obtain the observation equation that uses the space surface as the three-dimensional control information, and then list the corresponding distance observation equations for all connection points to obtain the space distance observation equation; Step 4: Calculate the directional derivative of the space surface according to the discrete points, including approximating the part of the spatial surface corresponding to the digital elevation model as the plane determined by the three vertices of the triangle, and the corresponding directional derivative and distance are calculated by the corresponding plane equation, thus Calculate and form the coefficient matrix in the spatial distance observation equation point by point; Step 5: Calculate the mean and standard deviation of the distance observations. The implementation is as follows: According to the accuracy of the digital ground model and the resolution of the remote sensing image, the accuracy of the distance between the connection point and the space surface is estimated, and the statistical histogram of the distance between the connection point and the corresponding space surface of the digital ground model is calculated based on this interval, and the effective connection within the distance range is calculated based on the statistical histogram. The mean and standard deviation of the distance from the point to the space surface; Step 6, according to the result obtained in step 5, for the accuracy of the distance observation value and the influence of gross error on the adjustment positioning result, calculate the weight of the three-dimensional control observation value of the space surface; Step 7: Use the coordinate observation equation of the connection points established in step 2 and the spatial distance observation equation established in step 3 as the joint observation equation of the remote sensing image regional network adjustment, and use the least squares principle to form a normal equation to solve the digital ground model. The remote sensing image regional network adjustment corresponding to the spatial curved surface as the three-dimensional control information, the remote sensing image positioning parameters and the object coordinate of the connection point are obtained. 2. The digital ground model according to claim 1, as a three-dimensionally controlled image area network adjustment method, is characterized in that: in step 1, images are quickly sorted according to the connection points between remote sensing images, and the implementation is as follows, First, establish a list of images associated with each scene image according to the connection points, calculate the number of images associated with each scene image, as the total number of connections of the images, and select the image with the largest total number of connections as the first scene image; Then traverse the unsorted images in the sorted image association list, calculate the number of common images associated with the image and the sorted images as the number of common connections, and select the image with the largest number of common connections or the largest total number of connections as the next scene image; Repeat the search and calculation until all images are sorted. 3. The digital ground model according to claim 1 and 2 is used as a three-dimensionally controlled image area network adjustment method, characterized in that: in step 2, according to the point coordinates of the connection points and the remote sensing image imaging model, a multi-piece forward intersection method is adopted. The initial value of the object coordinate of the connection point is calculated, and the observation equation of the image point coordinate of the connection point is further obtained, and the first group of observation equations is established. The matrix expression is as follows: V ₁ =B ₁₁ X ₁ +B ₁₂ X ₂ -l ₁ , P ₁ Among them, V ₁ is the residual vector of image point coordinates, X ₁ is the correction vector of the positioning parameter of the stereo image, X ₂ is the correction vector of the object coordinate of the connection point, and B ₁₁ and B ₁₂ are respectively X ₁ , B 12 in the first group of observation equations. The coefficient matrix corresponding to X ₂ is calculated from the first-order derivative corresponding to the imaging model, l ₁ is the constant term vector, and P ₁ is the corresponding weight matrix; Let the coordinates of all image points be independent observations, and the corresponding weight matrix P ₁ is a diagonal matrix; the weight p _i corresponding to the coordinate observations of each connection point is calculated as follows,

Among them, σ ₀ is the error in the unit weight, m is the number of observations corresponding to the connection point, and v _xi and v _yi are the residuals of the image point coordinates in the x and y directions. 4. according to the described digital ground model of claim 3, as the image area network adjustment method of three-dimensional control, it is characterized in that: the realization mode of step 3 is as follows, Select the three-dimensional Cartesian coordinate system as the object space coordinate system, and the corresponding three coordinate axes are X, Y, Z respectively; take the digital ground model S as a surface in the three-dimensional space, and express it as the equation S(X, Y, Z) =0, the distance d from the point P in the three-dimensional space to the surface is defined as,

Among them, X _p , Y _p , Z _p are the spatial coordinates of the point P, S(X _p , Y _p , Z _p ) is the value of the surface equation corresponding to the point P, S _x , S _y , S _z are the surface equations relative to the The first-order directional derivatives of X, Y, Z; From this, the observation equation that uses the space curved surface as the three-dimensional control information is derived as follows:

Among them, ΔX, ΔY, ΔZ are the correction amount of the object coordinate of the connection point, p _d is the weight of the distance observation value, and v _d is the observation residual error; List the corresponding distance observation equations for all connection points, and obtain the spatial distance observation equations, as the second set of observation equations, the matrix form is, V ₂ =B ₂₂ X ₂ -l ₂ , P ₂ Among them, V ₂ is the residual vector of the distance observation value, X ₂ is the correction vector of the object coordinate of the connection point, and B ₂₂ is the coefficient matrix corresponding to X ₂ in the second group of observation equations, which is calculated by the first-order directional derivative of the space surface equation. , l ₂ is the constant term vector, and P ₂ is the corresponding weight matrix. 5. according to the described digital ground model of claim 1 and 2, it is characterized in that as the image area network adjustment method of three-dimensional control: in step 5, if the statistical histogram has obvious peak value, then calculate the distance range corresponding to the maximum peak value, If there is no obvious peak, the range corresponding to the connection points with relatively concentrated distances is counted. 6. The digital ground model according to claim 4 is used as a three-dimensionally controlled image area network adjustment method, characterized in that: in step 6, the following formula is used to calculate the weight p _d of the distance observation value,

Among them, σ ₀ is the error in the unit weight, exp( ) is the exponential function of the natural number e; α is a positive number less than 1, _d is the distance calculated in step 3, and ed and σ _d are the distance observations, respectively The mean and standard deviation of ; Assuming that the distance from each connection point to the digital ground model is an independent observation value, the weight matrix P ₂ is a diagonal matrix, and its diagonal elements are calculated point by point according to the above formula. 7. The digital ground model according to claim 6 is used as a three-dimensionally controlled image area network adjustment method, characterized in that: in step 7, a point-by-point elimination method is adopted to eliminate the unknown X ₂ , and the composition only includes the correction of remote sensing image positioning parameters The reduced normal equation of vector X ₁ ; solve this reduced normal equation and correct the positioning parameters of the stereo image, and use the multi-slice forward intersection method to update the object coordinate of the connection point to realize the intersection of two sets of unknowns. 8. A digital ground model as a three-dimensionally controlled image area network adjustment system, characterized in that: for realizing a digital ground model as described in any one of claims 1-7 as a three-dimensionally controlled image area network adjustment poor method.

Images