Control patch generation and geometric positioning method based on fusion of spaceborne laser and orthophoto images

Abstract

The present invention discloses a control slice generation and geometric positioning method that fuses satellite-borne laser and orthophoto images. The method comprises the following steps: first, screening the satellite-borne laser point cloud to obtain a high-quality laser point set; second, projecting the screened high-quality satellite-borne laser points onto the orthophoto image through coordinate transformation, and removing low-feature response points; finally, generating a control slice based on the orthophoto image and laser control points, and using the constructed control slice to perform geometric correction on the target image to achieve high-precision geometric positioning of the image. Unlike using only laser altimetry data for elevation control, the present invention achieves precise leveling and height control of remote sensing images by fusing satellite-borne laser data with orthophoto images, while simultaneously overcoming the drawback of limited DEM accuracy and being more suitable for geometric positioning of sub-meter images.

Description

Technical Field

[0001] The present invention relates to the field of photogrammetry and remote sensing, and in particular to a control slice generation and geometric positioning method that integrates satellite-borne laser and orthophoto images, wherein level control positioning combining satellite-borne laser and orthophoto images is a key technology of the present invention. Background Art

[0002] With the continuous development and widespread application of remote sensing technology, high-resolution remote sensing imagery is playing an increasingly important role in a variety of fields, including environmental monitoring, urban planning, precision agriculture, and disaster response. To ensure the effectiveness of these applications, the geometric positioning accuracy of remote sensing imagery is crucial. This is particularly true in scenarios such as large-scale geographic information system construction, precision mapping, and real-time monitoring. The geometric positioning accuracy of imagery directly impacts the reliability of analysis results and the accuracy of decision-making. However, due to complex terrain, changing ground environments, and the difficulty of acquiring traditional ground control points, achieving high-precision geometric positioning of remote sensing imagery has become a key challenge in current remote sensing image processing and applications.

[0003] Traditional methods for geometric positioning and correction of remote sensing images rely primarily on ground control points. However, acquiring high-quality ground control points often requires significant human and material resources, especially in complex terrain or remote areas, where obtaining ground control points becomes increasingly difficult and costly. Furthermore, due to the sparse and uneven distribution of ground control points, these methods often struggle to ensure accurate geometric correction when applied to large areas or complex terrain.

[0004] Therefore, it is necessary to develop a high-precision geometric positioning method for remote sensing images that is suitable for complex terrain or remote areas without ground control points, large areas or complex terrain. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for generating and geometrically positioning control slices by fusing satellite-borne laser and orthophoto data. The method is a method for generating control slices by fusing satellite-borne laser altimetry data with orthophoto data, and then using these slices as control data to achieve precise leveling and height control positioning of the remote sensing image to be processed. This method combines the high-precision ground coordinates of satellite-borne laser data with the optical characteristic information of the orthophoto data to produce a control slice containing a large number of laser three-dimensional control points, thereby achieving high-precision geometric correction and positioning of the remote sensing image. The method is suitable for high-precision geometric positioning of remote sensing images in complex terrain or remote areas without ground control points, and in large areas or complex terrain. The method addresses the difficulty in geometrically positioning remote sensing images caused by the difficulty and high cost of collecting ground control points in complex terrain or remote areas, and also addresses the difficulty in ensuring the accuracy of geometric correction when processing large areas or complex terrain.

[0006] In order to achieve the above-mentioned purpose, the technical solution of the present invention is: a control slice generation and geometric positioning method integrating satellite-borne laser and orthophoto images, characterized in that it includes the following steps:

[0007] Step 1: preprocessing of onboard laser data;

[0008] Screen the satellite-borne laser point cloud, remove noise and abnormal points, retain high-precision and high-confidence laser points, and obtain a high-quality laser point set;

[0009] Step 2: Generation of image control points;

[0010] Through coordinate transformation, the screened high-quality satellite-borne laser points are converted to orthophoto images, and the Harris response value of the corresponding image point projected on the image is calculated using the local neighborhood image to evaluate whether the image point projected on the image is a significant feature point, and eliminate points with low feature response.

[0011] Step 3, construct the control piece;

[0012] Generate control patches based on orthophotos and laser control points. The control patches contain precise ground coordinates and image coordinate information.

[0013] Step 4: Perform high-precision geometric correction on the image to be processed;

[0014] The constructed control piece is used to perform geometric correction on the target image to achieve high-precision geometric positioning of the image.

[0015] In the above technical solution, in step 1, the satellite-borne laser data preprocessing method is as follows:

[0016] First, the laser points are screened based on the attribute labels. The attribute labels of the laser points are screened to ensure high confidence in the data.

[0017] Then, the elevation points are re-screened in combination with the prior terrain data; the elevation errors of the laser points are further screened and eliminated with the prior data as a reference, thereby obtaining a more accurate and reliable laser point set.

[0018] In the above technical solution, the prior data includes SRTM (Shuttle Radar Topography Mission), global water mask data, etc.

[0019] In the above technical solution, laser point screening based on attribute tags includes the following steps:

[0020] 1) First, screening based on the reference DEM elevation labels;

[0021] In step 1, the laser points are filtered based on the elevation labels of the reference DEM. The elevation difference threshold is set based on the accuracy of the reference DEM. Typically, the elevation difference threshold is set to twice the root mean square error of elevation, but can be relaxed to three times in complex areas. The filtering formula is:

[0022] ΔH=|H dem -H p |<3σ (1)

[0023] Among them, H dem It is the elevation label of the reference DEM, which describes the corresponding reference DEM elevation; H p is the elevation of the corresponding laser elevation control point; σ is the root mean square error of the elevation of the reference DEM;

[0024] 2) Next, screening based on quality evaluation labels;

[0025] For publicly available satellite-borne laser point products, each point often has a quality evaluation label assigned when it is generated. To ensure laser point accuracy, the highest-quality laser points are typically selected (i.e., the highest-level laser points, each with a label that evaluates its quality, such as 1, 2, 3, or 4, with higher levels indicating better quality). Therefore, laser points with a level of 4 are selected as the highest-quality laser points). Different satellite-borne laser data, and even different-level products from the same satellite-borne laser, may have different label formats.

[0026] In the above technical solution, the elevation points are re-screened in combination with the prior terrain data, including the following steps:

[0027] 1) First, combine the screening of global water masks;

[0028] Considering that the laser points falling in water bodies often have low accuracy and are not suitable for image geometric correction, the global water mask data is used to filter out the laser points in the water;

[0029] 2) Next, combine with SRTM screening;

[0030] Assume that the geographic coordinates of a laser point are known to be (X, Y), and the affine transformation parameters of SRTM are X0, Y0, R x 、R y 、S x 、S y , the geographic coordinates can be converted to the pixel coordinates of the image using the following formula:

[0031]

[0032] In the above formula, (X, Y) must be consistent with the SRTM ground coordinate unit. If not, the geographic coordinates of the laser point must be converted to the SRTM ground coordinate system.

[0033] Through coordinate projection, each laser control point (x i ,y i ,z i ) reference height H i , and then calculate the elevation difference:

[0034] ΔH i =z i -H i (4 )

[0035] To further eliminate gross errors, a statistical frequency histogram analysis can be performed on the elevation differences of all laser points in the analysis area. Based on the frequency distribution, gross elevation error points can be filtered out. Based on the regional characteristics and laser point quality, an appropriate frequency threshold can be selected to consider laser points outside the two sides as gross errors.

[0036] In the above technical solution, in step 2, after the laser point is projected onto the image, the Harris response value is calculated using the local neighborhood image, and weak feature points with low response values ​​are eliminated;

[0037] The method for calculating the Harris response value R of each projection point includes the following steps:

[0038] First, calculate the gradient information of the image. Assuming the gray value function of the image is I(x,y), the gradient at the image pixel point (x,y) can be expressed as:

[0039] G=[I x I y ] (5)

[0040] in, are the gradients of the image in the x and y directions respectively; in order to calculate the feature changes in the local area, the autocorrelation matrix M can be constructed:

[0041]

[0042] Among them, w(x,y) is a weight function (usually a Gaussian function) used to weightedly calculate the gradient information within the neighborhood W;

[0043] According to the autocorrelation matrix M, the Harris corner response function R can be defined:

[0044] R = det(M) - k (trace(M)) 2 (7)

[0045] in, is the determinant of the matrix, is the trace of the matrix, k is an empirical constant, usually between [0.04, 0.06].

[0046] In the above technical solution, the method for determining whether a projection point is a weak feature point is:

[0047] According to the Harris response value R of each projection point, set the threshold R threshold =5%·R max , when the projection point satisfies R≥R threshold , then this projection point is an obvious feature point and is retained as an image control point; otherwise, this projection point is a weak feature point and is removed.

[0048] In the above technical solution, in step 3, for each image control point, first cut out an image block of size n×n pixels from the orthophoto with the control point as the center. The size of n depends on the orthophoto resolution, ensuring that the image block is larger than 1km. 2 ;

[0049] Each control piece is composed of the image control point, the corresponding image block, and the image coordinates and object coordinates of the image control point.

[0050] In the above technical solution, in step 4, based on the control piece generated in step 3, the control piece is matched with the image to be processed by an image matching method. The matching method is as follows:

[0051] First, the same-name feature points between the control image and the image to be processed are identified and extracted through an automatic or semi-automatic image matching algorithm;

[0052] Then, based on these matching points, a regional network is constructed and rigorous adjustment is performed;

[0053] Then, during the adjustment process, robust statistical methods are used to dynamically remove image control points with large errors or poor matching quality to reduce the impact of errors on model accuracy.

[0054] Finally, through adjustment calculations, high-precision geometric positioning image results are obtained. It should be noted that satellite-borne lasers mainly provide elevation control, while orthophotos provide texture features and plane control.

[0055] The present invention has the following advantages:

[0056] (1) Provides a new image geometric positioning solution; the present invention achieves high-precision geometric positioning of images by fusing satellite-borne laser data and orthophotos and using the generated control slices, making full use of the advantages of multi-source data and further expanding the application scope of laser altimetry data (suitable for high-precision geometric positioning of remote sensing images in complex terrain or remote areas without ground control points, large areas or complex terrain conditions). By fusing with orthophotos to generate control slices, horizontal height control is achieved, rather than just elevation control.

[0057] (2) Compared with the positioning method based on orthophoto + DEM, the method of the present invention overcomes the defect of limited DEM accuracy (the geometric positioning accuracy of the existing technology can only reach the meter level), and is more suitable for the geometric positioning of sub-meter images (the geometric positioning accuracy of the method of the present invention is one order of magnitude higher than the positioning accuracy of the existing technology), ensuring the geometric accuracy of the entire image (the plane accuracy of the positioning of the present invention can be achieved within 1 pixel, and the elevation accuracy can be achieved to about 0.5 meters) and consistency (mainly ensuring the geometric consistency within the entire image and ensuring that the local accuracy and global accuracy of the image are relatively consistent due to the rich image control points);

[0058] (3) This method has strong iterativeness. As the planar accuracy of spaceborne lidar and the positioning accuracy of acquired orthophotos continue to improve, the positioning effect of the control piece produced based on the method of the present invention will also be continuously optimized and improved. Therefore, with the advancement of technology and the improvement of data quality, the positioning accuracy and application scope of this method will only continue to increase, ensuring its long-term effectiveness and competitiveness in future applications.

[0059] (4) The present invention achieves overall image correction by integrating satellite-borne laser altimetry data with orthophotos. This overcomes the limitation of the prior art that only uses laser altimetry data for correction, which can only perform elevation correction on a very small area of ​​the image (the area with the laser point) and cannot achieve comprehensive correction of the entire image.

[0060] (5) The method of the present invention uses open source orthophotos, which are sub-meter images themselves, and further improves the geometric positioning effect by integrating satellite-borne laser altimetry data. It overcomes the problem that the existing geometric positioning method based on open source DEM is not suitable for high-precision geometric positioning of sub-meter images due to the limited resolution and accuracy of open source DEM. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is the overall technical flow chart of the present invention;

[0062] Figure 2 A high-quality laser point distribution map extracted from a certain area in an embodiment of the present invention;

[0063] Figure 3 This is a partial control slice file image obtained by fusing satellite-borne laser altimetry and orthophotos in an embodiment of the present invention;

[0064] Figure 4 is a distribution map of image control points for image correction to be processed in an embodiment of the present invention;

[0065] Figure 5 This is a distribution map of ground control points used for image geometric positioning accuracy evaluation in an embodiment of the present invention. DETAILED DESCRIPTION

[0066] The following detailed description of the embodiments of the present invention is given in conjunction with the accompanying drawings, which do not limit the present invention but are merely examples. The description makes the advantages of the present invention clearer and easier to understand.

[0067] Satellite-borne laser altimetry data is gradually becoming an important control source for the geometric correction of remote sensing images due to its high-precision elevation information. By combining satellite-borne laser data with high-resolution optical images, the geometric positioning accuracy of the image can be significantly improved, thereby reducing dependence on traditional ground control points, especially in areas with complex terrain or where ground control points are difficult to obtain. However, in current technical applications, satellite-borne laser altimetry data is usually only used for elevation control, and its potential in all-round geometric correction has not been fully tapped. Research on how to organically integrate satellite-borne laser data with orthophotos to achieve high-precision horizontal and vertical positioning and correction of remote sensing images is still in its preliminary stages. Existing methods have problems such as low data utilization and limited fusion accuracy in the fusion process, and have failed to give full play to the unique advantages of satellite-borne laser data in improving image correction accuracy;

[0068] Control sheets are a new type of photogrammetric control data source. They refer to image data processed through aerial triangulation and contain precise geometric positioning parameters and encrypted point information. Because they contain the ground coordinates of the encrypted points and corresponding image data, they can be used directly as control information for image correction. Spaceborne laser altimetry data only contains ground coordinates and lacks corresponding image position information, while orthophotos contain ground texture information. By geometrically registering and fusing the two, and utilizing both the orthophoto texture information and the spaceborne laser geometric coordinate information, high-precision control sheets can be generated. However, how to fuse spaceborne laser points and orthophotos to generate control sheets remains an urgent problem.

[0069] The present invention provides a method that can fuse satellite-borne laser and orthophoto to achieve control piece generation and high-precision geometric positioning of images. The core idea is to use satellite-borne laser data and open-source orthophoto as the basic data sources, use the attribute labels of the satellite-borne laser data itself and prior data to screen and obtain high-quality laser point data, use the orthophoto as the base map and satellite-borne laser to align and extract high-feature response image control points, generate control pieces by fusing satellite-borne laser data and open-source orthophoto, use the generated control pieces to achieve high-precision level and height control positioning of the image to be processed, and then use them as control data to achieve accurate level and height control positioning of the remote sensing image to be processed. The specific method is as follows: first, the satellite-borne laser point cloud is screened to obtain a high-quality laser point set; second, through coordinate transformation, the screened high-quality satellite laser points are projected onto the orthophoto, and low-feature response points are eliminated; finally, a control piece is generated based on the orthophoto and laser control points, and the constructed control piece is used to perform geometric correction on the target image to achieve high-precision geometric positioning of the image. Unlike using only laser altimetry data for elevation control, the method of the present invention achieves precise leveling and elevation control of remote sensing images by fusing satellite-borne laser data with orthophotos, while overcoming the defect of limited DEM accuracy. It is more suitable for the geometric positioning of sub-meter images, and is suitable for high-precision geometric correction and positioning of remote sensing images in complex terrain or remote areas without ground control points, large areas or complex terrain.

[0070] The present invention is mainly implemented through the following technical solutions: a control slice generation and geometric positioning method integrating satellite-borne laser and orthophoto images, comprising the following steps:

[0071] Step 1: Preprocessing of spaceborne laser data: Screen the spaceborne laser point cloud to remove noise and outliers, retain high-precision, high-confidence laser points, and obtain a high-quality laser point set.

[0072] Step 2: Generate image control points. Through coordinate transformation, the screened high-quality satellite laser points are converted to orthophoto images, and their Harris response values ​​are calculated using local neighborhood images to evaluate whether they are obvious feature points and eliminate points with low feature response.

[0073] Step 3: Construct a control patch. Generate a control patch based on the orthophoto and laser control points. The control patch contains precise ground coordinates and image coordinate information.

[0074] Step 4: Perform high-precision geometric correction on the image to be processed. Use the constructed control slice to perform geometric correction on the target image to achieve high-precision geometric positioning of the image.

[0075] In step 1, the preprocessing of spaceborne laser data consists of two main steps: attribute label-based screening and elevation point screening based on prior terrain. First, the laser points are screened based on their attribute labels to ensure high data confidence. Then, using prior data such as SRTM (Shuttle Radar Topography Mission) and global water mask data as a reference, gross elevation errors in the laser points are further screened and eliminated, resulting in a more accurate and reliable laser point set.

[0076] (1) Laser point screening based on attribute tags

[0077] Referring to ICESat-2's laser point data preprocessing technology, laser points typically have multiple attribute tags, including control point quality confidence tags. The screening process is as follows:

[0078] 1) Filtering based on reference DEM elevation labels

[0079] In step 1, the laser points are filtered based on the elevation labels of the reference DEM. The elevation difference threshold is set based on the accuracy of the reference DEM. Typically, the elevation difference threshold is set to twice the root mean square error of elevation, but can be relaxed to three times in complex areas. The filtering formula is:

[0080] ΔH=|H dem -H p |<3σ (1)

[0081] Among them H dem It is the elevation label of the reference DEM, which describes the corresponding reference DEM elevation; H p is the elevation of the corresponding laser elevation control point; σ is the root mean square error of the elevation of the reference DEM.

[0082] 2) Screening based on quality evaluation labels

[0083] For publicly available satellite-borne laser point products, each point often has a quality evaluation label given when it is generated. Usually, in order to ensure the accuracy of the laser points, the highest quality laser points are selected.

[0084] It's important to note that different satellite laser data, and even different levels of products from the same satellite laser, may have different label formats. For example, the ATL03 and ATL08 products from ICESat-2 use the same primary attribute labels as described in this article, while the ATL08 includes multiple labels such as cloud coverage and slope for segmented photons. Therefore, this article will only use the most common attribute labels as an example.

[0085] (2) Re-screening based on prior data

[0086] Due to the large number of satellite-borne laser points, there are certain errors in the attribute labels provided by data providers when processing laser points. On this basis, some prior data is used for secondary screening to further ensure the quality of laser points.

[0087] 1) Combined with the screening of global water masks

[0088] Considering that the laser points falling in water bodies often have low accuracy and are not suitable for image geometric correction, the global water mask data is used to filter out the laser points in the water.

[0089] 2) Combined with SRTM screening

[0090] For SRTM TIFF image files, they contain geographic reference information, such as geographic coordinate system (such as WGS 84) and image affine transformation parameters. Affine transformation parameters are used to describe the mapping relationship between image coordinates and geographic coordinates, usually including six parameters: two parameters for translation (X0 and Y0, respectively, the X and Y coordinates of the upper left corner of the image), two parameters for rotation (R x and R y , which is usually close to 0 because images are usually aligned with the coordinate axes), and two parameters for scaling (S x and S y , respectively the horizontal and vertical pixel resolutions of the image).

[0091] Assume that the geographic coordinates of a laser point are known to be (X, Y), and the affine transformation parameters of SRTM are X0, Y0, R x 、R y 、S x 、S y , the geographic coordinates can be converted to the pixel coordinates of the image using the following formula:

[0092]

[0093] In the above formula, (X,Y) must be consistent with the SRTM ground coordinate system. If not, the laser point geographic coordinates must be converted to the SRTM ground coordinate system.

[0094] Through coordinate projection, each laser control point (x i ,y i ,z i ) reference height H i , and then calculate the elevation difference:

[0095] ΔH i =z i -H i (4).

[0096] To further eliminate gross errors, a statistical frequency histogram analysis can be performed on the elevation differences of all laser points within the analysis area. These gross elevation errors can then be filtered out based on the frequency distribution. Because the photon counter generates a large number of laser points and the density of these points varies across different regions, simply using zero values ​​on either side of the main frequency as a filtering criterion is inappropriate. A more appropriate approach is to set a frequency limit for gross errors based on a certain percentage of the total photon count. Specifically, an appropriate frequency threshold can be selected based on regional characteristics and laser point quality, with laser points outside these two sides considered gross errors.

[0097] In step 2, the high-quality laser point obtained in step 1 is projected onto the orthophoto using the coordinate transformation of formulas (2) and (3). After the laser point is projected onto the image, the Harris response value is calculated using the local neighborhood image, and weak feature points with low response values ​​are eliminated. First, the gradient information of the image is calculated. Assuming that the gray value function of the image is I(x,y), the gradient at the image pixel point (x,y) can be expressed as:

[0098] G=[I x I y ] (5)

[0099] in, are the gradients of the image in the x and y directions respectively. In order to calculate the feature changes in the local area, the autocorrelation matrix M can be constructed:

[0100]

[0101] Among them, w(x,y) is a weight function (usually a Gaussian function) used to weightedly calculate the gradient information within the neighborhood W.

[0102] According to the autocorrelation matrix M, the Harris corner response function R can be defined:

[0103] R = det(M) - k (trace(M)) 2 (7)

[0104] in, is the determinant of the matrix, is the trace of the matrix, k is an empirical constant, usually between [0.04, 0.06]. According to the Harris response value R of each projection point, the threshold R is set threshold =5%·R max , when the projection point satisfies R≥R threshold It is retained as an image control point.

[0105] In step 3, for each image control point, first cut out an image block of size n × n pixels from the orthophoto with the control point as the center. The size of n depends on the orthophoto resolution, ensuring that the image block is larger than 1km. 2 Each control slice is composed of the image control point, the corresponding image block, and the image coordinates and object coordinates of the image control point.

[0106] In step 4, based on the control piece generated in step 3, it is matched with the image to be processed by the image matching method. First, the feature points with the same name between the control piece and the image to be processed are identified and extracted through an automatic or semi-automatic image matching algorithm. Then, based on these matching points, a regional network is constructed and a strict adjustment process is performed. During the adjustment process, a robust statistical method is used to dynamically eliminate image control points with large errors or poor matching quality to reduce the impact of errors on model accuracy. Finally, through adjustment calculation, high-precision geometric positioning image results are obtained. It should be noted that the satellite-borne laser mainly provides elevation control, while the orthophoto provides texture features and plane control.

[0107] Example

[0108] The present invention will now be described in detail by taking the application of the present invention to geometric positioning of images in a remote area as an example, which can also provide guidance for the application of the present invention to geometric positioning of images in other areas.

[0109] In a remote area of ​​this embodiment, it is difficult and costly to obtain ground control points, and it is not possible to use ground control points for geometric positioning of images.

[0110] This example selected the ICESat-2 ATL03 laser point data of a certain area and the 1-meter resolution open source orthophotos for the generation of control slices. At the same time, the IKONOS satellite 1-meter resolution images and aerial ground control points of the area were collected to verify the geometric positioning accuracy of the method. Since the spacing between each set of laser tracks of ICESat-2 is 3 kilometers and the track spacing is about 1 kilometer, in order to ensure that the satellite-borne laser data fully covers the area, this example downloaded a total of 21 laser point data sets from January 2022 to October 2023. The overall process of geometric positioning of images using the present invention in this embodiment is shown in the attached figure. Figure 1 As shown, the core steps include the following aspects:

[0111] Step 1: Preprocess the 21 downloaded ATL03 satellite-borne laser point datasets. The initial data contained 7,846,683 laser points. For ATL03 data products, the laser point attributes only include quality confidence labels. By filtering for this label, 318,912 laser points were retained. Secondary filtering was performed using the collected 30-meter resolution SRTM DEM and water mask data, retaining 221,743 laser points. Finally, to meet the subsequent control point density requirements, a 50×50 meter grid was used for thinning, ultimately retaining 3,351 laser points.

[0112] Step 2: The high-quality satellite laser points selected in step 1 are projected onto the orthophoto using a general imaging model. Subsequently, the Harris operator is used to detect and remove low-feature response points, and finally 2163 image control points are extracted. Their spatial distribution is shown in the attached figure. Figure 2 As shown ( Figure 2 The red dots distributed on the image are all image control points).

[0113] Step 3: Generate a control slice based on the orthophoto and laser control points. The control slice contains accurate ground coordinates and image coordinate information. The generated control slice file is as follows: Figure 3 As shown (in Figure 3 In the Hobart txt file, the object space coordinates and image space coordinates of the image control points in each control slice are recorded; Hobart1-Hobart23 are the partial images corresponding to the control slice itself).

[0114] Step 4: Match the generated control slices with the IKONOS satellite image. The initial number of control slices was 2163. Since some image control points were outside the range of the IKONOS image, the actual number of control slices available for matching was 1979. Finally, 341 control slices were successfully matched. Geometric correction was performed based on the successfully matched control slices, and gross error points were eliminated through iterative calculation. Finally, 278 accurate image control points were obtained, as shown in the attached figure. Figure 4 As shown ( Figure 4 The blue dots in the image are precise image control points).

[0115] For the final geometrically positioned image, the example uses the following Figure 5 The accuracy of the ground control points is evaluated using the distribution shown in the figure ( Figure 5 The triangle points in the figure represent ground control points, and the numbers in the lower right corner of the triangle points are the numbers of the ground control points. The schemes are compared with other schemes, and the accuracy comparison is shown in Table 1 below.

[0116] Table 1 Evaluation and comparative analysis of image geometric positioning accuracy of the method of the present invention

[0117]

[0118] The accuracy comparison experiments were divided into four groups. Experiment 1 performed geometric positioning processing on the IKONOS image without any control data. Experiment 2 used STRM data as control data to perform positioning processing on the IKONOS image. Both experiments 1 and 2 used 17 ground control points as checkpoints. The results showed that the root mean square error of both the plane and elevation was within 2.5 meters.

[0119] Experiment 3 simulates a scenario with ground control points, using six ground control points for geometric correction and the remaining 11 ground control points as checkpoint data. Both the plane error and elevation error are controlled within 0.25 meters. After processing the IKONOS image using the control slice generated by the method of the present invention (Experiment 4), the plane accuracy can reach sub-meter level (within one pixel), and the elevation accuracy is not much different from that of the ground control points, demonstrating that the present invention can achieve high-precision geometric correction and positioning of remote sensing images and has extremely high engineering practicality. For the industry, the plane positioning accuracy is within one pixel, and the elevation reaches the decimeter level, which has met the higher accuracy requirements and fully demonstrated the effectiveness of the present invention. At the same time, the geometric positioning accuracy of the existing technology (Experiment 1, Experiment 2) is about 2.5 meters, while the geometric positioning accuracy of the method of the present invention is at the decimeter level, which is one order of magnitude higher than the geometric positioning accuracy of the existing technology. It can be seen that compared with the existing technology, the present invention not only greatly improves the geometric positioning accuracy when performing remote sensing image geometric positioning in complex terrain or remote areas, large areas or complex terrain, but also saves costs (the present invention does not require the setting of ground control points), improves work efficiency, and expands the applicability of remote sensing image geometric positioning in complex terrain or remote areas, large areas or complex terrain areas without ground control points.

[0120] The present invention primarily addresses the difficulty and high cost of collecting ground control points in complex terrain or remote areas. It proposes a method for generating control slices by fusing satellite-borne laser data with orthophotos, and for performing precise leveling and height control positioning of the image to be processed based on the generated control slices. This method achieves high-precision geometric correction and positioning of remote sensing images by combining the high-precision elevation information of satellite-borne laser data with the optical characteristics of orthophotos, making full use of the complementarity between laser data and image data, and effectively improving the positioning accuracy and geometric correction effect of remote sensing images in complex terrain areas. The combination of satellite-borne laser and orthophoto leveling and height control positioning is the key technology of the present invention.

[0121] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.

[0122] Other parts not described belong to the prior art.

Claims

1. A control image generation and geometric positioning method integrating spaceborne laser and orthophoto images, characterized by: The following steps are included: Step 1: preprocessing of onboard laser data; Screen the satellite-borne laser point cloud, remove noise and abnormal points, retain high-precision and high-confidence laser points, and obtain a high-quality laser point set; Step 2: Generation of image control points; Through coordinate transformation, the screened high-quality satellite-borne laser points are converted to orthophoto images, and the Harris response value of the corresponding image point projected on the image is calculated using the local neighborhood image to evaluate whether the image point projected on the image is a significant feature point, and eliminate points with low feature response. Step 3, construct the control piece; Generate control patches based on orthophotos and laser control points. The control patches contain precise ground coordinates and image coordinate information. Step 4: Perform high-precision geometric correction on the image to be processed; The constructed control piece is used to perform geometric correction on the target image to achieve high-precision geometric positioning of the image.

2. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 1 is characterized by: In step 1, the onboard laser data preprocessing method is as follows: First, laser points are screened based on attribute labels to ensure high confidence in the data; Then, the elevation points are re-screened in combination with the prior terrain data; the elevation errors of the laser points are further screened and eliminated with the prior data as a reference, thereby obtaining a more accurate and reliable laser point set.

3. The control slice generation and geometric positioning method for integrating satellite-borne laser and orthophoto images according to claim 2 is characterized by: Prior data include SRTM and global water mask data.

4. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 3 is characterized by: Laser point screening based on attribute tags includes the following steps: 1) Based on the reference DEM elevation label, the screening formula is: ΔH=|H dem -H p |<3σ (1) Among them, H dem It is the elevation label of the reference DEM, which describes the corresponding reference DEM elevation; H p is the elevation of the corresponding laser elevation control point; σ is the root mean square error of the elevation of the reference DEM; 2) Screening based on quality evaluation labels In order to ensure the accuracy of the laser points, the highest quality laser points are selected.

5. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 4 is characterized in that: Combined with the prior terrain data, the elevation points are screened again, including the following steps: 1) Combined with the screening of global water masks Use global water mask data to filter out laser points in water; 2) Combined with SRTM screening Assume that the geographic coordinates of a laser point are (X, Y), and the affine transformation parameters of SRTM are X0, Y0, R x 、R y 、S x 、S y , the geographic coordinates are converted to the pixel coordinates of the image using the following formula: In the above formula, (X, Y) must be consistent with the SRTM ground coordinate unit; If they are inconsistent, first convert the geographic coordinates of the laser point to the ground coordinate system where the SRTM is located; Through coordinate projection, each laser control point (x i ,y i , z i ) reference height H i , and then calculate the elevation difference: ΔH i =z i -H i (4)。 6. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 5, characterized in that: The method to determine whether the projection point is a weak feature point is: According to the Harris response value R of each projection point, set the threshold R threshold =5%·R max , when the projection point satisfies R≥R threshold , then this projection point is an obvious feature point and is retained as an image control point; otherwise, this projection point is a weak feature point and is removed.

7. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 6, characterized in that: In step 3, for each image control point, first cut out an image block of size n × n pixels from the orthophoto with the control point as the center. The size of n depends on the orthophoto resolution. Ensure that the image block is larger than 1km. 2 ; Each control piece is composed of the image control point, the corresponding image block, and the image coordinates and object coordinates of the image control point.

8. The control slice generation and geometric positioning method for integrating spaceborne laser and orthophoto images according to claim 7, characterized in that: In step 4, based on the control piece generated in step 3, the control piece is matched with the image to be processed by an image matching method. The matching method is as follows: First, the same-name feature points between the control image and the image to be processed are identified and extracted through an automatic or semi-automatic image matching algorithm; Then, based on these matching points, a regional network is constructed and rigorous adjustment is performed; Then, during the adjustment process, robust statistical methods are used to dynamically remove image control points with large errors or poor matching quality to reduce the impact of errors on model accuracy. Finally, through adjustment calculation, high-precision geometric positioning image results are obtained.

Images