Information extraction method for multi-source remote sensing data

Abstract

The invention relates to the technical field of multi-source remote sensing data extraction, and provides an information extraction method for multi-source remote sensing data, which comprises the following steps of: 1, acquiring high-altitude image data of a target area, and extracting a plurality of matching blocks from the high-altitude image data; 2, acquiring unmanned aerial vehicle image data of a target area, and extracting a plurality of feature blocks from the unmanned aerial vehicle image data; step 3, obtaining a measurement coordinate of each matching block in the high-altitude image data to obtain first coordinate information, and obtaining a measurement coordinate of each feature block in the unmanned aerial vehicle image data to obtain second coordinate information; and step 5, constructing a probability model for mutual registration of each matching block and each feature block for each matching group, and performing tensor decomposition on the probability model to obtain registration information corresponding to the matching blocks and the feature blocks one by one. According to the method, the multi-source remote sensing data fusion precision is remarkably improved through a grouping probabilistic registration mechanism, and the restriction of the positioning precision difference on the multi-source data fusion is broken through.

Description

Technical Field

[0001] The present application relates to the technical field of multi-source remote sensing data extraction, and in particular to a method for extracting information from multi-source remote sensing data. Background Art

[0002] The contents of this section merely provide background information related to the present application and may not constitute prior art.

[0003] Multi-source remote sensing data refers to heterogeneous datasets acquired through collaborative observations of the same area using heterogeneous sensors (such as optical, radar, infrared, and hyperspectral sensors) or cross-platform carriers (satellites, aircraft, and ground stations). By integrating multidimensional information such as spatial, spectral, temporal, and radiometric resolution, it effectively overcomes the inherent limitations of a single data source (such as optical imagery's susceptibility to cloud and fog interference and radar data's lack of spectral detail), significantly improving the accuracy of ground object classification, dynamic monitoring capabilities, and the reliability of environmental parameter inversion. It has been widely used in fields such as resource surveys, disaster warning, and precision agriculture.

[0004] However, the core challenge of multi-source data processing lies in the precise matching and fusion of cross-source data. Existing solutions typically rely on direct alignment of geographic coordinates, forcibly overlaying data from different sources at a preset location to construct a 3D scene model. This approach suffers from fundamental flaws: the positioning accuracy of different remote sensing systems varies significantly (for example, the coordinate deviation between satellite imagery and drone radar), and current technology struggles to adaptively match these differences. Consequently, it is impossible to extract effective spatial correspondence from heterogeneous data. Furthermore, fusion errors are propagated to downstream applications, severely degrading the geometric and texture accuracy of 3D scene rendering. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a method for extracting information from multi-source remote sensing data. The method for extracting information from multi-source remote sensing data disclosed in this application can solve the technical problems raised by the background technology.

[0006] The purpose of this application is achieved through the following technical solutions:

[0007] A method for extracting information from multi-source remote sensing data, comprising:

[0008] Step 1: Obtain high-altitude image data of the target area and extract several matching blocks from the high-altitude image data;

[0009] The matching block at least includes a number of extreme value points exceeding a preset threshold;

[0010] Step 2: Obtain UAV image data of the target area and extract several feature blocks from the UAV image data;

[0011] The feature block at least includes a number of extreme value points exceeding a preset threshold;

[0012] Step 3: Obtain the measured coordinates of each matching block in the high-altitude image data to obtain the first coordinate information, and obtain the measured coordinates of each feature block in the drone image data to obtain the second coordinate information;

[0013] Step 4: Perform fuzzy matching on the matching blocks and the feature blocks according to the first coordinate information and the second coordinate information, and merge the matching blocks and the feature blocks whose coordinate positions are not more than a preset threshold into one matching group;

[0014] Step 5: For each matching group, a probability model for the mutual registration of each matching block and feature block is constructed, and the probability model is subjected to tensor decomposition to obtain the one-to-one registration information of the matching block and feature block.

[0015] This method significantly improves the accuracy of multi-source remote sensing data fusion through a grouped probabilistic registration mechanism. First, elevation image matching blocks and drone image feature blocks are dynamically grouped based on spatial proximity to avoid the positioning errors associated with traditional hard coordinate matching. A probabilistic model of the spatial relationship between blocks is then constructed within each group. Tensor decomposition is used to calculate the optimal matching probability, converting the accuracy differences between satellite elevation data (low positioning accuracy) and drone imagery (high resolution but weak positioning) into computable probability weights, enabling adaptive calibration of heterogeneous data. The resulting high-confidence registration information significantly reduces matching errors, directly improving the geometric consistency and texture fidelity of 3D scene reconstruction, and overcoming the constraints of positioning accuracy differences on multi-source data fusion.

[0016] Furthermore, the real-world physical size corresponding to the matching block is consistent with the real-world physical size corresponding to the feature block; and the size of the matching block is set according to the matching accuracy.

[0017] In the technical solution provided in the present application, the real-world physical size corresponding to the matching block is consistent with the real-world physical size corresponding to the feature block, which can ensure that the number of matching blocks and feature blocks is equal within the same range, facilitating a one-to-one correspondence between the matching blocks and the feature blocks.

[0018] Furthermore, the number of extreme points in the feature block and the matching block satisfies the following conditions:

[0019] ;

[0020] x, y represent the center coordinates of the feature block or matching block, i and y i represents the coordinates of the i-th extreme point, K represents the kernel function, n represents the total number of extreme points, and h represents the bandwidth parameter.

[0021] In the technical solution provided in the present application, the distribution of extreme points in the feature block and the matching block is restricted, so as to effectively ensure that there are a sufficient number of extreme points in the matching block and the feature block to match each other, increase the accuracy of subsequent matching blocks and feature blocks when matching each other, and use the kernel function to evaluate the density of extreme points, thereby ensuring that there are a sufficient number of extreme points in the matching block and the feature block to match each other.

[0022] Furthermore, the first coordinate information is satellite positioning information obtained by the satellite, and the second coordinate information is radar positioning information obtained by the UAV; the sizes of the feature block and the matching block are determined based on the minimum measurement accuracy of the satellite positioning information.

[0023] In the technical solution provided in this application, the sizes of the feature blocks and the matching blocks are determined based on the minimum measurement accuracy of the satellite positioning information, which can ensure that the real sizes of the physical world corresponding to the feature blocks and the matching blocks can match each other, avoiding the low accuracy of subsequent matching due to the difference in actual sizes between the two.

[0024] In some possible embodiments, the method for obtaining the matching block includes the following steps:

[0025] Step 11: Select the mountain boundaries, river boundaries, depression boundaries, and road boundaries from the aerial image data;

[0026] Step 12: Extract several extreme points from the river boundary, depression boundary, and road boundary to generate a first extreme point set;

[0027] Step 13: Extract extreme points from the remaining areas of the high-altitude image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a third extreme point set;

[0028] Step 14: Establishing a correspondence between each first extreme point set and each extreme point in the third extreme point set;

[0029] Step 15: Extract a number of matching blocks that meet the number conditions of extreme points from the elevation image based on the first extreme point set.

[0030] This application significantly improves the accuracy and robustness of feature block matching by innovatively constructing "matching block information" that includes multi-dimensional relative position relationships. The core of this approach is to utilize the presence of a large number of extreme points at specific boundary locations to construct matching blocks that meet the requirements. This allows subsequent parts to be re-matched based on the corresponding relationships within the mesh structure after completing the matching of the corresponding boundaries, thus ensuring matching accuracy.

[0031] Furthermore, the method for obtaining the feature block includes the following steps:

[0032] Step 21: Select the mountain boundaries, river boundaries, depression boundaries, and road boundaries from the drone image data;

[0033] Step 22: extracting a number of extreme points from the river boundary, the depression boundary, and the road boundary to generate a second extreme point set;

[0034] Step 23: Extract extreme points from the remaining areas of the UAV image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a fourth extreme point set;

[0035] Step 24: extracting a number of feature blocks that meet the number conditions of extreme points from the elevation image based on the second extreme point set.

[0036] In the technical solution provided in this application, the method for extracting feature blocks is roughly the same as that for extracting matching blocks. Therefore, when the feature blocks and matching blocks are matched with each other, the special boundaries are matched with each other first. After the special boundaries are successfully matched with each other, the internal features are matched with each other. In practice, this can increase the matching accuracy.

[0037] Step 4 includes the following steps:

[0038] Step 41: Obtain mountain boundaries, river boundaries, depression boundaries, and road boundaries from the high-altitude image data, and use the mountain boundaries, river boundaries, depression boundaries, and road boundaries as matching areas;

[0039] Step 42: Divide the matching area into a plurality of matching grids based on a preset grid, and set the size of the matching grid according to a preset threshold;

[0040] Step 43: Merge the matching blocks and feature blocks in the dimensionality reduction matching grid into one matching group.

[0041] In the technical solution provided in the present application, a preset grid is used to divide the matching groups, which is highly accurate in practice and can ensure that the matching blocks and feature blocks in different areas are matched with each other.

[0042] Furthermore, step 5 includes the following steps:

[0043] Step 51: The similarity between the feature block and the matching block is used as the first matching element, the similarity between the texture features of the feature block and the matching block is used as the second matching element, and the ratio of the distance between the feature block and the reference object to the distance between the matching block and the reference object is used as the third matching element; the reference object is a specific physical marker pre-set on the target area;

[0044] Step 52: Establish a probability model based on the first matching element, the second matching element, and the third matching element; perform tensor decomposition on the probability model to reduce the model dimension.

[0045] This application significantly improves the matching accuracy and efficiency by introducing spatial context information through innovative multi-dimensional matching element fusion and model optimization strategies: On the basis of traditional similarity (first matching element) and texture similarity (second matching element), the direction difference between feature blocks and matching blocks relative to the common reference object (third matching element) is newly added, making full use of the implicit spatial relationship features to make the matching more comprehensive and robust.

[0046] Furthermore, step 52 specifically includes the following steps:

[0047] Step 521: Obtain all feature blocks a and matching blocks b in the matching group, calculate the first matching element, the second matching element, and the third matching element for each feature block a and b, and convert the first matching element, the second matching element, and the third matching element into the first matching probability, the second matching probability, and the third matching probability;

[0048] Step 522: constructing a probability model of the feature blocks and the matching blocks in the matching group based on the first matching probability, the second matching probability, and the third matching probability;

[0049] 523: Decompose the probability model into tensors to reduce the model dimension.

[0050] In the technical solution provided in this application, by performing tensor decomposition on the probability model, the model dimension can be effectively reduced, the density of the extracted information can be increased, and the difficulty of fusing high-altitude image data and drone image data can be reduced.

[0051] Furthermore, step 522 includes the following steps:

[0052] Step 5221: Set initial parameters;

[0053] 、B= 、 、 、 ;

[0054] Among them, A represents the first extreme point set of the matching block, B represents the third extreme point set of the matching block, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the first extreme point set, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the second extreme point set; 、 、 Respectively and The first matching probability, the second matching probability and the third matching probability;

[0055] Step 5222: Build a probability model :

[0056] .

[0057] Step 523 includes the following steps:

[0058] Step 5231: Probabilistic Model Perform slice decomposition:

[0059] , slice 1;

[0060] , slice 2;

[0061] , slice 3;

[0062] Step 5232: Based on the slicing results, the probability model Decompose into a low-rank representation:

[0063] ;

[0064] Among them, R represents the decomposition rank, represents the feature block factor vector, represents the matching block factor vector, Matching type factor vector, r represents the sum index, which is used to traverse all decomposition ranks;

[0065] Step 5233: Probability model Reconstruct the probability model after tensor decomposition :

[0066] .

[0067] In the technical solution provided in this application, through the rank index r, tensor decomposition compresses the high-dimensional matching relationship into a low-dimensional latent space, while retaining the key matching patterns, realizing efficient and interpretable matching modeling, thereby being able to explore the implicit connections between the data and provide greater value for subsequent data applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 Flowchart of the information extraction method for multi-source remote sensing data. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific implementation methods. The same figure marks in the accompanying drawings represent the same components. It should be noted that the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the described embodiments of this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0070] Compared to the embodiments shown in the drawings, feasible embodiments within the scope of protection of the present application may have fewer components, other components not shown in the drawings, different components, differently arranged components, or differently connected components, etc. In addition, two or more components in the drawings may be implemented in a single component, or a single component shown in the drawings may be implemented as multiple separate components.

[0071] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning understood by persons of ordinary skill in the field to which this application belongs. The words "first", "second" and similar terms used in the specification and claims of this application do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "a" or "an" do not necessarily indicate a quantitative limitation. "Up", "down" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0072] Example 1:

[0073] refer to Figure 1 The first embodiment of the present application discloses a method for extracting information from multi-source remote sensing data, comprising:

[0074] Step 1: Acquire high-altitude image data of the target area and extract a number of matching blocks from the high-altitude image data; the matching blocks at least include extreme value points exceeding a preset threshold number.

[0075] The method of obtaining the matching block includes the following steps:

[0076] Step 11: Select the mountain boundaries, river boundaries, depression boundaries and road boundaries from the aerial image data.

[0077] Load the digital elevation model (DEM) data of the target area and use the terrain analysis tool group to extract the mountain boundaries, river boundaries, depression boundaries, and road boundaries.

[0078] Step 12: Extract several extreme points from the river boundary, depression boundary and road boundary to generate the first extreme point set.

[0079] The extreme point extraction algorithm provided in Example 2 is used to extract extreme points from mountain boundaries, river boundaries, depression boundaries, and road boundaries. The number of extreme points is as large as possible to increase the accuracy of the description of the feature area. In this way, a first extreme point set can be obtained, which describes the feature information of the mountain boundaries, river boundaries, depression boundaries, and road boundaries.

[0080] Step 13: Extract extreme points from the remaining areas of the high-altitude image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a third extreme point set.

[0081] Similarly, extreme points also need to be extracted in other areas that are not mountain boundaries, river boundaries, depression boundaries, and road boundaries. These extreme points are not involved in matching first, but are gathered together as the third extreme point set.

[0082] Step 14: Establish a correspondence between each first extreme point set and each extreme point in the third extreme point set.

[0083] Establish the correspondence between each extreme point in the first extreme point set and the third extreme point set. The correspondence here is mainly a positional relationship. That is, after describing the positions of all extreme points in the first extreme point set on a blank map, the positions of the remaining extreme points can be restored on the blank map according to the correspondence.

[0084] Step 15: Based on the first extreme point set, extract several matching blocks from the elevation image that meet the number of extreme points. A matching block is not a single pixel, but a specific area containing many pixels and extreme points. Matching matching blocks with larger areas can effectively measure the size difference between satellite imagery and drone imagery.

[0085] Step 2: Obtain drone image data of the target area and extract several feature blocks from the drone image data. The feature blocks at least include extreme points exceeding a preset threshold number;

[0086] The method of obtaining the feature block includes the following steps:

[0087] Step 21: Select the mountain boundaries, river boundaries, depression boundaries, and road boundaries from the drone image data;

[0088] Step 22: extracting a number of extreme points from the river boundary, the depression boundary, and the road boundary to generate a second extreme point set;

[0089] Step 23: Extract extreme points from the remaining areas of the UAV image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a fourth extreme point set;

[0090] Step 24: extracting a number of feature blocks that meet the number conditions of extreme points from the elevation image based on the second extreme point set.

[0091] The method of extracting the feature blocks in step 2 is the same as the method of extracting the matching blocks in step 1, except that the objects are different. Step 2 deals with drone image data, which is the impact data of aerial photography by humans and machines at high altitudes. Relatively speaking, drone image data is clearer and can represent more image information, but the flight altitude of drones is unstable and stable elevation information cannot be obtained. Generally, the image data collected by drones is used as material information and rendered onto high-altitude image data. The specific rendering method will not be described in detail in this application. This application is mainly used to determine the correspondence between drone image data and high-altitude image data, that is, it solves the problem of how to accurately render drone image data onto high-altitude image data.

[0092] The above describes how to obtain matching blocks and feature blocks. However, the following requirements must be met when obtaining matching blocks and feature blocks:

[0093] (1): The real-world physical size of the matching block is consistent with the real-world physical size of the feature block; the size of the matching block is set according to the matching accuracy.

[0094] (2) The number of extreme points in the feature block and the matching block meets the following conditions:

[0095] ;

[0096] x, y represent the center coordinates of the feature block or matching block, i and y i represents the coordinates of the i-th extreme point, K represents the kernel function, n represents the total number of extreme points, and h represents the bandwidth parameter.

[0097] For condition 1, it is only necessary to set it reasonably according to the measurement accuracy of the high-altitude image data and the drone image data. For example, in this application, the size of the matching block and the feature block in the real physical world is set to 10 meters. 10 meters. In other modes, you can set it according to your needs.

[0098] For condition 2, it is necessary to increase the collection performance of extreme points. This application mainly collects extreme points at mountain boundaries, river boundaries, depression boundaries and road boundaries. The terrain changes at these locations are obvious and are more likely to form extreme points.

[0099] Step 3: Obtain the measured coordinates of each matching block in the high-altitude image data to obtain the first coordinate information, and obtain the measured coordinates of each feature block in the drone image data to obtain the second coordinate information.

[0100] The first coordinate information and the second coordinate information are actually the coordinates of the matching block and the feature block in the physical world. In this solution, the coordinate information is understood as the longitude and latitude coordinates of the center points of the matching block and the feature block.

[0101] Step 4: Perform fuzzy matching on the matching blocks and the feature blocks according to the first coordinate information and the second coordinate information, and merge the matching blocks and the feature blocks whose coordinate positions are not more than a preset threshold into one matching group.

[0102] Within a rough range, matching blocks and feature blocks correspond to each other, but this correspondence can be subject to comparison errors due to measurement offsets. Thus, within a matching group, matching blocks and feature blocks can correspond to each other, but the correspondence can be offset. To address this, this solution divides the target area into several matching groups. Each matching group matches its corresponding feature blocks with matching blocks. This reduces the overall number of matching blocks and feature blocks that need to be considered during matching, while ensuring matching accuracy.

[0103] Step 4 includes the following steps:

[0104] Step 41: Obtain mountain boundaries, river boundaries, depression boundaries, and road boundaries from the high-altitude image data, and use the mountain boundaries, river boundaries, depression boundaries, and road boundaries as matching areas;

[0105] Step 42: Divide the matching area into a plurality of matching grids based on a preset grid, and set the size of the matching grid according to a preset threshold;

[0106] Step 43: Merge the matching blocks and feature blocks in the dimensionality reduction matching grid into one matching group.

[0107] The matching group actually matches the feature block and the matching block in a larger range. For example, the size of the feature block and the matching block in the real world is 10 meters. 10 meters, each grid size is 100 meters If the distance is 100 meters, there are 100 matching groups and 100 feature groups that need to be matched within a matching group.

[0108] Step 5: For each matching group, a probability model for the mutual registration of each matching block and feature block is constructed, and the probability model is subjected to tensor decomposition to obtain the one-to-one registration information of the matching block and feature block.

[0109] There are many ways to align the feature blocks and matching blocks in a matching group, each of which is theoretically possible. To this end, this application provides a probabilistic model for matching the feature blocks and matching blocks. After obtaining the probabilistic model, it can achieve higher accuracy when subsequently fusing high-altitude image data with drone image data.

[0110] Step 5 includes the following steps:

[0111] Step 51: The similarity between the feature block and the matching block is used as the first matching element, the similarity between the texture features of the feature block and the matching block is used as the second matching element, and the ratio of the distance between the feature block and the reference object to the distance between the matching block and the reference object is used as the third matching element; the reference object is a specific physical marker pre-set in the target area.

[0112] The first matching element is the similarity between the feature block and the matching block. This involves converting all pixels within the feature block and the matching block into vector representations and then calculating the similarity between the two vectors. This similarity measures the global similarity between the two, but lacks detailed features.

[0113] The second matching element is mainly the similarity of the texture features of the feature block and the matching block, that is, the texture information is extracted from the feature block, and then the texture information is extracted from the matching block, and the similarity is obtained by comparing the two texture information;

[0114] The texture information is extracted as follows:

[0115] Texture information is R;

[0116] .

[0117] Among them, i and j represent the horizontal and vertical coordinates of the pixel points of the feature block or matching block respectively, and N represents the maximum grayscale value. Is in the feature block or matching block frequency.

[0118] The third matching element is the ratio of the distance between the feature block and the reference object to the distance between the matching block and the reference object. The distance here is a vector distance, that is, a length with a direction. The reference object is a preset landmark, such as a set ruler.

[0119] The third matching element can measure the corresponding positions of the feature block and the matching block in the real world. When the third matching element is close to 1, the probability of the two matching is higher.

[0120] Step 52: Establish a probability model based on the first matching element, the second matching element, and the third matching element, and perform tensor decomposition on the probability model to reduce the model dimension.

[0121] Furthermore, step 52 specifically includes the following steps:

[0122] Step 521: Obtain all feature blocks a and matching blocks b in the matching group, calculate the first matching element, second matching element and third matching element for feature blocks a and feature blocks b one by one, and convert the first matching element, second matching element and third matching element into the first matching probability, second matching probability and third matching probability.

[0123] When the first matching element and the second matching element are converted into the first matching probability and the second matching probability, they only need to be multiplied by the preset weight parameters respectively.

[0124] The closer the third matching element is to 1, the higher the probability that the matching block and the feature block match each other. Therefore, when converting the third matching element into the third matching probability, the following formula needs to be used:

[0125] Step 522: Construct a probability model of the feature blocks and the matching blocks in the matching group based on the first matching probability, the second matching probability, and the third matching probability.

[0126] Furthermore, step 522 includes the following steps:

[0127] Step 531: Obtain all feature blocks a and matching blocks b in the matching group, calculate the first matching element, the second matching element, and the third matching element for each feature block a and b, and convert the first matching element, the second matching element, and the third matching element into the first matching probability, the second matching probability, and the third matching probability;

[0128] Step 532: Construct a probability model of the feature blocks and the matching blocks in the matching group based on the first matching probability, the second matching probability, and the third matching probability.

[0129] Specifically, step 532 includes the following steps:

[0130] Step 5321: Set initial parameters;

[0131] 、B= 、 、 、 ;

[0132] Among them, A represents the first extreme point set of the matching block, B represents the third extreme point set of the matching block, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the first extreme point set, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the second extreme point set; 、 、 Respectively and The first matching probability, the second matching probability and the third matching probability;

[0133] Step 5322: Build a probability model :

[0134] ;

[0135] After building a probabilistic model, the data dimension of the probabilistic model is too high, making it very difficult to perform feature fusion later. Dimensionality reduction of the probabilistic model is required.

[0136] 533: Perform tensor decomposition on the probability model to reduce the model dimension.

[0137] Step 533 includes the following steps:

[0138] Step 531: Probabilistic Model Perform slice decomposition:

[0139] , slice 1;

[0140] , slice 2;

[0141] , slice 3;

[0142] Step 532: Based on the slicing results, the probability model Decompose into a low-rank representation:

[0143] ;

[0144] Among them, R represents the decomposition rank, represents the feature block factor vector, represents the matching block factor vector, Matching type factor vector, r represents the sum index, which is used to traverse all decomposition ranks.

[0145] Specifically, Used to describe the potential representation of feature blocks, Used to describe the potential representation of matching blocks, Used to describe the match type weight.

[0146] Step 533: Step 5233: Probability model Reconstruct the probability model after tensor decomposition :

[0147] .

[0148] The technical solution provided by this application reconstructs the joint probability, which can more intuitively illustrate the correspondence between high-altitude image data and drone image data, providing a basis for subsequent 3D model rendering.

[0149] In the technical solution provided by this application, the matching relationship between the first extreme point set and the second extreme point set is a probability model after tensor decomposition. It is indicated that after the correspondence between the first extreme value point set and the second extreme value point set is determined, the correspondence between the third extreme value point set and the fourth extreme value point set is also determined.

[0150] In this way, this application uses a probabilistic model to extract the correlation characteristics between high-altitude image data and drone image data, providing a basis for subsequent drone image data rendering of high-altitude image data.

[0151] Example 2: When performing image registration, this application needs to extract extreme points, and needs to extract as many extreme points as possible. For this purpose, this application provides the following technical solutions:

[0152] Calculate the gradient magnitude and gradient direction of each pixel in the image to be extracted, where the gradient magnitude includes the gradient value at 0°, the gradient value at 45°, the gradient value at 90°, and the gradient value at 135°.

[0153] ; ;

[0154] ; .

[0155] Correspondingly, the calculation formulas for the gradient amplitude and gradient direction of the pixel point are:

[0156] ;

[0157] ;

[0158] Among them, U(x, y) is the gradient amplitude at the pixel point (x, y), α(x, y) is the gradient direction at the pixel point (x, y), is the gradient value of 0°, is a gradient value of 45°, is a gradient value of 90°, The gradient value is 135°;

[0159] All pixels of the image to be extracted are traversed one by one. If the gradient amplitude of a pixel is a local maximum in the gradient direction, the pixel is regarded as an extreme point.

[0160] In this application, when extracting extreme points, compared with the existing extraction algorithm, four directions are selected for extraction, so that a larger number of extreme points can be extracted, thereby increasing the accuracy of the registration of high-altitude image data and drone image data.

[0161] The above are merely preferred embodiments of the present application and are not intended to limit the present application. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A method for extracting information from multi-source remote sensing data, characterized in that: include: Step 1: Obtain high-altitude image data of the target area and extract several matching blocks from the high-altitude image data; The matching block at least includes a number of extreme value points exceeding a preset threshold; Step 2: Obtain UAV image data of the target area and extract several feature blocks from the UAV image data; The feature block at least includes a number of extreme value points exceeding a preset threshold; Step 3: Obtain the measured coordinates of each matching block in the high-altitude image data to obtain the first coordinate information, and obtain the measured coordinates of each feature block in the drone image data to obtain the second coordinate information; Step 4: Perform fuzzy matching on the matching blocks and the feature blocks according to the first coordinate information and the second coordinate information, and merge the matching blocks and the feature blocks whose coordinate positions are not more than a preset threshold into one matching group; Step 5: For each matching group, a probability model for the mutual registration of each matching block and feature block is constructed, and the probability model is subjected to tensor decomposition to obtain the one-to-one registration information of the matching block and feature block.

2. The method for extracting information from multi-source remote sensing data according to claim 1, wherein: The real-world physical size of the matching block is consistent with the real-world physical size of the feature block; the size of the matching block is set according to the matching accuracy.

3. The method for extracting information from multi-source remote sensing data according to claim 2, wherein: The number of extreme points in the feature block and matching block meets the following conditions: ； x, y represent the center coordinates of the feature block or matching block, i and y i represents the coordinates of the i-th extreme point, K represents the kernel function, n represents the total number of extreme points, and h represents the bandwidth parameter.

4. The method for extracting information from multi-source remote sensing data according to claim 1, wherein: The first coordinate information is the satellite positioning information obtained by the satellite, and the second coordinate information is the radar positioning information obtained by the UAV; the sizes of the feature blocks and the matching blocks are determined based on the minimum measurement accuracy of the satellite positioning information.

5. The method for extracting information from multi-source remote sensing data according to claim 1, wherein: The method of obtaining the matching block includes the following steps: Step 11: Select the mountain boundaries, river boundaries, depression boundaries, and road boundaries from the aerial image data; Step 12: Extract several extreme points from the river boundary, depression boundary, and road boundary to generate a first extreme point set; Step 13: Extract extreme points from the remaining areas of the high-altitude image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a third extreme point set; Step 14: Establishing a correspondence between each first extreme point set and each extreme point in the third extreme point set; Step 15: Extract a number of matching blocks that meet the number conditions of extreme points from the elevation image based on the first extreme point set.

6. The method for extracting information from multi-source remote sensing data according to claim 5, characterized in that: The method of obtaining the feature block includes the following steps: Step 21: Select the mountain boundaries, river boundaries, depression boundaries, and road boundaries from the drone image data; Step 22: extracting a number of extreme points from the river boundary, the depression boundary, and the road boundary to generate a second extreme point set; Step 23: Extract extreme points from the remaining areas of the UAV image data excluding mountain boundaries, river boundaries, depression boundaries, and road boundaries to generate a fourth extreme point set; Step 24: extracting a number of feature blocks that meet the number conditions of extreme points from the elevation image based on the second extreme point set.

7. The method for extracting information from multi-source remote sensing data according to claim 5, wherein: Step 4 includes the following steps: Step 41: Obtain mountain boundaries, river boundaries, depression boundaries, and road boundaries from the high-altitude image data, and use the mountain boundaries, river boundaries, depression boundaries, and road boundaries as matching areas; Step 42: Divide the matching area into a plurality of matching grids based on a preset grid, and set the size of the matching grid according to a preset threshold; Step 43: Merge the matching blocks and feature blocks in the dimensionality reduction matching grid into one matching group.

8. The method for extracting information from multi-source remote sensing data according to claim 1, wherein: Step 5 includes the following steps: Step 51: The similarity between the feature block and the matching block is used as the first matching element, the similarity between the texture features of the feature block and the matching block is used as the second matching element, and the ratio of the distance between the feature block and the reference object to the distance between the matching block and the reference object is used as the third matching element; the reference object is a specific physical marker pre-set on the target area; Step 52: Establish a probability model based on the first matching element, the second matching element, and the third matching element; The probability model is decomposed into tensors to reduce the model dimension.

9. The method for extracting information from multi-source remote sensing data according to claim 8, wherein: Step 52 specifically includes the following steps: Step 521: Obtain all feature blocks A and matching blocks B in the matching group, calculate the first matching element, the second matching element, and the third matching element for each feature block A and feature block B, and convert the first matching element, the second matching element, and the third matching element into the first matching probability, the second matching probability, and the third matching probability; Step 522: constructing a probability model of the feature blocks and the matching blocks in the matching group based on the first matching probability, the second matching probability, and the third matching probability; 533: Perform tensor decomposition on the probability model to reduce the model dimension.

10. The method for extracting information from multi-source remote sensing data according to claim 9, wherein: Step 522 includes the following steps: Step 5221: Set initial parameters; Step 5221: Set initial parameters; 、B= 、 、 、 ； Among them, A represents the first extreme point set of the matching block, B represents the third extreme point set of the matching block, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the first extreme point set, Respectively represent the first extreme point, the second extreme point, and the Mth extreme point in the second extreme point set; 、 、 Respectively and The first matching probability, the second matching probability and the third matching probability; Step 5222: Build a probability model : Step 523 includes the following steps: Step 5231: Probabilistic Model Perform slice decomposition: , slice 1; , slice 2; , slice 3; Step 5232: Based on the slicing results, the probability model Decompose into a low-rank representation: ； Among them, R represents the decomposition rank, represents the feature block factor vector, represents the matching block factor vector, Matching type factor vector, r represents the sum index, which is used to traverse all decomposition ranks; Step 5233: Probability model Reconstruct the probability model after tensor decomposition : 。

Images