Methods and systems for detecting ill-conditioned observation areas in satellite imagery regional network adjustment

Abstract

This invention discloses a method and system for detecting ill-conditioned observation regions in satellite imagery regional network adjustment. The method includes: acquiring multiple satellite images to be detected within the regional network, and multiple connection points corresponding to the satellite images; for any connection point, obtaining an adjustment parameter variance estimation model with elevation constraints; based on the satellite images, constructing adjustment parameter constraint equations for an error correction model by setting a control image, thereby obtaining a control image accuracy propagation variance estimation model; and, based on the two-stage error propagation law, detecting the existence of ill-conditioned observation regions on the corresponding satellite images by evaluating the model estimation accuracy of the error correction model corresponding to different locations on each satellite image. This invention, before regional network adjustment, quantitatively expresses the propagation law of satellite images within the regional network through modeling, enabling the detection of satellite images with ill-conditioned observation regions within the regional network. This solves the problems of high labor costs and poor reliability associated with traditional post-processing manual detection methods.

Description

Technical Field

[0001] This invention relates to a method and system for detecting ill-conditioned observation areas in satellite image regional network adjustment, belonging to the field of high-precision processing technology for optical satellite remote sensing images. Background Technology

[0002] Satellite imagery regional network adjustment is a key technology for correcting geometric errors in regional images and improving their geometric accuracy. It is an essential step in the high-precision processing and application of satellite remote sensing images. Regional network adjustment uses corresponding image points between satellite images as the main observations and estimates adjustment parameters using the geometric constraints of spatial intersections of corresponding rays. With the development of automated digital technologies such as image matching, image matching is currently commonly used to automatically extract tie-point observations from overlapping areas of regional images to construct regional networks. However, due to potential differences in geometry, radiometry, imaging angle, texture, and ground features among satellite images within a regional network, problems such as uneven spatial distribution of tie points, low repetition rate, weak local image observation conditions, and even broken connections in local areas can easily arise during automatic matching network construction. This makes it impossible to guarantee the consistency of geometric accuracy within the regional network adjustment.

[0003] Due to the complex topological relationships within the regional network, these ill-conditioned areas are often obscured by the entire network, making them difficult to detect before regional network adjustment. In practice, this relies heavily on the experience of operators, requiring extensive manual checks and accuracy index statistics after regional network adjustment to identify quality problems. However, directly locating the problematic image is often challenging. Currently, large-scale, high-precision remote sensing imagery has become the foundation for intelligent remote sensing big data research and applications. In practical applications, the scale of regional networks is increasingly large, with the number of images to be adjusted often exceeding a thousand. Relying on manual checks has become extremely difficult and costly. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for detecting ill-conditioned observation areas in satellite image regional network adjustment. Before regional network adjustment, the propagation law of satellite images in the regional network is expressed quantitatively through modeling, and the theoretical accuracy of abnormal images is detected. In this way, satellite images with ill-conditioned observation areas in the regional network are automatically detected, which solves the problems of high labor cost and poor reliability of traditional post-process manual detection methods.

[0005] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0006] On the one hand, this invention discloses a method for detecting ill-conditioned observation regions in satellite imagery regional network adjustment, including:

[0007] Acquire multiple satellite images to be detected within the regional network, as well as multiple connection points corresponding to the satellite images;

[0008] For any connection point, based on the preset error correction model, an adjustment parameter variance estimation model for elevation constraints is constructed.

[0009] Based on the satellite imagery, by setting control images, the adjustment parameter constraint equations of the error correction model are constructed.

[0010] Based on the adjustment parameter variance estimation model and adjustment parameter constraint equation, the control image accuracy propagation variance estimation model is obtained.

[0011] Based on the control image accuracy propagation variance estimation model, and based on the error propagation law, the theoretical accuracy of the adjustment parameters of the error correction model for all satellite images is obtained by calculating the correction cofactor matrix. Then, based on the error propagation law, the model estimation accuracy of the error correction model corresponding to different locations on each satellite image is evaluated to detect whether there are ill-conditioned observation areas on the corresponding satellite images.

[0012] Furthermore, the step of obtaining the variance estimation model of the adjustment parameters for elevation constraints includes:

[0013] The regional network contains satellite images with multiple degrees of overlap, so one connection point corresponds to multiple image points with the same name;

[0014] For any image point with the same name, the basic adjustment equation is constructed based on the preset error correction model and rational function model, and the spatial intersection constraint equation is obtained by parameter linearization.

[0015] For any connection point, based on the spatial intersection constraint equations of all corresponding image points, the spatial intersection constraint equation of the connection point is established. Then, the elevation constraint equation with additional prior accuracy is introduced to obtain the variance estimation model of the adjustment parameters of the elevation constraint.

[0016] Furthermore, the expression for the variance estimation model of the adjustment parameters of the elevation constraint is as follows:

[0017] ;

[0018] In the formula, Represents the residual vector of the spatial intersection constraint equation of the i-th connection point;

[0019] Let represent the partial derivative matrix of the spatial intersection constraint equations for the i-th connection point with respect to the adjustment parameters;

[0020] This represents the adjustment parameter correction number for the satellite image corresponding to the i-th tie point;

[0021] Let represent the partial derivative matrix of the spatial intersection constraint equation of the i-th connection point with respect to the ground coordinates;

[0022] Represents the object coordinates of the i-th connection point;

[0023] This represents the negative value of the current value vector of the fundamental adjustment equation for the i-th connection point;

[0024] The weights of the spatial intersection constraint equations for the i-th connection point are represented.

[0025] Represents the residual vector of the elevation constraint equation for the i-th connection point; Represents the elevation correction; This represents the weight of the elevation constraint equation for the i-th connection point.

[0026] Furthermore, based on the satellite imagery, the step of constructing the adjustment parameter constraint equations of the error correction model by setting control images includes:

[0027] Set any satellite image within the regional network as a control image, and divide the control image into multiple grids evenly, generating a uniformly distributed virtual image point at the center of each grid.

[0028] For any virtual image point, using the virtual image point as the observation value, by controlling the initial propagation accuracy of the control image, by constructing the basic adjustment equation of the virtual image point and linearizing the parameters, the adjustment parameter constraint equation of the virtual image point is obtained.

[0029] Based on the adjustment parameter constraint equations for all virtual image points, the adjustment parameter constraint equations corresponding to the control image are obtained.

[0030] Furthermore, the expression for the adjustment parameter constraint equation of the virtual image point is as follows:

[0031] ;

[0032] In the formula, Represents the residual vector of the adjustment parameter constraint equation for the k-th virtual image point;

[0033] This represents the partial derivative matrix of the adjustment parameter constraint equation for the k-th virtual image point with respect to the adjustment parameters;

[0034] This represents the adjustment parameter correction number for the control image corresponding to the k-th virtual image point;

[0035] This represents the negative value of the current value vector of the fundamental adjustment equation for the k-th virtual image point;

[0036] Let represent the weight matrix of the adjustment parameter constraint equation for the k-th virtual image point.

[0037] Furthermore, the expression for the control image accuracy propagation variance estimation model is as follows:

[0038] ;

[0039] , , , ,

[0040] , , ;

[0041] In the formula, The set of residual vectors representing each constraint equation; Represents the residual vector of the spatial intersection constraint equations for all connection points; Represents the residual vector of the adjustment parameter constraint equations for all virtual image points; This represents the residual vector of the additional constraint equations introduced for the object coordinates of the connection points;

[0042] This represents the set of partial derivative matrices of each constraint equation with respect to the adjustment parameters; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the adjustment parameters; This represents the partial derivative matrix of the adjustment parameter constraint equations for all virtual image points with respect to the adjustment parameters;

[0043] This represents the set of partial derivative matrices of the constraint equations with respect to ground coordinates; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the ground coordinates. This represents the coefficient matrix of the introduced additional constraint equations;

[0044] The set of negative values ​​representing the current value vector of each of the basic adjustment equations; This represents the negative value of the current value vector of the fundamental adjustment equations for all connection points; This represents the negative value of the current value vector of the fundamental adjustment equations for all virtual image points;

[0045] The set representing the weight matrices of each constraint equation; The weight matrix represents the spatial intersection constraint equations for all connection points; The weight matrix represents the adjustment parameter constraint equations for all virtual image points; This represents the weight matrix of the introduced additional constraint equations; Indicates elevation accuracy;

[0046] This represents the vector of adjustment parameter corrections for all images; This represents the correction vector for the object coordinates of all connection points.

[0047] Furthermore, the expression for the corrected cofactor matrix is:

[0048] ;

[0049] In the formula, Represents the corrected cofactor matrix; This represents the first transition matrix; This represents the second transition matrix; Indicates the third transition matrix; This represents the fourth transition matrix.

[0050] Furthermore, the step of detecting whether there are ill-conditioned observation areas on the corresponding satellite image by evaluating the model estimation accuracy of the error correction model corresponding to different locations on each satellite image includes:

[0051] Each satellite image to be detected is evenly divided into multiple grids, and a uniformly distributed virtual image point is generated at the center of each grid.

[0052] For any virtual image point, based on the error propagation law, the model covariance matrix of the virtual image point is calculated according to the correction cofactor matrix, and the model estimation accuracy of the error correction model of the virtual image point is obtained.

[0053] For each satellite image, based on the model estimation accuracy of the error correction model for all virtual image points included, the mean, maximum and minimum values ​​of the model estimation accuracy and the root mean square error value are statistically analyzed.

[0054] If the maximum value of the model estimation accuracy of any satellite image is greater than a preset maximum value threshold, then it is determined that there is an ill-observed area on the satellite image.

[0055] If the difference between the extreme value and the mean value of the model estimation accuracy of any satellite image is greater than a preset root mean square error value, then it is determined that there is an ill-conditioned observation area on the satellite image.

[0056] Furthermore, the expression for the model covariance matrix of the virtual image point is as follows:

[0057] ;

[0058] In the formula, This represents the model covariance matrix of the m-th virtual image point; This represents the partial derivative matrix of the error correction model with respect to the adjustment parameters; Represents the corrected covariance matrix;

[0059] The expression for the model estimation accuracy of the error correction model for the virtual image points is as follows:

[0060] ;

[0061] In the formula, This represents the model estimation accuracy of the error correction model corresponding to the row direction of the m-th virtual image point; This represents the model estimation accuracy of the error correction model corresponding to the column direction of the m-th virtual image point; Representing the model covariance matrix The element in the first row and first column; Show model covariance matrix The element in the last row and last column.

[0062] On the other hand, this invention discloses a detection system for ill-conditioned observation regions in satellite image regional network adjustment, applicable to the aforementioned method for detecting ill-conditioned observation regions in satellite image regional network adjustment, comprising:

[0063] The data acquisition module is used to acquire multiple satellite images to be detected within the regional network, as well as multiple connection points corresponding to the satellite images;

[0064] The connection point module, for any connection point, constructs an adjustment parameter variance estimation model for elevation constraints based on a preset error correction model.

[0065] The control image module is used to construct the adjustment parameter constraint equations of the error correction model based on the satellite image and by setting the control image.

[0066] The control image accuracy propagation module is used to obtain the control image accuracy propagation variance estimation model based on the adjustment parameter variance estimation model and adjustment parameter constraint equation.

[0067] The ill-observed area detection module is used to obtain the theoretical accuracy of the adjustment parameters of the error correction model of all satellite images by calculating the correction cofactor matrix based on the error propagation law, according to the control image accuracy propagation variance estimation model; and then, based on the error propagation law, to detect whether there are ill-observed areas on the corresponding satellite images by evaluating the model estimation accuracy of the error correction model corresponding to different locations on each satellite image.

[0068] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0069] The present invention relates to a method and system for detecting ill-conditioned observation areas in satellite imagery regional network adjustment. Before regional network adjustment, the method quantitatively expresses the propagation law of satellite images within the regional network through modeling, performs theoretical accuracy anomaly image detection, and automatically detects satellite images with ill-conditioned observation areas within the regional network. This solves the problems of high labor costs and poor reliability of traditional post-process manual detection methods. Starting from the mathematical essence of regional network adjustment, the method more fully considers the mutual influence of tie point distribution, satellite image geometric conditions, and error models, providing scientific guidance for regional network adjustment quality control. Attached Figure Description

[0070] Figure 1 This is a flowchart of the method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment provided in Embodiment 1 of the present invention;

[0071] Figure 2 This is a schematic diagram of generating virtual image points provided in Embodiment 1 of the present invention. Detailed Implementation

[0072] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0073] Example 1

[0074] This embodiment 1 provides a method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment, including:

[0075] Acquire multiple satellite images to be detected within the local area network, as well as multiple connection points corresponding to the satellite images;

[0076] For any connection point, based on the preset error correction model, an adjustment parameter variance estimation model for elevation constraints is constructed.

[0077] Based on satellite imagery, by setting control images, adjustment parameter constraint equations for the error correction model are constructed.

[0078] Based on the adjustment parameter variance estimation model and adjustment parameter constraint equation, the control image accuracy propagation variance estimation model is obtained.

[0079] Based on the control image accuracy propagation variance estimation model, and based on the error propagation law, the theoretical accuracy of the adjustment parameters of the error correction model for all satellite images is obtained by calculating the correction cofactor matrix. Then, based on the error propagation law, the model estimation accuracy of the error correction model corresponding to different locations on each satellite image is evaluated to detect whether there are ill-conditioned observation areas on the corresponding satellite images.

[0080] The technical concept of this invention is to quantitatively express the propagation law of satellite imagery within the regional network through modeling before regional network adjustment, and to detect anomalous images with theoretical accuracy, thereby automatically detecting satellite images within the regional network that exhibit observational ill-formed regions. This solves the problems of high labor costs and poor reliability associated with traditional post-processing manual detection methods. Starting from the mathematical essence of regional network adjustment, it more fully considers the interaction between the distribution of tie points, the geometric conditions of satellite imagery, and the error model, providing scientific guidance for the quality control of regional network adjustment.

[0081] like Figure 1 As shown, the specific steps are as follows:

[0082] Step 1: Acquire multiple satellite images to be detected within the local area network, as well as multiple connection points corresponding to the satellite images.

[0083] The process involves acquiring satellite imagery to be inspected within the regional network, obtaining tie points through automatic image matching, and then combining this with the image error correction model to be used in the regional network adjustment. The error correction model commonly used in regional network adjustment is usually based on the image points of satellite imagery. The translation transformation model or affine transformation model. Commonly used affine transformation models are as follows:

[0084]

[0085] In the formula, This represents the line direction error correction model; Indicates the coefficients of the first model; Indicates the coefficients of the second model; Indicates the coefficients of the third model; This represents the column direction error correction model; Indicates the coefficients of the fourth model; Indicates the coefficients of the fifth model; Indicates the coefficients of the sixth model;

[0086] in, These are all model coefficients, and also the adjustment parameters for each satellite image in the regional network adjustment.

[0087] By incorporating the affine transformation model into the image side of the rational function model of satellite imagery, a single-scene image imaging model with an additional image side error correction model can be constructed. The specific form of the model is as follows:

[0088]

[0089] In the formula, Indicates a line number like a dot; This represents the line direction error correction model; A rational function model of satellite imagery with respect to the row direction of image points; Represents a column number like a dot; This represents the column direction error correction model; A rational function model of satellite imagery with respect to the direction of image point columns;

[0090] Representing image points The latitude of the corresponding ground point in the geodetic coordinate system; Representing image points The longitude of the corresponding ground point in the geodetic coordinate system; Representing image points The elevation of the corresponding ground point in the spatial geodetic coordinate system.

[0091] Step 2: For any connection point, construct the variance estimation model of the adjustment parameters for the elevation constraint based on the preset error correction model.

[0092] Specifically, using the tie points as observations, the basic adjustment equations are constructed based on the spatial intersection geometry of the light rays at the tie points. Considering the presence of highly overlapping satellite imagery within a complex regional network, a tie point corresponds to multiple image points with the same name.

[0093] For any image point with the same name, the basic adjustment equation is constructed and the spatial intersection constraint equation is obtained by parametric linearization;

[0094] For any connection point, based on the spatial intersection constraint equations of all corresponding image points, a spatial intersection constraint equation for the connection point is established. Then, an elevation constraint equation with additional prior accuracy is introduced to obtain the variance estimation model of the adjustment parameters of the elevation constraint.

[0095] Specifically, assuming the first The first connection point The same image point is Its basic adjustment equation can be expressed as:

[0096]

[0097] In the formula, Indicates the same image point The corresponding basic adjustment equations for the image row direction; Indicates the same image point The corresponding image column direction basic adjustment equation.

[0098] Then, by further linearizing the parameters, we can obtain the corresponding image points. Spatial intersection constraint equations:

[0099]

[0100]

[0101] ,

[0102] In the formula, Indicates the same image point The residual vector of the spatial intersection constraint equation; Indicates the same image point Adjustment parameter corrections for the satellite imagery in which it is located; Indicates the first The object coordinates of each connection point; Indicates the same image point The weights of the spatial intersection constraint equations;

[0103] Indicates the same image point The partial derivative matrix of the spatial intersection constraint equations with respect to the adjustment parameters; Indicates the same image point The partial derivative matrix of the spatial intersection constraint equations with respect to ground coordinates; Indicates the same image point The negative value of the current value vector of the fundamental adjustment equation;

[0104] Indicates the sign of partial derivatives; These are all model coefficients of the error correction model, and also adjustment parameters for each satellite image in the regional network adjustment.

[0105] Further using common object coordinates To constrain, establish the first The spatial intersection constraint equations for the connection points are expressed as follows:

[0106]

[0107] In the formula, Represents the residual vector of the spatial intersection constraint equation of the i-th connection point;

[0108] Let represent the partial derivative matrix of the spatial intersection constraint equations for the i-th connection point with respect to the adjustment parameters;

[0109] This represents the adjustment parameter correction number for the satellite image corresponding to the i-th tie point;

[0110] Let represent the partial derivative matrix of the spatial intersection constraint equation of the i-th connection point with respect to the ground coordinates;

[0111] Represents the object coordinates of the i-th connection point;

[0112] This represents the negative value of the current value vector of the fundamental adjustment equation for the i-th connection point;

[0113] Let represent the weights of the spatial intersection constraint equations for the i-th connection point.

[0114] However, due to the correlation between the geometric accuracy of satellite imagery in the plane and elevation directions, the spatial intersection constraint equations for the aforementioned connection points, even under good regional network connectivity conditions, cannot accurately express the geometric accuracy transmitted from other imagery within the network, thus failing to accurately detect ill-conditioned observations within the network. To address this problem, this invention introduces an additional prior accuracy elevation constraint on the spatial intersection constraint equations. While overcoming the coupling between the geometric accuracy of imagery in the plane and elevation directions, it establishes an adjustment parameter variance estimation model within the regional network that can quantitatively describe the propagation of accuracy for the elevation constraint, as detailed below:

[0115]

[0116] In the formula, Represents the residual vector of the elevation constraint equation for the i-th connection point; Represents the elevation correction; This represents the weight of the elevation constraint equation for the i-th connection point.

[0117] Step 3: Based on satellite imagery, construct the adjustment parameter constraint equations for the error correction model by setting control images.

[0118] To detect ill-conditioned observation areas in a regional network, this invention first selects any satellite image within the regional network as a control image. By directly adding a certain precision constraint to the adjustment parameters in the error correction model of the control image, the effect of controlling the image is achieved, giving it high prior positioning accuracy. Ill-conditioned observations are then detected by propagating this accuracy within the regional network.

[0119] Specifically, the steps to obtain the control image accuracy propagation variance estimation model include:

[0120] Designate any satellite image within the regional network as a control image, such as... Figure 2 As shown, the control image is uniformly divided into multiple grids, and a uniformly distributed virtual image point is generated at the center of each grid. The coordinates of the image point are... Since the role of control imagery is to provide a high initial propagation accuracy for the regional network, meaning the imaging model of the control imagery has high accuracy, the error correction model added to the image side of the rational function model... It should approach 0, that is Based on this constraint, using virtual image points as observations, we can obtain the first... The fundamental adjustment equation for each virtual image point:

[0121]

[0122] In the formula, This represents the basic adjustment equation for the image row direction corresponding to the kth virtual image point; This represents the basic adjustment equation for the image column direction corresponding to the kth virtual image point;

[0123] The adjustment parameter constraint equation for the error correction model constructed for the k-th virtual image point is as follows:

[0124] ;

[0125] ,

[0126] In the formula, Represents the residual vector of the adjustment parameter constraint equation for the k-th virtual image point;

[0127] This represents the partial derivative matrix of the adjustment parameter constraint equation for the k-th virtual image point with respect to the adjustment parameters;

[0128] This represents the adjustment parameter correction number for the control image corresponding to the k-th virtual image point;

[0129] This represents the negative value of the current value vector of the fundamental adjustment equation for the k-th virtual image point;

[0130] Let represent the weight matrix of the adjustment parameter constraint equation for the k-th virtual image point.

[0131] Step 4: Based on the adjustment parameter variance estimation model and adjustment parameter constraint equation, obtain the control image accuracy propagation variance estimation model.

[0132] The variance estimation model of adjustment parameters for all connection points within the integrated regional network and the constraint equations of adjustment parameters for all control images are expressed as follows:

[0133]

[0134] Combining the above equations, we can further refine the model for estimating the variance of image accuracy propagation, as shown in the following expression:

[0135]

[0136] , , , ,

[0137] , ,

[0138] In the formula, The set of residual vectors representing each constraint equation; Represents the residual vector of the spatial intersection constraint equations for all connection points; Represents the residual vector of the adjustment parameter constraint equations for all virtual image points; This represents the residual vector of the additional constraint equations introduced for the object coordinates of the connection points;

[0139] This represents the set of partial derivative matrices of each constraint equation with respect to the adjustment parameters; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the adjustment parameters; This represents the partial derivative matrix of the adjustment parameter constraint equations for all virtual image points with respect to the adjustment parameters;

[0140] This represents the set of partial derivative matrices of the constraint equations with respect to ground coordinates; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the ground coordinates. This represents the partial derivative matrix of the introduced additional constraint equations with respect to ground coordinates;

[0141] The set of negative values ​​representing the current value vector of each of the basic adjustment equations; This represents the negative value of the current value vector of the fundamental adjustment equations for all connection points; This represents the negative value of the current value vector of the fundamental adjustment equations for all virtual image points;

[0142] The set representing the weight matrices of each constraint equation; The weight matrix represents the spatial intersection constraint equations for all connection points; The weight matrix represents the adjustment parameter constraint equations for all virtual image points; This represents the weight matrix of the introduced additional constraint equations; Indicates elevation accuracy;

[0143] This represents the vector of adjustment parameter corrections for all images; This represents the correction vector for the object coordinates of all connection points.

[0144] Step 5: Based on the control image accuracy propagation variance estimation model and the error propagation law, calculate the theoretical accuracy of the adjustment parameters of the error correction model for all satellite images by calculating the correction cofactor matrix.

[0145] According to the error propagation law, the correction number The corresponding estimation accuracy is the same as the adjustment parameters in the corresponding error correction model. Therefore, the theoretical accuracy of the adjustment parameters in the error correction model can be obtained by estimating the accuracy of the correction.

[0146] According to the law of error propagation, the expression for the correction cofactor matrix is ​​as follows:

[0147]

[0148] In the formula, Represents the corrected cofactor matrix; This represents the first transition matrix; This represents the second transition matrix; Indicates the third transition matrix; This represents the fourth transition matrix.

[0149] Further based on the corrected covariance matrix Cofactor matrix Relationship, ,in The mean square error of the observations can be set to a constant in this invention based on the matching accuracy of the observations at the connection points. Since the adjustment parameters and their corrections have the same covariance matrix, their main diagonal elements can be used to represent the theoretical accuracy of the adjustment parameters in the error correction model, thereby obtaining the theoretical accuracy of the nth adjustment parameter in the error correction model. The expression is as follows:

[0150]

[0151] In the formula, This represents the theoretical accuracy of the nth adjustment parameter in the error correction model; Represents the corrected covariance matrix The element in the nth row and nth column; This represents the standard error of the observations; Represents the cofactor matrix The element in the nth row and nth column.

[0152] Step 6: Based on the law of error propagation, the model estimation accuracy of the error correction model corresponding to different locations on each satellite image is evaluated to detect whether there are ill-conditioned observation areas on the corresponding satellite image.

[0153] Whether the observation conditions are pathological can be quantitatively expressed with a certain level of precision through the error correction model of each satellite image.

[0154] Therefore, in step five, the theoretical accuracy of the adjustment parameters in the error correction model containing all images has been obtained. This is further converted into the model estimation accuracy in the error correction model corresponding to each satellite image, which is used to detect ill-conditioned observation areas.

[0155] 1. Calculation of model estimation accuracy in error correction model.

[0156] To estimate the accuracy of the adjustment parameters in the error correction model and to express the differences in observation conditions within a single satellite image, the same strategy as generating virtual image points for the control image is adopted here. Virtual image points are generated on each satellite image to be detected. The number of virtual image points can be the same as that of the control image. By evaluating the model estimation accuracy of the error correction model corresponding to the virtual image points, ill-conditioned observation areas are detected.

[0157] The specific steps are as follows:

[0158] For any given satellite image, it is uniformly divided into multiple grids, and a uniformly distributed virtual image point is generated at the center of each grid.

[0159] For any virtual image point, based on the error propagation law, the model covariance matrix of each virtual image point in each satellite image is estimated. The expression for the model covariance matrix of the m-th virtual image point is as follows:

[0160]

[0161] In the formula, This represents the model covariance matrix of the m-th virtual image point; Represents the corrected covariance matrix; The matrix representing the partial derivatives of the error correction model with respect to the adjustment parameters is expressed as follows:

[0162]

[0163] Finally, the model estimation accuracy of the error correction model corresponding to the m-th virtual image point is obtained. for:

[0164]

[0165] In the formula, This represents the model estimation accuracy of the error correction model corresponding to the row direction of the m-th virtual image point; This represents the model estimation accuracy of the error correction model corresponding to the column direction of the m-th virtual image point; Representing the model covariance matrix The element in the first row and first column; Show model covariance matrix The element in the last row and last column.

[0166] It is important to note here that the model covariance matrix... There are only two rows and two columns of elements.

[0167] 2. Pathological observation and detection.

[0168] For each satellite image, based on the model estimation accuracy of the error correction model for all virtual image points contained therein, the corresponding mean, maximum and minimum values ​​and root mean square error values ​​are statistically analyzed, and pathological observation and detection are carried out at two levels: between images and within images.

[0169] Between images: Since the model estimation accuracy of weakly connected images usually differs significantly from that of normally connected images, weakly connected satellite images can be detected through deliberate judgment or accuracy clustering. Here, the maximum or minimum value of the model estimation accuracy is used for accuracy clustering. If the maximum or minimum value of the model estimation accuracy for any satellite image is greater than a preset maximum or minimum value threshold, then an ill-conditioned observation region is identified on that satellite image.

[0170] The table below shows the maximum and minimum values ​​of the model estimation accuracy corresponding to the ten satellite images obtained using Example 1:

[0171]

[0172] It is quite obvious that the maximum value of the model estimation accuracy corresponding to the 6th satellite image is significantly different from that of other satellite images. The maximum value threshold can be set to 1 to filter out the 6th satellite image and identify the presence of ill-conditioned observation areas in the satellite image.

[0173] Inside the image: The difference in accuracy at different locations inside the image is mainly due to the uneven distribution of connection points. Therefore, the root mean square error and the maximum and minimum values ​​are used as indicators for detection. If the difference between the maximum and minimum values ​​and the mean values ​​of the model estimation accuracy corresponding to any satellite image is greater than a preset multiple of the root mean square error value, then it is determined that there is an ill observation area on the satellite image.

[0174] In this embodiment, the preset multiplier is three times.

[0175] Example 2

[0176] This embodiment 2 provides a system for detecting ill-conditioned observation regions in satellite image regional network adjustment, applicable to the method for detecting ill-conditioned observation regions in satellite image regional network adjustment of embodiment 1, including:

[0177] The data acquisition module is used to acquire multiple satellite images to be detected within the regional network, as well as multiple connection points corresponding to the satellite images;

[0178] The connection point module, for any connection point, constructs an adjustment parameter variance estimation model for elevation constraints based on a preset error correction model.

[0179] The control image module is used to construct the adjustment parameter constraint equations of the error correction model based on the satellite image and by setting the control image.

[0180] The control image accuracy propagation module is used to obtain the control image accuracy propagation variance estimation model based on the adjustment parameter variance estimation model and adjustment parameter constraint equation.

[0181] The ill-observed area detection module is used to obtain the theoretical accuracy of the adjustment parameters of the error correction model of all satellite images by calculating the correction cofactor matrix based on the error propagation law, according to the control image accuracy propagation variance estimation model; and then, based on the error propagation law, to detect whether there are ill-observed areas on the corresponding satellite images by evaluating the model estimation accuracy of the error correction model corresponding to different locations on each satellite image.

[0182] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0183] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0184] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0185] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0186] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment, characterized by: include: Acquire multiple satellite images to be detected within the regional network, as well as multiple connection points corresponding to the satellite images; For any connection point, based on the preset error correction model, an adjustment parameter variance estimation model for elevation constraints is constructed. Based on the satellite imagery, by setting control images, the adjustment parameter constraint equations of the error correction model are constructed. Based on the adjustment parameter variance estimation model and adjustment parameter constraint equation, the control image accuracy propagation variance estimation model is obtained. Based on the control image accuracy propagation variance estimation model, and based on the error propagation law, the theoretical accuracy of the adjustment parameters of the error correction model for all satellite images is obtained by calculating the correction cofactor matrix. Based on the law of error propagation, the model estimation accuracy of the error correction model corresponding to different locations on each satellite image is evaluated to detect whether there are ill-observed areas on the corresponding satellite image. The steps for obtaining the variance estimation model of the adjustment parameters for elevation constraints include: The regional network contains satellite images with multiple degrees of overlap, so one connection point corresponds to multiple image points with the same name; For any image point with the same name, the basic adjustment equation is constructed based on the preset error correction model and rational function model, and the spatial intersection constraint equation is obtained by parameter linearization. For any connection point, based on the spatial intersection constraint equations of all corresponding image points, the spatial intersection constraint equation of the connection point is established. Then, the elevation constraint equation with additional prior accuracy is introduced to obtain the variance estimation model of the adjustment parameters of the elevation constraint. The steps for constructing the adjustment parameter constraint equations of the error correction model based on the satellite imagery and by setting control images include: Set any satellite image within the regional network as a control image, and divide the control image into multiple grids evenly, generating a uniformly distributed virtual image point at the center of each grid. For any virtual image point, using the virtual image point as the observation value, by controlling the initial propagation accuracy of the control image, by constructing the basic adjustment equation of the virtual image point and linearizing the parameters, the adjustment parameter constraint equation of the virtual image point is obtained. Based on the adjustment parameter constraint equations for all virtual image points, the adjustment parameter constraint equations corresponding to the control image are obtained.

2. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 1, characterized in that, The expression for the variance estimation model of the adjustment parameters of the elevation constraint is as follows: ； In the formula, Represents the residual vector of the spatial intersection constraint equation of the i-th connection point; Let represent the partial derivative matrix of the spatial intersection constraint equations for the i-th connection point with respect to the adjustment parameters; This represents the adjustment parameter correction number for the satellite image corresponding to the i-th tie point; Let represent the partial derivative matrix of the spatial intersection constraint equation of the i-th connection point with respect to the ground coordinates; Represents the object coordinates of the i-th connection point; This represents the negative value of the current value vector of the fundamental adjustment equation for the i-th connection point; The weights of the spatial intersection constraint equations for the i-th connection point are represented. Represents the residual vector of the elevation constraint equation for the i-th connection point; Represents the elevation correction; This represents the weight of the elevation constraint equation for the i-th connection point.

3. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 1, characterized in that, The expression for the adjustment parameter constraint equation of the virtual image point is as follows: ； In the formula, Represents the residual vector of the adjustment parameter constraint equation for the k-th virtual image point; This represents the partial derivative matrix of the adjustment parameter constraint equation for the k-th virtual image point with respect to the adjustment parameters; This represents the adjustment parameter correction number for the control image corresponding to the k-th virtual image point; This represents the negative value of the current value vector of the fundamental adjustment equation for the k-th virtual image point; Let represent the weight matrix of the adjustment parameter constraint equation for the k-th virtual image point.

4. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 1, characterized in that, The expression for the control image accuracy propagation variance estimation model is as follows: ； , , , , , , ； In the formula, The set of residual vectors representing each constraint equation; Represents the residual vector of the spatial intersection constraint equations for all connection points; Represents the residual vector of the adjustment parameter constraint equations for all virtual image points; This represents the residual vector of the additional constraint equations introduced for the object coordinates of the connection points; This represents the set of partial derivative matrices of each constraint equation with respect to the adjustment parameters; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the adjustment parameters; This represents the partial derivative matrix of the adjustment parameter constraint equations for all virtual image points with respect to the adjustment parameters; This represents the set of partial derivative matrices of the constraint equations with respect to ground coordinates; This represents the partial derivative matrix of the spatial intersection constraint equations for all connection points with respect to the ground coordinates. This represents the coefficient matrix of the introduced additional constraint equations; The set of negative values ​​representing the current value vector of each of the basic adjustment equations; This represents the negative value of the current value vector of the fundamental adjustment equations for all connection points; This represents the negative value of the current value vector of the fundamental adjustment equations for all virtual image points; The set representing the weight matrices of each constraint equation; The weight matrix represents the spatial intersection constraint equations for all connection points; The weight matrix represents the adjustment parameter constraint equations for all virtual image points; This represents the weight matrix of the introduced additional constraint equations; Indicates elevation accuracy; This represents the vector of adjustment parameter corrections for all images; This represents the correction vector for the object coordinates of all connection points.

5. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 4, characterized in that, The expression for the corrected cofactor matrix is: ； In the formula, Represents the corrected cofactor matrix; This represents the first transition matrix; This represents the second transition matrix; Indicates the third transition matrix; This represents the fourth transition matrix.

6. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 1, characterized in that, The step of detecting whether there are ill-conditioned observation areas on the corresponding satellite image by evaluating the model estimation accuracy of the error correction model corresponding to different locations on each satellite image includes: Each satellite image to be detected is evenly divided into multiple grids, and a uniformly distributed virtual image point is generated at the center of each grid. For any virtual image point, based on the error propagation law, the model covariance matrix of the virtual image point is calculated according to the correction cofactor matrix, and the model estimation accuracy of the error correction model of the virtual image point is obtained. For each satellite image, based on the model estimation accuracy of the error correction model for all virtual image points included, the mean, maximum and minimum values ​​of the model estimation accuracy and the root mean square error value are statistically analyzed. If the maximum value of the model estimation accuracy of any satellite image is greater than a preset maximum value threshold, then it is determined that there is an ill-observed area on the satellite image. If the difference between the extreme value and the mean value of the model estimation accuracy of any satellite image is greater than a preset root mean square error value, then it is determined that there is an ill-conditioned observation area on the satellite image.

7. The method for detecting ill-conditioned observation areas in satellite imagery regional network adjustment according to claim 6, characterized in that, The expression for the model covariance matrix of the virtual image points is as follows: ； In the formula, This represents the model covariance matrix of the m-th virtual image point; This represents the partial derivative matrix of the error correction model with respect to the adjustment parameters; Represents the corrected covariance matrix; The expression for the model estimation accuracy of the error correction model for the virtual image points is as follows: ； In the formula, This represents the model estimation accuracy of the error correction model corresponding to the row direction of the m-th virtual image point; This represents the model estimation accuracy of the error correction model corresponding to the column direction of the m-th virtual image point; Representing the model covariance matrix The element in the first row and first column; Show model covariance matrix The element in the last row and last column.

8. A system for detecting ill-conditioned observation regions in satellite image regional network adjustment, applicable to the method for detecting ill-conditioned observation regions in satellite image regional network adjustment as described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to acquire multiple satellite images to be detected within the regional network, as well as multiple connection points corresponding to the satellite images; The connection point module, for any connection point, constructs an adjustment parameter variance estimation model for elevation constraints based on a preset error correction model. The control image module is used to construct the adjustment parameter constraint equations of the error correction model based on the satellite image and by setting the control image. The control image accuracy propagation module is used to obtain the control image accuracy propagation variance estimation model based on the adjustment parameter variance estimation model and adjustment parameter constraint equation. The ill-condition observation area detection module is used to obtain the theoretical accuracy of the adjustment parameters of the error correction model of all satellite images by calculating the correction cofactor matrix based on the error propagation law, according to the control image accuracy propagation variance estimation model. Based on the law of error propagation, the model estimation accuracy of the error correction model corresponding to different locations on each satellite image is evaluated to detect whether there are ill-conditioned observation areas on the corresponding satellite image.

Images