Complex Terrain Analysis Method Based on the Fusion of Airborne LiDAR and Spaceborne Multispectral Images

Abstract

The present invention discloses a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images, comprising: acquiring multispectral images and point cloud data; generating a depth map based on the point cloud data; performing rough registration on the depth map and the multispectral images to obtain rough registration point pairs; establishing a two-dimensional to three-dimensional conversion model by using the rough registration point pairs and the elevation values of the point cloud data; determining the fine registration coefficient through a search algorithm with the maximum mutual information value between the point cloud depth map and the multispectral images as the measure; converting the multispectral images according to the fine registration coefficient and the conversion model to obtain converted images; performing dynamic feature fusion on the converted images and the point cloud data to generate fusion data; generating a contour topographic map by using elevation information and displaying it in the multispectral images; and analyzing the complex scene in cooperation with the contour topographic map and the fusion data. The present invention reduces the time cost, improves the registration and fusion efficiency of heterogeneous data, and provides support for the analysis of complex scene conditions.

Description

Technical Field

[0001] The present invention belongs to the technical field of terrain analysis, and particularly relates to a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images. Background Art

[0002] LiDAR (Light Detection And Ranging) is an indispensable "scout" for future information-based system operations. With advantages such as high resolution, good concealment, strong anti-active interference ability, good low-altitude detection performance, small size, and light weight, it has been widely used in the military field. However, with the evolution of battlefield informatization and the complex and changeable nature of actual combat, purposeful and targeted camouflage forms are more diverse. The construction of three-dimensional terrain in a complex environment with single data gradually cannot meet the actual needs, and dynamically obtaining real-scene information and dynamically presenting environmental information have also become the key to grasping the overall battle situation. Adding dynamic real-scene and thermal infrared information to the construction of three-dimensional terrain in a complex environment will make the information in the complex environment more abundant, thus being more conducive to the actual work arrangement and overall planning. Therefore, researching complex terrain analysis based on the registration and fusion of airborne dynamic LiDAR data and spaceborne static multispectral images has great practical application value.

[0003] Currently, the registration techniques of LiDAR and multispectral images mainly include registration methods based on mutual information, feature-based registration methods, intensity-based registration methods, etc. Specifically, the registration method based on mutual information mainly registers for the gray information, gradient information, etc. of the images. The feature-based registration method mainly realizes registration by extracting special points, lines, and planes in the images. The intensity-based registration method mainly maps LiDAR point clouds into two-dimensional images using elevation values or intensity values, and then uses relatively mature two-dimensional image registration methods to register the point cloud data and optical images. To a certain extent, the above methods all have good performances when matching point cloud data and multispectral images. By selecting a registration method with better performance and fusing specific regions in LiDAR and multispectral images according to requirements, the actual situation in a complex environment will be more prominent and more conducive to grasping the overall situation.

[0004] Multi-source data fusion has always been a research hotspot in related fields. With the development of LiDAR technology, the fusion of point cloud data and optical remote sensing data has attracted extensive attention. Existing technologies have applied deep learning to data fusion and achieved certain research results. Among them, feature-level fusion is one of the most widely used methods at present. This method extracts the spectral features and spatial features of optical remote sensing images and the elevation features of LiDAR point cloud data and performs fusion. Another commonly used fusion strategy is decision-level fusion, which is widely used in multi-source data image classification. After classifying according to the different features of different data, it performs fusion through a decision-making mechanism or a weight mechanism to obtain a better classification result.

[0005] The above methods do not achieve scene dynamic registration and fusion, do not perform heterologous image matching in selective specific regions, have high algorithm complexity and large workload, and have limitations in information richness, unable to meet actual needs. Summary of the Invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images. The technical problems to be solved by the present invention are realized through the following technical solutions:

[0007] The present invention provides a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images, including:

[0008] Obtain multispectral images using spaceborne sensors and obtain point cloud data using airborne lidar, and the area covered by the point cloud data is included in the area covered by the multispectral images;

[0009] Generate a point cloud depth map based on the point cloud data;

[0010] Coarsely register the point cloud depth map and the multispectral image to obtain coarsely registered point pairs;

[0011] Use the coarsely registered point pairs and the elevation values of the point cloud data to establish a two-dimensional to three-dimensional conversion model;

[0012] Taking the maximum mutual information value between the point cloud depth map and the multispectral image as a measure, determine the fine registration coefficient through a search algorithm;

[0013] According to the fine registration coefficient and the conversion model, convert the multispectral image to obtain a converted image;

[0014] Perform dynamic feature fusion on the converted image and the point cloud data to generate fusion data;

[0015] Generate a contour topographic map and display it in the multispectral image for collaborative analysis with the fusion data.

[0016] In an embodiment of the present invention, before the step of generating a point cloud depth map based on point cloud data, it further includes:

[0017] Preprocess the multispectral image and the point cloud data respectively.

[0018] In an embodiment of the present invention, the step of preprocessing the multispectral image and the point cloud data respectively includes:

[0019] Perform radiometric calibration, atmospheric correction, and geometric correction on the multispectral image to obtain a preprocessed multispectral image;

[0020] Use a filtering algorithm to filter and denoise the point cloud, and downsample the filtered point cloud data to obtain preprocessed point cloud data; the filtering algorithm is at least one of median filtering, band-pass filtering, bilateral filtering, statistical filtering, radius filtering, conditional filtering, and Gaussian filtering.

[0021] In an embodiment of the present invention, the step of determining the registration coefficient through a search algorithm with the maximum mutual information between the point cloud depth map and the multispectral image as a measure includes:

[0022] Calculate the mutual information between the point cloud depth map and the multispectral image;

[0023] After obtaining the rough registration coefficient during the rough registration of the point cloud depth map and the multispectral image, substitute the rough registration coefficient as the initial registration coefficient into the transformation model, and search within a preset range using the Powell algorithm;

[0024] Transform the multispectral image using the registration coefficient searched in each round, and calculate the mutual information between the point cloud depth map and the transformed multispectral image;

[0025] When the mutual information reaches the maximum, determine the corresponding registration coefficient as the fine registration coefficient.

[0026] In an embodiment of the present invention, the step of calculating the mutual information between the point cloud depth map and the transformed multispectral image includes:

[0027] Calculate the edge information entropy H(M) of the transformed multispectral image and the edge information entropy H(D) of the point cloud depth map respectively;

[0028] Calculate the conditional information entropy H(DM) based on the edge information entropy H(M) of the transformed multi-spectral image and the edge information entropy H(D) of the point cloud depth map, and calculate H(M,D) according to the conditional information entropy H(DM);

[0029] Calculate the mutual information I(D;M) between the transformed multi-spectral image and the point cloud depth map according to the edge information entropy H(D) of the point cloud depth map, the edge information entropy H(M) of the transformed multi-spectral image, and H(M,D):

[0030] I(D;M)=H(M)+H(D)-H(M,D).

[0031] In an embodiment of the present invention, calculate the edge information entropy H(M) of the transformed multi-spectral image and the edge information entropy H(D) of the point cloud depth map respectively according to the following formulas:

[0032]

[0033]

[0034] Wherein, P(·) represents the probability of a discrete random variable, M represents the transformed multi-spectral image, and D represents the point cloud depth map.

[0035] In an embodiment of the present invention, the steps of dynamically fusing the transformed image and the point cloud data to generate fused data include:

[0036] Assign the elevation information in the point cloud data to the corresponding multi-spectral image to generate fused data.

[0037] In an embodiment of the present invention, the steps of dynamically fusing the transformed image and the point cloud data to generate fused data include:

[0038] Fuse the point cloud data with the multi-spectral image of the corresponding area to form a fused image presenting the point cloud data on the multi-spectral image.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] The present invention provides a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images. By leveraging the advantages of heterogeneous data, registration, and fusion technologies, dynamic registration and fusion of partial regions between airborne LiDAR dynamically acquired data scenarios and large-scale multispectral (including infrared data) images are achieved. Thus, dynamic point cloud data can be acquired, transmitted, registered, and fused in real time, ultimately enabling the dynamic construction and analysis of a three-dimensional real-scene topographic map based on multispectral data. This method reduces the time cost, improves the registration and fusion efficiency of heterogeneous data, and provides reliable support for the analysis of complex scenarios.

[0041] The following will further elaborate on the present invention in conjunction with the accompanying drawings and embodiments. Brief Description of the Drawings

[0042] Figure 1 is a flowchart of a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images provided by an embodiment of the present invention;

[0043] Figure 2 is a flowchart of a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images provided by an embodiment of the present invention;

[0044] Figure 3 is a matching schematic diagram of point cloud data and multispectral images provided by an embodiment of the present invention;

[0045] Figure 4 is a fusion schematic diagram of point cloud data and multispectral images provided by an embodiment of the present invention. Detailed Embodiments

[0046] The following further describes the present invention in detail in conjunction with specific embodiments, but the implementation manners of the present invention are not limited thereto.

[0047] Figure 1 is a flowchart of a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images provided by an embodiment of the present invention, Figure 2 is a flowchart of a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images provided by an embodiment of the present invention. As Figure 1-2 shown, the present invention provides a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images, including:

[0048] S1. Obtain multispectral images using a spaceborne sensor and obtain point cloud data using an airborne lidar, and the area covered by the point cloud data is included in the area covered by the multispectral images;

[0049] S2. Generate a point cloud depth map based on the point cloud data;

[0050] S3. Coarsely register the point cloud depth map and the multi-spectral image to obtain coarsely registered point pairs;

[0051] S4. Use the coarsely registered point pairs and the elevation values of the point cloud data to establish a two-dimensional to three-dimensional conversion model;

[0052] S5. Taking the maximum value of the mutual information between the point cloud depth map and the multi-spectral image as a measure, determine the fine registration coefficient through a search algorithm;

[0053] S6. According to the fine registration coefficient and the conversion model, convert the multi-spectral image to obtain a converted image;

[0054] S7. Dynamically fuse the converted image and the point cloud data to generate fused data;

[0055] S8. Generate a contour topographic map and display it in the multi-spectral image for collaborative analysis with the fused data. In this embodiment, first, a multi-spectral image of a large scene is obtained using a spaceborne sensor, and point cloud data of multiple regions is obtained using an airborne lidar. On the one hand, during the acquisition of the multi-spectral image, solar radiation will enter the object surface through the atmosphere in a certain way and then be reflected back to the spaceborne sensor. Therefore, affected by factors such as atmospheric aerosols, terrain, and adjacent ground objects, the multi-spectral image contains comprehensive information from multiple aspects such as the object surface, the atmosphere, and the sun. On the other hand, due to factors such as equipment, operator experience, and environmental factors, the acquired point cloud data often contains a large number of isolated points, discrete points, and some voids. It can be seen that after obtaining the multi-spectral image and the point cloud data, preprocessing of the two is still required.

[0056] Furthermore, the generation of the point cloud depth map is an important process for the registration process of the point cloud data and the multi-spectral image. In step S2, it can be generated through a third-party software such as Cloud Compare, or the commonly used model for generating the point cloud depth map, the "pinhole" simulation imaging model, can also be used. When using the "pinhole" simulation for imaging, it is assumed that the multi-spectral image and the point cloud data have been accurately registered, and the point cloud data can completely construct a DSM (Digital Surface Model, digital surface model). The DSM refers to a ground elevation model that includes the heights of surface buildings, bridges, and trees. By relying on the collinearity equation and the image orientation elements, and using the elevation (depth distance) as the basis for the pixel gray value, a simulated gray image consistent with the multi-spectral image, that is, the point cloud depth map, can be generated. Affected by various factors, there are visual differences between this simulated gray image and the multi-spectral image, but they are relatively similar in terms of the geometric shape and position of the ground objects and the image resolution.

[0057] In steps S3 - S5, the SIFT algorithm is used to perform rough registration on the point cloud depth map and the multi - spectral image, obtaining rough registration point pairs. Then, using the rough registration point pairs and the elevation information (i.e., the z - coordinate) of the point cloud data, a two - dimensional to three - dimensional transformation model is established to transform the multi - spectral image. Subsequently, the Powell algorithm is used to search for the fine registration coefficients of the transformation model.

[0058] Finally, the multi - spectral image is transformed using the fine registration coefficients to obtain a transformed image. Further, the point cloud data and the multi - spectral image are subjected to dynamic feature fusion. After generating a contour topographic map using software such as ArcGIS, Graphic Earth, and LiDAR360, it is displayed in the pre - processed multi - spectral image and analyzed in coordination with the fusion data for relevant personnel to conduct coordinated analysis of complex environments such as battlefields.

[0059] It should be understood that in this embodiment, by using the dynamic LiDAR point cloud data collected by the unmanned aerial vehicle in combination with static multi - spectral data and digital elevation data, terrain factors of surface geometric features and water system features can be selected as needed. For example, terrain factors of surface geometric features: average slope, average slope change, roughness index, terrain elevation ratio, terrain, elevation deviation; terrain factors of water system features: water network density, rugosity, river network bifurcation degree, structure, average valley depth. Optionally, a classification tree is established using dynamic hierarchical classification, and the topographic characteristics of the relevant area can be quickly obtained through multi - level classification technology, such as the degree of terrain undulation in the target area, the density of the river network in the area and its relationship with the road distribution, the distribution characteristics of surrounding residents, and the attributes and distribution of soil and vegetation. Specific area, length and other attribute information of relevant categories can be obtained through quantitative calculation.

[0060] Optionally, before the step of generating a point cloud depth map based on the point cloud data, it further includes:

[0061] Pre - processing the multi - spectral image and the point cloud data respectively.

[0062] Exemplarily, the steps of pre - processing the multi - spectral image and the point cloud data respectively include:

[0063] Performing radiometric calibration, atmospheric correction, and geometric correction on the multi - spectral image to obtain the pre - processed multi - spectral image;

[0064] Using a filtering algorithm to perform filtering and denoising on the point cloud, and down - sampling the filtered point cloud data to obtain the pre - processed point cloud data; the filtering algorithm is at least one of median filtering, band - pass filtering, bilateral filtering, statistical filtering, radius filtering, conditional filtering, and Gaussian filtering.

[0065] Specifically, the preprocessing process of the multi-spectral image includes radiometric calibration, atmospheric correction, geometric correction, etc. Exemplarily, radiometric calibration is a process of converting the digital quantization value of the multi-spectral image into physical quantities such as radiance value, reflectance, or surface temperature; atmospheric correction can separate the reflection information of a certain object surface from the information of the atmosphere and the sun to understand the spectral properties of the object surface; for the image distortion caused by the performance differences of spaceborne sensors, the attitude changes of the carrier vehicle, and the target object, etc., it can be eliminated or improved by using geometric correction.

[0066] For the point cloud data, in this embodiment, filtering and denoising processing and downsampling are performed on it, such as median filtering algorithm, band-pass filtering algorithm, bilateral filtering algorithm, statistical filtering algorithm, radius filtering algorithm, conditional filtering algorithm, Gaussian filtering algorithm, etc. Further, considering that the acquired heterogeneous data is imaged by different sensors and has different resolutions, in order to ensure better registration of the point cloud data and the multi-spectral, in this embodiment, downsampling is performed on the filtered point cloud data.

[0067] Figure 3 It is a matching schematic diagram of the point cloud data and the multi-spectral image provided by the embodiment of the present invention. It should be noted that in this embodiment, downsampling is performed on the point cloud data instead of upsampling the multi-spectral image. This is because when acquiring multiple point cloud data with multiple resolutions, it is only necessary to downsample the corresponding point cloud data to the resolution of the multi-spectral image, and as Figure 3 shown, multiple point cloud data can be simultaneously matched to the same multi-spectral image. When downsampling the point cloud data, if n types of resolution point cloud data are acquired and downsampled n times and then registered with the preprocessed multi-spectral image, there will also be 1 registration result. If the multi-spectral image is upsampled, it is necessary to perform upsampling of n types of resolution multi-spectral data to make its resolution the same as the acquired point cloud data, and after registration, there will be n registration results (because there are n types of resolutions), which is not conducive to quickly viewing the three-dimensional terrain real scene maps of different regions, and the overall processing time cost will increase significantly.

[0068] Optionally, in the above step S5, the step of determining the fine registration coefficient by a search algorithm with the maximum mutual information between the point cloud depth map and the multi-spectral image as the measure includes:

[0069] S501. Calculate the mutual information between the point cloud depth map and the multi-spectral image;

[0070] S502. After obtaining the rough registration coefficient during the rough registration of the point cloud depth map and the multi-spectral image, substitute the rough registration coefficient into the transformation model as the initial registration coefficient, and use the Powell algorithm to search within a preset range;

[0071] S503. Transform the multi-spectral image using the registration coefficients found in each round, and calculate the mutual information between the point cloud depth map and the transformed multi-spectral image;

[0072] S504. When the mutual information reaches the maximum, determine the corresponding registration coefficient as the fine registration coefficient.

[0073] Optionally, the step of calculating the mutual information between the point cloud depth map and the transformed multi-spectral image includes:

[0074] Calculate the marginal information entropy H(M) of the transformed multi-spectral image and the marginal information entropy H(D) of the point cloud depth map respectively;

[0075] Calculate the conditional information entropy H(DM) based on the marginal information entropy H(M) of the transformed multi-spectral image and the marginal information entropy H(D) of the point cloud depth map, and calculate H(M,D) according to the conditional information entropy H(DM);

[0076] Calculate the mutual information I(D;M) between the point cloud depth map and the transformed multi-spectral image according to the marginal information entropy H(D) of the point cloud depth map, the marginal information entropy H(M) of the transformed multi-spectral image, and H(M,D):

[0077] I(D;M) = H(M) + H(D) - H(M,D).

[0078] Specifically, calculate the marginal information entropy H(M) of the above-mentioned transformed multi-spectral image and the marginal information entropy H(D) of the point cloud depth map respectively according to the following formulas:

[0079]

[0080]

[0081] In the formula, P(·) represents the probability of a discrete random variable, M represents the transformed multi-spectral image, and D represents the point cloud depth map.

[0082] In the above step S7, the step of performing dynamic feature fusion on the transformed image and the point cloud data to generate fused data includes:

[0083] Assign the elevation information in the point cloud data to the corresponding area of the multi-spectral image after fine registration to generate fused data.

[0084] In other embodiments of the present invention, the step of performing dynamic feature fusion on the transformed image and the point cloud data to generate fused data may further include:

[0085] Fuse the point cloud data with the multi-spectral image of the corresponding area to form a fused image presenting the point cloud data on the multi-spectral image.

[0086] Figure 4 It is a schematic diagram of the fusion of point cloud data and multispectral images provided by the embodiments of the present invention. Specifically, as Figure 4 shown, for the fusion of point cloud data and multispectral images, there are two forms: one is to pay more attention to the elevation information of LiDAR point cloud data, so the elevation information of the corresponding points can be assigned to the corresponding multispectral data; the other is to fuse the LiDAR point cloud data with the multispectral data of the corresponding area to form the LiDAR point cloud data that can be presented on the multispectral data.

[0087] It can be seen from the above embodiments that the beneficial effects of the present invention are as follows:

[0088] The present invention provides a complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images. By utilizing the advantages of heterogeneous data, registration and fusion technologies, dynamic registration and fusion of partial areas of large-scale multispectral (including infrared data) images are realized in the scenario of dynamically collecting data by airborne LiDAR, so as to realize the dynamic acquisition, transmission, registration and fusion of dynamic point cloud data, and finally achieve the dynamic construction and analysis of three-dimensional real-scene topographic maps based on multispectral data. This method reduces the time cost, improves the registration and fusion efficiency of heterogeneous data, and provides reliable support for the analysis of complex scene situations.

[0089] In the description of the present invention, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, "a plurality" means two or more unless otherwise specifically defined.

[0090] The description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification.

[0091] Although the present application has been described in conjunction with various embodiments herein, however, in the process of implementing the claimed present application, those skilled in the art can understand and realize other variations of the disclosed embodiments by viewing the accompanying drawings, the disclosed content, and the appended claims.

[0092] The above content is a further detailed description of the present invention in combination with specific preferred embodiments. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as falling within the protection scope of the present invention.

Claims

1. A complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images, characterized in that, include: Acquire multispectral images using a spaceborne sensor and acquire point cloud data using an airborne laser radar, wherein an area covered by the point cloud data is included in an area covered by the multispectral images; Generate a point cloud depth map based on point cloud data; performing coarse registration on the point cloud depth map and the multispectral image to obtain coarse registration point pairs; Establishing a two-dimensional to three-dimensional conversion model using the coarse registration point pairs and the elevation values of the point cloud data; Determining a precise registration coefficient through a search algorithm using the maximum mutual information between the point cloud depth map and the multispectral image as a measure; Converting the multispectral image according to the precise registration coefficient and the conversion model to obtain a converted image; Performing dynamic feature fusion on the converted image and the point cloud data to generate fused data; generating a contour topographic map and displaying it in the multispectral image, and analyzing it in conjunction with the fused data; The step of determining the registration coefficient by a search algorithm using the maximum mutual information between the point cloud depth map and the multispectral image as a measure includes: Calculating the mutual information between the point cloud depth map and the multispectral image; After obtaining a coarse registration coefficient when coarsely registering the point cloud depth map with the multispectral image, substituting the coarse registration coefficient into the conversion model as an initial registration coefficient, and searching within a preset range using the Powell algorithm; Transforming the multispectral image using the registration coefficients searched in each round, and calculating the mutual information between the point cloud depth map and the transformed multispectral image; When the mutual information reaches a maximum, the corresponding registration coefficient is determined as the fine registration coefficient.

2. The complex terrain analysis method based on the fusion of airborne LiDAR and satellite-borne multispectral images according to claim 1 is characterized in that: Before the step of generating a point cloud depth map based on point cloud data, it also includes: The multispectral image and the point cloud data are preprocessed respectively.

3. The complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images according to claim 2, characterized in that, The steps of preprocessing the multispectral image and the point cloud data respectively include: performing radiometric calibration, atmospheric correction, and geometric correction on the multispectral image to obtain a preprocessed multispectral image; The point cloud is filtered and denoised using a filtering algorithm, and the filtered point cloud data is downsampled to obtain preprocessed point cloud data; the filtering algorithm is at least one of a medium-pass filter, a band-pass filter, a bilateral filter, a statistical filter, a radius filter, a conditional filter, and a Gaussian filter.

4. The complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images according to claim 1, characterized in that, The step of calculating the mutual information between the point cloud depth map and the transformed multispectral image comprises: Calculate the edge information entropy of the transformed multispectral image respectively And the edge information entropy of the point cloud depth map ; Edge information entropy based on the transformed multi-spectral image and the edge information entropy of the point cloud depth map Calculate the conditional information entropy and calculate based on the conditional information entropy ; According to the edge information entropy of the point cloud depth map , the edge information entropy of the transformed multispectral image and , calculate the mutual information between the transformed multispectral image and the point cloud depth map : 。 5. The complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images according to claim 4, characterized in that Calculate the edge information entropy of the transformed multi-spectral image according to the following formulas respectively and the edge information entropy of the point cloud depth map : ； ； Where, represents the probability of a discrete random variable, represents the multispectral image after the transformation, represents the transformed point cloud depth map.

6. The complex terrain analysis method based on the fusion of airborne LiDAR and spaceborne multispectral images according to claim 1, wherein The step of performing dynamic feature fusion of the converted image and the point cloud data to generate fused data includes: The elevation information in the point cloud data is assigned to the corresponding multispectral image to generate fused data.

7. The complex terrain analysis method based on the fusion of airborne LiDAR and satellite-borne multispectral images according to claim 1, characterized in that: The step of performing dynamic feature fusion of the converted image and the point cloud data to generate fused data includes: The point cloud data is fused with the multispectral image of the corresponding area to form a fused image that presents the point cloud data on the multispectral image.

Images