Orthoimage generation method and device based on scene reconstruction

Abstract

The invention discloses an orthoimage generation method and device based on scene reconstruction, and relates to the technical field of image processing. The method comprises the following steps: carrying out coordinate conversion on images, and establishing an adjacency relation between the images; performing feature point matching on the images based on the adjacency relation to obtain a matching pair file, and determining an initial matching pair and poses of other images relative to the initial matching pair in the matching pair file to obtain an object space sparse point cloud; filtering and rasterizing the object space sparse point cloud to obtain a digital surface model; performing point cloud extraction and surface reconstruction on the digital surface model to obtain a surface grid model; performing texture mapping on the surface mesh model to obtain a texture three-dimensional model, and converting the texture three-dimensional model to a target coordinate system to obtain a target three-dimensional model; and performing orthographic projection on the target three-dimensional model to obtain a two-dimensional orthoimage. According to the method, the three-dimensional model can be reconstructed more quickly under the condition that the model precision is ensured.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and device for generating orthophotos based on scene reconstruction. Background Art

[0002] With the continuous development of drone photogrammetry technology, the methods of generating orthophotos are becoming more and more abundant. At present, in the image processing process, using orthophotos to solve problems in urban planning, environmental analysis, land management, cadastral surveying and other fields can be more efficient and convenient.

[0003] At present, orthophotos are generated using photogrammetry methods. Usually, after motion restoration of structure, the original image is corrected and mosaicked using a pixel-by-pixel digital differential correction method, or the texture model is orthographically projected after 3D reconstruction to obtain an orthophoto. These traditional orthophoto generation methods all require dense matching. While dense matching finds points of the same name for all points on the image, the time required for dense matching increases proportionally with the increase in the number of images, which makes the entire orthophoto generation process time-consuming and cumbersome, and may not necessarily generate accurate orthophotos, which cannot meet the business needs of real-world applications. This is especially true for some special scenarios, such as emergency mapping, which require accuracy and real-time business needs. Summary of the invention

[0004] The embodiment of the present application provides a method and device for generating orthophotos based on scene reconstruction, which fully considers the relationship between various links and processing time in the process of orthophoto generation. The distance between shooting points is obtained through coordinate system conversion to establish the adjacency relationship between images, and the sparse point cloud is obtained by matching feature points according to the adjacency relationship. The digital surface model is obtained by rasterizing the sparse point cloud, and then the digital surface model is extracted pixel by pixel to extrude a three-dimensional surface mesh model, avoiding the dense matching process. It is simpler and more efficient than the traditional orthophoto generation process, and solves the problem that the orthophoto generation process in the prior art is cumbersome, time-consuming, and difficult to guarantee accuracy. It can enhance the user experience in emergency scenarios and improve the accuracy and real-time performance of orthophotos.

[0005] In a first aspect, an embodiment of the present application provides a method for generating an orthophoto based on scene reconstruction, comprising: performing coordinate transformation on the image and establishing an adjacency relationship between the images; performing feature point matching on the image based on the adjacency relationship to obtain a matching pair file, and determining the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-side sparse point cloud; filtering and rasterizing the object-side sparse point cloud to obtain a digital surface model; performing point cloud extraction and surface reconstruction on the digital surface model to obtain a surface mesh model; performing texture mapping on the surface mesh model to obtain a textured three-dimensional model, and converting the textured three-dimensional model to a target coordinate system to obtain a target three-dimensional model; and performing orthographic projection on the target three-dimensional model to obtain a two-dimensional orthophoto.

[0006] In a possible implementation, the establishing of the adjacency relationship between images includes: defining a grayscale set based on the information of the image, traversing all the pixels in the grayscale set and determining the adjacency type; wherein the adjacency type includes 4-adjacency, 8-adjacency and m-adjacency; if the pixel points satisfy the adjacency condition, the adjacency relationship between the images is created; wherein the adjacency condition includes that the pixel points are adjacent in position and satisfy a similarity criterion.

[0007] In a possible implementation, before performing feature point matching on the image based on the adjacency relationship to obtain a matching pair file, the method further includes: resampling the image information and then extracting feature points of the resampled image.

[0008] In a possible implementation, the matching pair file is obtained by matching feature points of images based on the adjacency relationship, including: matching feature points of adjacent images according to the adjacency relationship to obtain initial matching point pairs; determining a matching threshold according to the number of the initial matching point pairs; and filtering the initial matching point pairs by the matching threshold to obtain valid matching point pairs to constitute the matching pair file.

[0009] In a possible implementation, the method of determining the initial matching pair and the positions of other images relative to the initial matching pair to obtain the object-side sparse point cloud includes: sequentially traversing the images corresponding to the matching pair file, and determining the best matching point pair therein according to the number of the valid matching point pairs and their intersection angle information; calculating the positions and forward intersections of the other valid point pairs in the matching pair file relative to the best matching point pair; and optimizing the positions and the forward intersections to obtain the object-side sparse point cloud.

[0010] In a possible implementation, the filtering and rasterizing of the object-side sparse point cloud includes: performing statistical filtering on the object-side sparse point cloud; rasterizing the sparse point cloud after the statistical filtering, and taking the maximum value of the sparse point cloud in each grid as the value of the grid.

[0011] In a possible implementation, extracting point clouds and reconstructing the surface of the digital surface model to obtain a surface mesh model includes: extracting point clouds from the digital surface model pixel by pixel; and performing surface Poisson reconstruction based on the extracted point clouds to obtain the surface mesh model.

[0012] In the second aspect, an embodiment of the present application provides a device for generating orthophotos based on scene reconstruction, including: a data processing module, which is used to perform coordinate transformation on images and establish adjacency relationships between images; a motion recovery module, which is used to perform feature point matching on images based on the adjacency relationship to obtain a matching pair file, and determine the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-side sparse point cloud; a rasterization module, which is used to filter and rasterize the object-side sparse point cloud to obtain a digital surface model; a reconstruction module, which is used to perform point cloud extraction and surface reconstruction on the digital surface model to obtain a surface mesh model; a texture mapping module, which is used to perform texture mapping on the surface mesh model to obtain a textured three-dimensional model, and convert the textured three-dimensional model to a target coordinate system to obtain a target three-dimensional model; an orthophoto generation module, which is used to perform positive projection on the target three-dimensional model to obtain a two-dimensional orthophoto.

[0013] In a third aspect, an embodiment of the present application provides a device for a method for generating orthophotos based on scene reconstruction, the device comprising: a processor; a memory for storing processor executable instructions; when the processor executes the executable instructions, it implements the method described in the first aspect or any possible implementation method of the first aspect.

[0014] In a fourth aspect, an embodiment of the present application provides a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium includes a medium for storing a computer program or instructions, which, when executed, enables the method described in the first aspect or any possible implementation method of the first aspect to be implemented.

[0015] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: The embodiment of the present application adopts a method and device for generating orthophotos based on scene reconstruction. On the basis of feature point extraction and matching, the image is subjected to a process of determining the pose and generating a sparse point cloud in the object space, and then rasterizing to generate a digital surface model, extracting the point cloud of the digital surface model until a surface mesh model is generated. Finally, the textured three-dimensional model is orthographically projected to obtain an orthophoto. This effectively solves the problem of a proportional increase in the time consumption in a dense matching process as the number of images increases. Taking into account various influencing factors such as the fact that the production process of traditional orthophotos is too time-consuming and cannot meet the time efficiency requirements of special scenarios, the orthophoto generation speed and plane geometry accuracy are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flowchart of a method for generating an orthophoto based on scene reconstruction provided in an embodiment of the present application; Figure 2 The effect diagram of the sparse point cloud provided in the embodiment of the present application; Figure 3 A rendering of a surface mesh model for 2.5D scene reconstruction provided in an embodiment of the present application; Figure 4 The effect diagram of the texture mapping of the surface mesh model provided in the embodiment of the present application; Figure 5 The rendering of the orthophoto of the scene reconstruction provided in the embodiment of the present application; Figure 6 A schematic diagram of the structure of a device for generating orthophotos based on scene reconstruction provided in an embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] The following describes some of the techniques involved in the embodiments of the present application to facilitate understanding, and they should be considered as merely exemplary. Therefore, it should be appreciated by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of clarity and conciseness, some descriptions of well-known functions and structures are omitted in the following description.

[0020] Figure 1 It is a flowchart of a method for generating an orthophoto based on scene reconstruction provided in an embodiment of the present application, including steps 101 to 106. Figure 1 This is only an execution order shown in the embodiment of the present application, and does not represent the only execution order of the method for generating orthophotos for scene reconstruction. In the case where the final result can be achieved, Figure 1 The steps shown may be performed in parallel or in reverse, as follows.

[0021] Step 101: Convert the image coordinates and establish adjacency between images. In an embodiment of the present application, the ground image is obtained by drone photography, and the exifread library in Python is used to read the EXIF ​​(Exchangeable Image File Format) of the ground image or the initial position attitude angle of the camera. The GPS coordinates are regarded as the origin of the camera in the world coordinate system, and the specific direction of the camera is determined in combination with the rotation matrix, which is converted into world coordinates to facilitate unified processing of image data. Among them, EXIF ​​information includes: width, height, longitude and latitude of the ground image.

[0022] In addition, those skilled in the art can obtain ground images through satellite remote sensing, online data platforms, and scanning technologies. Define a grayscale set based on the information of the ground image, traverse all the pixels in the grayscale set and determine the adjacency type to better represent the positional relationship between surface elements. Among them, the adjacency types include 4-adjacency, 8-adjacency, and m-adjacency; if the adjacency conditions are met between the pixels, an adjacency relationship between the images is created; among them, the adjacency conditions include that the pixels are adjacent in position and must meet the similarity criteria.

[0023] Step 102: Match the feature points of the images based on the adjacency relationship to obtain a matching pair file, and determine the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-space sparse point cloud. Figure 2The effect diagram of sparse point cloud is shown. In order to restore the position and posture of the drone's onboard camera at the moment of photography to perform motion structure recovery, specifically, the ground image is resampled to extract features, and the maximum width or height of the image is set to 2048. Reusing the feature detection method, this application exemplarily uses the SIFTGPU (scale-invariant feature) algorithm to extract feature points from the resampled image. By making full use of the parallel computing power of the GPU, the speed of feature detection and descriptor generation is greatly improved, thereby improving the computational efficiency of the entire feature extraction process. Based on the adjacency relationship, feature points of adjacent images are matched to obtain initial matching point pairs, and the matching threshold is determined according to the number of initial matching point pairs. The initial matching pairs are filtered by the matching threshold to obtain valid matching point pairs to form a matching pair file.

[0024] Specifically, the threshold is set to n according to the actual number of matching points, and the matching pairs with matching points greater than or equal to n are filtered. The images corresponding to the matching pair files are traversed in order, and the best matching point pair is determined according to the number of valid matching point pairs and their intersection angle information. Among them, the best matching point pair needs to perform relative orientation of the image, and the process of determining the relative position and posture between the images can be achieved by Hartley's 8-point method (an algorithm based on linear algebra to solve the basic matrix) or Nister's 5-point method (an algorithm based on matching points to solve the basic matrix or essential matrix).

[0025] In addition, RANSAC (randomly adopted consensus algorithm) is also needed to remove wrong matching points to improve the accuracy of the best matching point pair. This process is used to replace dense matching, speed up the generation of the entire orthophoto while saving computing resources; calculate the pose and front intersection of other valid point pairs in the matching pair file relative to the best matching point pair; optimize the pose and front intersection. Specifically, the image is locally optimized before incremental SFM (structure restored from motion) to minimize the reprojection error and maintain a high local accuracy, that is, the difference between the actual position of the feature point on the image and the position predicted by the model, and fine optimization is performed on the currently processed part to improve the reconstruction accuracy of the local area. Global optimization is performed after incremental SFM to minimize the overall reprojection error of all feature points and improve overall accuracy and stability. This targeted combination of two methods at different stages can fully utilize the advantages of local and global adjustments and complement each other to obtain high-quality object space sparse point clouds.

[0026] Step 103: Filter and rasterize the object-side sparse point cloud to obtain a digital surface model. In the embodiment of the present application, the number of neighboring points and the standard deviation threshold in the object-side sparse point cloud are statistically filtered. Compared with the traditional filtering algorithm, the definition of outliers in statistical filtering is more complicated. The average distance between feature points is first calculated. If it conforms to the normal distribution, the high-density points are retained, otherwise the feature points are removed to achieve a better noise removal effect. At the same time, the signal-to-noise ratio and clarity of the data can be improved, and the visual effect of the image can be enhanced. The details are as follows.

[0027] ;in, , ; In the formula, represents the maximum threshold, Represents the average distance from a feature point to all feature points in the sparse point cloud. represents the standard deviation coefficient of distance, Represents a custom standard deviation coefficient. Represents the number of all feature points in the sparse point cloud, express The threshold of feature points. The threshold of feature points is greater than The maximum threshold indicates that the neighboring points around the point are relatively scattered and are considered outliers and are removed; the standard deviation coefficient is used to control the impact of the distance standard deviation on the distance threshold.

[0028] The sparse point cloud after statistical filtering is rasterized. Specifically, the grid is divided into regular cells according to the pixel size, and each point in the sparse point cloud is mapped to the corresponding grid cell. In the mapping process, direct mapping is used to determine the relationship between the sparse point cloud and the grid, and the maximum value of the sparse point cloud in each grid is used as the value of the grid. Exemplarily, this application uses the Point Cloud Processing Toolbox of MATLAB (a commercial mathematical software), which provides more point cloud processing functions, such as pcgrid(), which can directly combine point cloud objects with rasterization methods, such as setting the grid size and obtaining the rasterization results, which is more convenient for point cloud rasterization.

[0029] Step 104: extracting point clouds and reconstructing the surface of the digital surface model to obtain a surface mesh model. Extracting point clouds from the digital surface model pixel by pixel. , The coordinate calculation formula is as follows: , ; In the formula, , Represent point clouds x With point cloud The pixel coordinates of the point, , Respectively represent the coordinates of the upper left corner of the image ( , ), , Represents the pixel ground resolution ( , ). The Z value of the point cloud is the pixel coordinate where it is located ( x, ) corresponding to the pixel value, such as Figure 3 As shown, the surface Poisson reconstruction is performed based on the extracted point cloud to obtain the surface mesh model, which avoids the dense matching process in the traditional orthophoto generation process, completes the 2.5D scene reconstruction process from the digital surface model to the surface mesh model, and makes the entire orthophoto generation process simpler and faster.

[0030] Step 105: Texture mapping is performed on the surface mesh model to obtain a textured 3D model, and the textured 3D model is converted to the target coordinate system to obtain a target 3D model. In the embodiment of the present application, the surface mesh model needs to be uniformly colored during the mapping process to ensure that the color tone of the final generated digital surface model is consistent. The effect diagram of the texture mapping of the surface mesh model is shown in FIG. Figure 4 shown.

[0031] For example, color correction algorithms such as histogram equalization, grayscale world hypothesis, and color mapping can be used to adjust the texture color of each block according to the characteristics of each block and the reference image, so that it matches the color of the adjacent blocks or the overall model. The textures of different blocks are superimposed, gradient or layered by multi-scale texture fusion methods to make the color transition natural. The edge detection algorithm is used to detect the texture edge, and interpolation or repair is performed according to the color gradient so that the texture colors of adjacent blocks are seamlessly connected at the boundary. The global texture mapping method coordinates and optimizes the texture colors of all blocks to achieve an overall texture color uniformity effect.

[0032] In the embodiment of the present application, the coordinate system of the textured 3D model generated by default is EPSG:4490 (an ellipsoidal coordinate system), while in practical applications, UTM (Universal Transverse Mercator Projection) 3-degree zone or 6-degree zone projection is mostly used, so it is necessary to calculate the conversion matrix between EPSG:4490 and the target coordinate system, and convert the textured 3D model to a target 3D model containing the texture according to the conversion matrix. Exemplarily, the conversion can also be performed through GIS (Geographic Information System).

[0033] Step 106: Orthographically project the target three-dimensional model to obtain a two-dimensional orthographic image. Figure 5 As shown, the final two-dimensional orthophoto is obtained by orthographically projecting the target three-dimensional model and adaptively clipping it.

[0034] Although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative labor. The order of steps listed in this embodiment is only one way of executing the order of many steps and does not represent the only execution order. When the actual device or client product is executed, it can be executed in sequence or in parallel according to the method shown in this embodiment or the accompanying drawings (for example, in a parallel processor or multi-threaded processing environment).

[0035] like Figure 6 As shown, the embodiment of the present application also provides a device 600 for generating orthophotos based on scene reconstruction. The device includes: a data processing module 601, a motion recovery module 602, a rasterization module 603, a reconstruction module 604, a texture mapping module 605 and an orthophoto generation module 606, as follows.

[0036] The data processing module 601 is used to perform coordinate transformation on the images and establish adjacency relationships between the images.

[0037] The motion recovery module 602 is used to perform feature point matching on the images based on the adjacency relationship to obtain a matching pair file, and determine the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-space sparse point cloud.

[0038] The rasterization module 603 is used to filter and rasterize the object space sparse point cloud to obtain a digital surface model.

[0039] The reconstruction module 604 is used to extract point clouds and reconstruct the surface of the digital surface model to obtain a surface mesh model.

[0040] The texture mapping module 605 is used to perform texture mapping on the surface mesh model to obtain a textured three-dimensional model, and to convert the textured three-dimensional model to a target coordinate system to obtain a target three-dimensional model.

[0041] The orthophoto generation module 606 is used to perform orthographic projection on the target three-dimensional model to obtain a two-dimensional orthophoto.

[0042] The devices or modules described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. For the convenience of description, the above devices are described in various modules according to their functions. When implementing this application, the functions of each module can be implemented in the same or multiple software and / or hardware. Of course, the module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.

[0043] The method, device or module in the present application can be implemented in the form of a computer-readable program code. The controller can be implemented in any appropriate manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (English: Application Specific Integrated Circuit; Abbreviation: ASIC), a programmable logic controller and an embedded microcontroller. Examples of controllers include but are not limited to the following microcontrollers: ARC 625D, AtmelAT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included in it for implementing various functions can also be regarded as structures within the hardware component. Or even, the means for realizing various functions may be regarded as both a software module for realizing the method and a structure within a hardware component.

[0044] An embodiment of the present application also provides a device for a method of generating an orthophoto based on scene reconstruction, the device comprising: a processor; a memory for storing processor executable instructions; when the processor executes the executable instructions, the method described in the embodiment of the present application is implemented.

[0045] The embodiments of the present application also provide a non-volatile computer-readable storage medium on which a computer program or instruction is stored. When the computer program or instruction is executed, the method in the embodiments of the present application is implemented.

[0046] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist independently, or two or more modules may be integrated into one module.

[0047] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary hardware. Based on such an understanding, the technical solution of the present application can be essentially or partly reflected in the prior art in the form of a software product, or it can be reflected in the implementation process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods of each embodiment of the present application or some parts of the embodiments.

[0048] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. All or part of this application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, mobile communication terminals, multi-processor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0049] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, a person of ordinary skill in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present application.

Claims

1. A method for generating orthophotos based on scene reconstruction, characterized in that: include: Perform coordinate transformation on images and establish adjacency relationship between images; Performing feature point matching on the images based on the adjacency relationship to obtain a matching pair file, and determining the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-space sparse point cloud; Filtering and rasterizing the object space sparse point cloud to obtain a digital surface model; Performing point cloud extraction and surface reconstruction on the digital surface model to obtain a surface mesh model; Performing texture mapping on the surface mesh model to obtain a textured three-dimensional model, and converting the textured three-dimensional model to a target coordinate system to obtain a target three-dimensional model; The target three-dimensional model is orthographically projected to obtain a two-dimensional orthophoto.

2. The method according to claim 1, characterized in that The establishing of the adjacency relationship between the images includes: Define a grayscale set according to the information of the image, traverse all the pixels in the grayscale set and determine the adjacency type; wherein the adjacency type includes 4-adjacency, 8-adjacency and m-adjacency; If the pixel points satisfy the adjacency condition, the adjacency relationship between the images is established; wherein the adjacency condition includes that the pixel points are adjacent in position and satisfy a similarity criterion.

3. The method according to claim 1, characterized in that Before performing feature point matching on the image based on the adjacency relationship to obtain a matching pair file, the method further includes: The information of the image is resampled, and then the feature points of the resampled image are extracted.

4. The method according to claim 1, characterized in that The step of matching the image feature points based on the adjacency relationship to obtain a matching pair file includes: Performing feature point matching on adjacent images according to the adjacency relationship to obtain initial matching point pairs; Determining a matching threshold according to the number of the initial matching point pairs; The initial matching point pairs are filtered using the matching threshold to obtain valid matching point pairs, which constitute the matching pair file.

5. The method according to claim 1, characterized in that The step of determining the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-space sparse point cloud includes: Traversing the images corresponding to the matching pair files in order, and determining the best matching point pair therein according to the number of the valid matching point pairs and their intersection angle information; Calculate the positions and forward intersections of other valid point pairs in the matching pair file relative to the best matching point pair; The pose and the front intersection are optimized to obtain the object-space sparse point cloud.

6. The method according to claim 1, characterized in that The filtering and rasterizing of the object space sparse point cloud includes: Performing statistical filtering on the object space sparse point cloud; The sparse point cloud after statistical filtering is rasterized, and the maximum value of the sparse point cloud in each grid is used as the value of the grid.

7. The method according to claim 1, characterized in that The step of extracting point clouds and reconstructing surfaces of the digital surface model to obtain a surface mesh model includes: extracting a point cloud in the digital surface model pixel by pixel; The surface mesh model is obtained by performing surface Poisson reconstruction based on the extracted point cloud.

8. A device for generating orthophotos based on scene reconstruction, characterized in that: include: The data processing module is used to transform the coordinates of the images and establish the adjacency relationship between the images; A motion recovery module, used for performing feature point matching on the image based on the adjacency relationship to obtain a matching pair file, and determining the initial matching pair and the positions of other images relative to the initial matching pair to obtain an object-space sparse point cloud; A rasterization module, used for filtering and rasterizing the object space sparse point cloud to obtain a digital surface model; A reconstruction module, used for performing point cloud extraction and surface reconstruction on the digital surface model to obtain a surface mesh model; A texture mapping module is used to perform texture mapping on the surface mesh model to obtain a textured three-dimensional model, and to convert the textured three-dimensional model into a target coordinate system to obtain a target three-dimensional model; The orthophoto generation module is used to perform orthographic projection on the target three-dimensional model to obtain a two-dimensional orthophoto.

9. A device for generating an orthophoto based on scene reconstruction, characterized in that: include: processor; a memory for storing processor-executable instructions; When the processor executes the executable instructions, the method according to any one of claims 1 to 7 is implemented.

10. A non-volatile computer-readable storage medium, characterized in that: The device comprises a computer program or an instruction for storing the computer program or the instruction, which, when executed, enables the method according to any one of claims 1 to 7 to be implemented.

Images