A UAV-based Geographic Mapping Method and System

Abstract

The present invention discloses a geographic mapping method and system based on an unmanned aerial vehicle, relating to the technical field of geographic mapping; dividing a satellite image into multiple local regions; for each local region, dividing the local region through a Voronoi diagram to obtain multiple target sub-regions, and determining the route planning of the unmanned aerial vehicle for each target sub-region; positioning each unmanned aerial vehicle according to the satellite image, acquiring the images collected by each unmanned aerial vehicle after positioning, and stitching all the collected images to obtain a local map; fusing all the local maps into the satellite image to obtain a mapping map corresponding to the target region. After dividing the satellite image into local regions and then dividing the regions through the Voronoi diagram, each unmanned aerial vehicle is only responsible for one target sub-region, which improves the overall mapping speed. The stitching of local images and the fusion of satellite images ensure the accuracy and consistency of the final result, and improve the accuracy and efficiency of the overall mapping result.

Description

Technical Field

[0001] The present invention belongs to the technical field of geodetic surveying and mapping, and particularly relates to a geodetic surveying and mapping method and system based on an unmanned aerial vehicle (UAV). Background Art

[0002] With the rapid development of geographic information science and remote sensing technology, traditional geodetic surveying and mapping methods mainly rely on manual measurement and ground equipment. These methods are not only inefficient but also often limited by the working environment, with limited measurement accuracy and range. Although current surveying and mapping methods have introduced new technologies such as satellite positioning systems, lidar, and remote sensing images, there are still problems such as insufficient measurement accuracy and poor adaptability to complex terrains.

[0003] With the development of UAV technology, UAVs have been widely used in fields such as agriculture, environmental monitoring, and geodetic surveying and mapping. Using UAVs for surveying and mapping has the advantages of high efficiency, accuracy, and flexibility, and can perform efficient geographic data collection in complex terrains and dangerous environments. Existing technologies have begun to adopt a technical solution combining satellites and UAVs to achieve geodetic surveying and mapping.

[0004] However, existing technologies usually perform image acquisition through a single UAV. In this case, since the surveying and mapping standards of UAVs are unified, the data is easier to align and fuse. However, this also greatly increases the surveying and mapping time. Therefore, multiple UAVs need to operate in cooperation. However, when multiple UAVs perform image acquisition, due to inconsistent standards, it affects the accuracy of the final surveying and mapping results. Summary of the Invention

[0005] The object of the present invention is to solve the above problems and propose a geodetic surveying and mapping method and system based on an unmanned aerial vehicle (UAV).

[0006] In the first aspect of the implementation of the present invention, a geodetic surveying and mapping method based on an unmanned aerial vehicle (UAV) is first proposed. The method includes:

[0007] Obtain satellite images of a target area, and divide the satellite images into multiple local areas;

[0008] For each local area, divide the local area into multiple target sub-areas through a Voronoi diagram, and determine the route planning of the UAV for each target sub-area;

[0009] Locate each UAV according to the satellite images, and obtain the images collected by each UAV after positioning;

[0010] For the images collected by each UAV, splice all the collected images to obtain a local map;

[0011] Fuse all the local maps into the satellite images to obtain the surveying and mapping map corresponding to the target area.

[0012] Optionally, partitioning the satellite image into multiple local regions includes:

[0013] Preprocessing the satellite image to obtain an initial satellite image, and segmenting the initial satellite image according to a preset grid to obtain a set of sub-grid images;

[0014] For each sub-grid image in the set of sub-grid images, substituting the sub-grid image into a preset model to extract an image type and obtain a type label;

[0015] Fusing adjacent sub-grid images according to the type label to obtain multiple local regions.

[0016] Optionally, positioning each drone according to the satellite image includes:

[0017] Obtaining a set of images collected by the drone at the starting point of the route planning, and substituting each image in the set of images into a feature point detection model to extract multiple first feature point sets;

[0018] Substituting the satellite image into the feature point detection model to extract a second feature point set;

[0019] For each first feature point set, performing feature point matching on the first feature point set and the second feature point set through a feature matching algorithm;

[0020] If the feature point matching result is successful, then according to the matching point pairs, converting the coordinate space of the images collected by the drone to the coordinate space of the satellite image to obtain the corresponding position of the drone in the satellite image; the matching point pairs are the successfully matched first feature points and second feature points.

[0021] Optionally, for each image collected by the drone, stitching all the collected images to obtain a local map includes:

[0022] Obtaining the images collected by each drone in a time series, and performing binarization processing on each image to obtain a set of binary images;

[0023] For each image in the set of binary images, performing geometric analysis on the image to obtain boundary isocontours;

[0024] Aligning adjacent binary images according to the boundary isocontours, and repeating this step until all the binary images in the set of binary images are aligned to obtain an initial local map region;

[0025] Dynamically splice and adjust the initial local map according to a preset region of interest to obtain a local map region, determine the effective region corresponding to each acquisition image according to the local map region, and splice all the effective regions to obtain a local map.

[0026] Optionally, fusing all the local maps into the satellite image to obtain the mapping graph corresponding to the target region includes:

[0027] Extract the corresponding region in the satellite image according to the effective regions corresponding to all the acquisition images to obtain an effective region image pair; the effective region image pair includes an acquisition image and a satellite image;

[0028] For each effective region image pair, perform Gaussian pyramid decomposition on the effective region image pair to obtain a first Gaussian image set and a second Gaussian image set;

[0029] Calculate the difference between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the difference between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid;

[0030] For the first Laplacian pyramid and the second Laplacian pyramid, perform weighted fusion on the features at the same level to obtain a fused Laplacian pyramid;

[0031] Perform image reconstruction according to the fused Laplacian pyramid to obtain a local mapping graph, and obtain the mapping graph corresponding to the target region by acquiring all the local mapping graphs.

[0032] In the second aspect of the implementation of the present invention, a drone-based geographic mapping system is proposed, including:

[0033] A local region determination module, configured to acquire a satellite image of a target region, and perform region division on the satellite image to obtain a plurality of local regions;

[0034] A route planning module, configured to, for each local region, perform region division on the local region through a Voronoi diagram to obtain a plurality of target sub-regions, and determine the route planning of the drone for each target sub-region;

[0035] A drone positioning module, configured to position each drone according to the satellite image, and acquire the images collected by each drone after positioning;

[0036] An image splicing module, configured to, for the images collected by each drone, splice all the collected images to obtain a local map;

[0037] A mapping graph determination module, configured to fuse all the local maps into the satellite image to obtain the mapping graph corresponding to the target region.

[0038] Optionally, the local area determination module includes:

[0039] A preprocessing module, configured to preprocess the satellite image to obtain an initial satellite image, and segment the initial satellite image according to a preset grid to obtain a set of sub-grid images;

[0040] A type extraction module, configured to substitute each sub-grid image in the set of sub-grid images into a preset model to extract an image type and obtain a type label;

[0041] A grid image fusion module, configured to fuse adjacent sub-grid images according to the type label to obtain a plurality of local areas.

[0042] Optionally, the UAV positioning module includes:

[0043] A first feature point extraction module, configured to obtain a set of images collected by the UAV at the starting point of the route planning, and substitute each image in the set of images into a feature point detection model to extract a plurality of first feature point sets;

[0044] A second feature point extraction module, configured to substitute the satellite image into the feature point detection model to extract a second feature point set;

[0045] A feature point matching module, configured to perform feature point matching on each first feature point set and the second feature point set through a feature matching algorithm;

[0046] A UAV position determination module, configured to, if the feature point matching result is successful, convert the coordinate space of the image collected by the UAV to the coordinate space of the satellite image according to the matching point pairs, and obtain the corresponding position of the UAV in the satellite image; the matching point pairs are the successfully matched first feature points and second feature points.

[0047] Optionally, the image stitching module includes:

[0048] A binarization processing module, configured to obtain the images collected by each UAV in a time series, and perform binarization processing on each image to obtain a set of binary images;

[0049] A geometric analysis module, configured to perform geometric analysis on each image in the set of binary images to obtain boundary isograms;

[0050] An image alignment module, configured to align adjacent binary images of the image according to the boundary isograms, and repeat this step until all binary images in the set of binary images are aligned to obtain an initial local map area;

[0051] An effective area stitching module, configured to perform dynamic stitching adjustment on the initial local map according to a preset region of interest to obtain a local map area, determine an effective area corresponding to each acquisition image according to the local map area, and stitch all the effective areas to obtain a local map.

[0052] Optionally, the mapping chart determination module includes:

[0053] An effective area image pair determination module, configured to extract corresponding regions in the satellite image according to the effective areas corresponding to all acquisition images to obtain effective area image pairs; the effective area image pairs include acquisition images and satellite images;

[0054] A Gaussian pyramid decomposition module, configured to perform Gaussian pyramid decomposition on each effective area image pair to obtain a first Gaussian image set and a second Gaussian image set;

[0055] A Laplacian pyramid calculation module, configured to calculate the difference between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the difference between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid;

[0056] A feature fusion module, configured to perform weighted fusion on the features at the same level for the first Laplacian pyramid and the second Laplacian pyramid to obtain a fused Laplacian pyramid;

[0057] An image reconstruction module, configured to perform image reconstruction according to the fused Laplacian pyramid to obtain a local mapping chart, and obtain the mapping chart corresponding to the target area by acquiring all local mapping charts.

[0058] Advantages of the present invention:

[0059] The present invention proposes a geographic mapping method based on unmanned aerial vehicles (UAVs). Satellite images of a target area are acquired, and the satellite images are divided into multiple local areas; for each local area, the local area is divided into multiple target sub-areas through a Voronoi diagram, and the route planning of the UAV for each target sub-area is determined; each UAV is positioned according to the satellite images, and the images collected by each UAV after positioning are acquired; for the images collected by each UAV, all the collected images are stitched to obtain a local map; all the local maps are fused into the satellite images to obtain a mapping diagram corresponding to the target area. By dividing the satellite images into local areas and determining which type of UAV is used for which area according to the category, the mapping efficiency is improved. Then, through the Voronoi diagram for area division, each UAV is only responsible for one target sub-area, so that the operation efficiency can be improved, and at the same time, the UAVs are prevented from repeating operations on the same area, improving the overall mapping speed. Moreover, through the precise positioning of the UAVs before data collection and the flight path planning, the flight synchronization error can be reduced, and the alignment accuracy of the data can be improved. Finally, through the stitching of the local images and the fusion with the satellite images, the accuracy and consistency of the final result are ensured, improving the accuracy and efficiency of the overall mapping result. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The present invention will be further described below with reference to the accompanying drawings.

[0061] Figure 1 FIG. [X] is a flowchart of a geographic mapping method based on unmanned aerial vehicles according to an embodiment of the present invention;

[0062] Figure 2 FIG. [Y] is a framework diagram of a geographic mapping system based on unmanned aerial vehicles according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the present invention, descriptions such as "first" and "second" are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on what can be achieved by those of ordinary skill in the art. When the combination of technical solutions is contradictory or cannot be achieved, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0064] Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0065] An embodiment of the present invention provides a geographic mapping method based on an unmanned aerial vehicle. Refer to Figure 1 , Figure 1 which is a flowchart of a geographic mapping method based on an unmanned aerial vehicle provided by an embodiment of the present invention. The method includes the following steps:

[0066] S101, obtain a satellite image of a target area, and divide the satellite image into multiple local areas;

[0067] S102, for each local area, divide the local area into multiple target sub-areas through a Voronoi diagram, and determine the route planning of the unmanned aerial vehicle for each target sub-area;

[0068] S103, locate each unmanned aerial vehicle according to the satellite image, and obtain the images collected by each unmanned aerial vehicle after positioning;

[0069] S104, for the images collected by each unmanned aerial vehicle, splice all the collected images to obtain a local map;

[0070] S105, fuse all the local maps into the satellite image to obtain a mapping map corresponding to the target area.

[0071] Based on a geographic mapping method based on an unmanned aerial vehicle (UAV) provided by an embodiment of the present invention, by dividing a local area of a satellite image and determining which type of UAV is used for which area according to categories, the mapping efficiency is improved. Then, by dividing the area using a Voronoi diagram, each UAV is only responsible for one target sub-area, thereby improving the operation efficiency and avoiding the UAVs from repeating operations on the same area, improving the overall mapping speed. Furthermore, by precisely positioning the UAV before data collection and planning the flight path, the flight synchronization error can be reduced and the alignment accuracy of the data can be improved. Finally, through the stitching of local images and the fusion with satellite images, the accuracy and consistency of the final result are ensured, improving the accuracy and efficiency of the overall mapping result.

[0072] In one implementation, satellite images can be used to obtain the basic information of the target area in a large range and quickly, reducing the time and cost of manual measurement. Through the area division of the Voronoi diagram, the working range and path planning of each UAV can be accurately determined, ensuring that the UAVs will not interfere with each other during task execution, thereby improving the efficiency and accuracy of the mapping operation; through the precise positioning of each UAV, the accuracy of image acquisition can be ensured, and the areas that cannot be carefully captured in the satellite image can be covered.

[0073] In one implementation, by having multiple UAVs execute tasks in different sub-areas, multiple areas can be scanned in detail simultaneously, reducing the overall mapping time. The local images collected by the UAVs can be stitched and fused to obtain information with higher resolution and more perspectives. By fusing multiple local images into the satellite image, a more detailed and comprehensive mapping diagram can be obtained.

[0074] In one implementation, using the Voronoi diagram for area division is a classic geometric segmentation method, which can effectively avoid overlapping operations between UAVs. The division of each sub-area makes each UAV only responsible for one target sub-area, thereby improving the operation efficiency and avoiding interference between UAVs, which helps to improve the accuracy of the UAV flight path planning and ensure that each UAV can complete the image acquisition task according to the predetermined route.

[0075] In one implementation, when the UAV is taking pictures, in order to ensure the accuracy of the shooting, the captured images often contain overlapping parts. Therefore, the stitching algorithm needs to process the data of these overlapping areas to ensure the accuracy of the finally stitched images and improve the quality of the final mapping diagram.

[0076] In one implementation, before each drone collects images, positioning the drone through satellite images is a key step in solving the positioning error. High-precision positioning can ensure that the drone can accurately cover the target area during the mission execution, avoid error accumulation, and reduce the data fusion error caused by positioning problems.

[0077] In one embodiment, dividing the satellite image into multiple local regions includes:

[0078] Preprocess the satellite image to obtain an initial satellite image, and segment the initial satellite image according to a preset grid to obtain a set of sub-grid images;

[0079] For each sub-grid image in the set of sub-grid images, substitute the sub-grid image into a preset model to extract the image type to obtain a type label;

[0080] Fuse adjacent sub-grid images according to the type label to obtain multiple local regions.

[0081] In one implementation, preprocess the satellite image to remove noise and irrelevant information, and then segment the image according to a preset grid to make the image processing more systematic and standardized. The size and resolution of each sub-grid can be adjusted according to specific requirements to balance the processing accuracy and efficiency; the preprocessing is common satellite image processing methods such as denoising and image enhancement to enhance the image clarity; the size of the preset grid is determined by technicians.

[0082] In one implementation, the image types include building areas, road areas, water areas, and vegetation areas. Among them, building areas and road areas are defined as the first type label, and water areas and vegetation areas are defined as the second type label; the regional scenes corresponding to the first type label are often more complex, so multi-rotor drones are used. The regional scenes corresponding to the second type label are basically water and vegetation areas, and fixed-wing drones are usually simply used; the preset model can be common regional type recognition models on the market such as U-Net, DeepLabV3+, and SegNet.

[0083] In one implementation, by performing type label recognition on adjacent sub-grid images, regional fusion can be performed more intelligently. For example, adjacent sub-grid images of different types such as buildings and roads can be recognized and merged into a large local area for more efficient subsequent analysis and application, providing a smoother and more reasonable regional division. By reasonably merging adjacent sub-grids, local regions that more conform to the actual geographical environment can be obtained.

[0084] In one embodiment, positioning each drone according to the satellite image includes:

[0085] Obtain the image set collected by the drone at the starting point of the route planning. For each image in the image set, substitute it into the feature point detection model to extract feature points and obtain multiple first feature point sets;

[0086] Substitute the satellite image into the feature point detection model to extract feature points and obtain a second feature point set;

[0087] For each first feature point set, perform feature point matching on the first feature point set and the second feature point set through a feature matching algorithm;

[0088] If the feature point matching result is successful, then according to the matching point pairs, convert the coordinate space of the images collected by the drone to the coordinate space of the satellite image to obtain the corresponding position of the drone in the satellite image; the matching point pairs are the first feature points and the second feature points that match successfully.

[0089] In one implementation, by matching the feature points in the images collected by the drone with those in the satellite image, the image coordinates of the drone can be converted to the coordinate system of the satellite image, thereby achieving high-precision positioning and providing a basis for subsequent image stitching and fusion; the feature point matching process can be automatically executed, reducing the intervention of human operations, especially in complex and dangerous environments. The drone can autonomously perform positioning and adjust its path during the task execution, improving the automation level of task completion.

[0090] In one implementation, before substituting each image in the image set into the feature point detection model and substituting the satellite image into the feature point detection model, scale normalization is performed on the satellite image and each image in the image set. To ensure accuracy, usually the size of each image in the image set is used as the standard.

[0091] In one implementation, the image set collected by the drone at the starting point of the route planning is the images randomly captured by the drone flying at the current starting point. Based on these collected images, the drone can achieve coordinate mapping with the satellite image. The feature point detection model can be SIFT, SURF, ORB, etc.; the feature matching algorithm can be FLANN, SuperGlue, RANSAC, etc.; the feature point matching is based on the feature points of each image in the image set. For example, when there is an image in the image set with 100 feature points, and then these feature points are matched with the feature points in the satellite image. If 80 feature points are successfully matched, there is a threshold for successful matching. The threshold is the proportion of all feature points in the current image that are successfully matched, and the proportion should be greater than or equal to 80%. Then, the coordinate space of this image can be transformed into the coordinate space of the satellite image; for each unmanned driving area, a coordinate system is established with the starting point as the origin and the initial driving direction as the positive direction of the Y-axis; the coordinate space of the satellite image is the coordinate system established based on the satellite image; when the corresponding position of the drone in the satellite image is determined, the corresponding area of the image captured by the drone in the satellite image can be determined according to the driving path of the drone, providing a basis for subsequent image fusion.

[0092] In one embodiment, for each image collected by the drone, stitching all the collected images to obtain a local map includes:

[0093] Obtain the images collected by each drone in the time series, and perform binarization processing on each image to obtain a binary image set;

[0094] For each image in the binary image set, perform geometric analysis on the image to obtain boundary isocontours;

[0095] Align the binary images adjacent to the image according to the boundary isocontours, and repeat this step until all binary images in the binary image set are aligned to obtain an initial local map area;

[0096] Perform dynamic stitching adjustment on the initial local map according to the preset region of interest to obtain a local map area, determine the effective area corresponding to each collected image according to the local map area, and stitch all the effective areas to obtain a local map.

[0097] In one implementation, binarization can simplify the complexity of image processing, convert the image into only black and white, thereby removing noise, background, and irrelevant information in the image. Through binarization, key target areas such as object boundaries and contours are more prominent, which helps subsequent image analysis and feature extraction; geometric analysis can extract edges and contours in the image. The extraction of boundary isocontours helps to understand the shape characteristics of objects or scenes in the image, and can provide necessary structural information for subsequent alignment and stitching to ensure that features such as edges and angles between images can be effectively matched.

[0098] In one implementation, to determine the effective region corresponding to each acquired image according to the local map region, specifically, obtain the overlapping part of each acquired image and other images after dynamic stitching adjustment, determine the proportion of the overlapping part in the current image. If the proportion exceeds the preset ratio, remove the overlapping part in the current image, and record the remaining part as the effective region. If the proportion does not exceed the preset ratio, retain the overlapping part, and record the retained image region as the effective region.

[0099] In one implementation, aligning images through boundary isocontours can ensure precise spatial docking of adjacent images, thus avoiding stitching errors and improving the quality and accuracy of image stitching. Dynamic stitching adjustment can optimize local images according to the preset region of interest (ROI). The region of interest generally refers to the area that needs to be focused on in the image. Dynamic adjustment can avoid interference from irrelevant regions, ensure that the local map region after stitching accurately reflects the details of the target region, and at the same time reduce the distortion generated during stitching.

[0100] In one implementation, to perform geometric analysis on the image to obtain boundary isocontours, specifically, extract the contours through regions with the same pixel values in the image. Each contour is a closed path composed of multiple points, representing the boundary of an object in the image. In this way, the contours of objects can be identified from the image and used as isocontours for subsequent processing. The extracted contours can be multiple discrete point sets, representing the boundaries of objects or regions. The contour lines separate different regions in the image because when using drones to take pictures for mapping and supplementing satellite images, the objects to be supplemented are often regular graphics such as buildings. The preset region of interest is the place in the satellite image that needs to be identified in detail, such as the area of houses and buildings. The specific preset region of interest is determined by technicians. Dynamic stitching adjustment is based on the overlapping region. If the image overlapping region is larger than the preset first area, usually a wider region of interest is required. At this time, the width of the original region of interest will be changed to 1.2 times the original to better smooth the exposure difference and tone change. If the image overlapping region is smaller than the preset second area, usually the width of the region of interest needs to be reduced. At this time, the width of the original region of interest will be changed to 0.8 times the original to improve the stitching efficiency and reduce unnecessary calculations.

[0101] In one implementation, through stitching the local map regions, multiple images can be synthesized into a larger and more detailed image, and more comprehensive regional information can be obtained by using the image data at multiple time points, thus forming a seamless overall image, which can provide higher precision and coverage.

[0102] In one embodiment, fusing all local maps into the satellite image to obtain the mapping diagram corresponding to the target region includes:

[0103] Extract the corresponding regions in the satellite image according to the valid regions corresponding to all the acquired images to obtain a valid region image pair; the valid region image pair includes the acquired image and the satellite image;

[0104] For each valid region image pair, perform Gaussian pyramid decomposition on the valid region image pair to obtain a first Gaussian image set and a second Gaussian image set;

[0105] Calculate the differences between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the differences between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid;

[0106] For the first Laplacian pyramid and the second Laplacian pyramid, perform weighted fusion on the features at the same level to obtain a fused Laplacian pyramid;

[0107] Perform image reconstruction according to the fused Laplacian pyramid to obtain a local mapping, and obtain the mapping corresponding to the target region by acquiring all the local mappings.

[0108] In one implementation, by performing Gaussian pyramid and Laplacian pyramid decomposition on the image, the image can be decomposed into different spatial frequency bands from low frequency to high frequency, so that when fusing at different levels, it can better retain low-frequency structural information such as large-scale background and geographical morphology and highlight high-frequency details such as textures and feature points. It can be adaptively processed according to the characteristics of different images such as the resolution and exposure differences between drone images and satellite images, avoiding detail loss or over-smoothing caused by direct weighted averaging.

[0109] In one implementation, for each layer of the Laplacian pyramid, it includes a low-frequency layer and a high-frequency layer. Among them, the Laplacian pyramid obtained through the first preset number of layers of the Gaussian pyramid is the high-frequency layer, and the Laplacian pyramid obtained through the Gaussian pyramid after the preset number of layers is the low-frequency layer. The weight corresponding to the Laplacian pyramid of the first Laplacian pyramid in the low-frequency layer is 0.7, and the weight corresponding to the second Laplacian pyramid is 0.3; the weight corresponding to the Laplacian pyramid of the first Laplacian pyramid in the high-frequency layer is 0.2, and the weight corresponding to the second Laplacian pyramid is 0.8.

[0110] In one implementation, satellite images usually contain a larger field of view but lower resolution; while drone images provide higher resolution but a smaller coverage area. Through multi-resolution pyramid decomposition and difference calculation of the Laplacian pyramid, it can effectively extract and enhance the local details of the image, making the features of the target region clearer and helping to improve the matching accuracy of the image.

[0111] In one implementation manner, the specific process of obtaining a local surveying and mapping map through image reconstruction based on the fused Laplacian pyramid is as follows: starting from the bottom layer of the fused Laplacian pyramid, reconstruct layer by layer upwards. For each layer of the image, use upsampling to expand it to the same size as the upper layer, and then overlay the two images until the top layer image is finally obtained.

[0112] In one implementation manner, during the process of image fusion, through the layer-by-layer fusion of the Laplacian pyramid, smooth transition can be achieved at the seams, avoiding obvious stitching marks caused by differences in brightness or texture between different images, making the finally generated local surveying and mapping map more natural and seamless. Especially when stitching drone images and satellite images, it can effectively eliminate the disharmony caused by the image seams; by decomposing the image into multiple local regions and processing layer by layer, the processing difficulty of large images can be effectively reduced. By performing Gaussian pyramid decomposition on each pair of effective region images, the features of each local region can be evenly retained at both low-frequency and high-frequency levels. The finally generated local surveying and mapping maps are of higher quality. After merging all the local surveying and mapping maps, the surveying and mapping map of the overall target area is obtained, which can not only maintain the accuracy but also handle larger geographical areas.

[0113] Based on the same inventive concept, the embodiments of the present invention also provide a drone-based geographical surveying and mapping system. Refer to Figure 2 , Figure 2 which is a framework diagram of a drone-based geographical surveying and mapping system provided by the embodiments of the present invention, including:

[0114] A local area determination module, configured to obtain a satellite image of a target area and divide the satellite image into multiple local areas;

[0115] A route planning module, configured to, for each local area, divide the local area into multiple target sub-areas through a Voronoi diagram and determine the route planning of the drone for each target sub-area;

[0116] A drone positioning module, configured to position each drone according to the satellite image and obtain the images collected by each drone after positioning;

[0117] An image stitching module, configured to, for the images collected by each drone, stitch all the collected images to obtain a local map;

[0118] A surveying and mapping map determination module, configured to fuse all the local maps into the satellite image to obtain the surveying and mapping map corresponding to the target area.

[0119] A geographic mapping system based on an unmanned aerial vehicle provided by an embodiment of the present invention divides a local area of a satellite image, determines which areas use what types of unmanned aerial vehicles according to categories, improves the mapping efficiency, and then divides the area through a Voronoi diagram, so that each unmanned aerial vehicle is only responsible for one target sub-area, thereby improving the operation efficiency, avoiding the repeated operation of the unmanned aerial vehicle on the same area, and improving the overall mapping speed. Then, through the precise positioning of the unmanned aerial vehicle before data collection and the flight path planning, the flight synchronization error can be reduced, and the alignment accuracy of the data can be improved. Finally, through the stitching of local images and the fusion of satellite images, the accuracy and consistency of the final result are ensured, and the accuracy and efficiency of the overall mapping result are improved.

[0120] In one embodiment, the local area determination module includes:

[0121] A preprocessing module for preprocessing the satellite image to obtain an initial satellite image and segmenting the initial satellite image according to a preset grid to obtain a set of sub-grid images;

[0122] A type extraction module for substituting each sub-grid image in the set of sub-grid images into a preset model to extract the image type and obtain a type label;

[0123] A grid image fusion module for fusing adjacent sub-grid images according to the type label to obtain a plurality of local areas.

[0124] In one embodiment, the unmanned aerial vehicle positioning module includes:

[0125] A first feature point extraction module for obtaining an image set collected by the unmanned aerial vehicle at the starting point of the route planning, and substituting each image in the image set into a feature point detection model to extract a plurality of first feature point sets;

[0126] A second feature point extraction module for substituting the satellite image into a feature point detection model to extract a second feature point set;

[0127] A feature point matching module for, for each first feature point set, performing feature point matching on the first feature point set and the second feature point set through a feature matching algorithm;

[0128] An unmanned aerial vehicle position determination module for, if the feature point matching result is successful, converting the coordinate space of the image collected by the unmanned aerial vehicle to the coordinate space of the satellite image according to the matching point pair to obtain the corresponding position of the unmanned aerial vehicle in the satellite image; the matching point pair is the successfully matched first feature point and second feature point.

[0129] In one embodiment, the image stitching module includes:

[0130] The binarization processing module is used to obtain the images collected by each drone in the time series, and perform binarization processing on each image to obtain a set of binary images;

[0131] The geometric analysis module is used to perform geometric analysis on each image in the set of binary images to obtain boundary isocontours;

[0132] The image alignment module is used to align adjacent binary images of the image according to the boundary isocontours, and repeat this step until all binary images in the set of binary images are aligned to obtain an initial local map area;

[0133] The effective area stitching module is used to perform dynamic stitching adjustment on the initial local map according to a preset region of interest to obtain a local map area, determine the effective area corresponding to each collected image according to the local map area, and stitch all effective areas to obtain a local map.

[0134] In one embodiment, the mapping chart determination module includes:

[0135] The effective area image pair determination module is used to extract the corresponding regions in the satellite image according to the effective areas corresponding to all the collected images to obtain an effective area image pair; the effective area image pair includes the collected image and the satellite image;

[0136] The Gaussian pyramid decomposition module is used to perform Gaussian pyramid decomposition on each effective area image pair to obtain a first Gaussian image set and a second Gaussian image set;

[0137] The Laplacian pyramid calculation module is used to calculate the difference between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the difference between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid;

[0138] The feature fusion module is used to perform weighted fusion on the features of the same level for the first Laplacian pyramid and the second Laplacian pyramid to obtain a fused Laplacian pyramid;

[0139] The image reconstruction module is used to perform image reconstruction according to the fused Laplacian pyramid to obtain a local mapping chart, and obtain the mapping chart corresponding to the target area by acquiring all local mapping charts.

[0140] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. A method for geographical mapping based on drones, characterized in that, The method includes: Obtain a satellite image of a target area, and divide the satellite image into multiple local areas; For each local area, divide the local area by a Voronoi diagram to obtain multiple target sub-areas, and determine the route planning of the unmanned aerial vehicle (UAV) for each target sub-area; Locate each UAV according to the satellite image, and obtain the images collected by each UAV after positioning; For the images collected by each UAV, splice all the collected images to obtain a local map; Fuse all the local maps into the satellite image to obtain a mapping chart corresponding to the target area; For the images collected by each UAV, splicing all the collected images to obtain a local map includes: Obtain the images collected by each UAV in a time series, and perform binary processing on each image to obtain a binary image set; For each image in the binary image set, perform geometric analysis on the image to obtain boundary isograms; Align the binary images adjacent to the image according to the boundary isograms, and repeat this step until all the binary images in the binary image set are aligned to obtain an initial local map area; Perform dynamic splicing adjustment on the initial local map according to a preset region of interest to obtain a local map area, determine the valid area corresponding to each collected image according to the local map area, and splice all the valid areas to obtain a local map.

2. The method for geodetic survey based on an unmanned aerial vehicle according to claim 1, wherein Dividing the satellite image into multiple local areas includes: Preprocess the satellite image to obtain an initial satellite image, and segment the initial satellite image according to a preset grid to obtain a set of sub-grid images; For each sub-grid image in the set of sub-grid images, substitute the sub-grid image into a preset model for image type extraction to obtain a type label; Fuse adjacent sub-grid images according to the type label to obtain multiple local areas.

3. The method for geodetic surveying based on an unmanned aerial vehicle according to claim 1, characterized in that, Locating each UAV according to the satellite image includes: Obtain a set of images collected by the UAV at the starting point of the route planning, and substitute each image in the set of images into a feature point detection model for feature point extraction to obtain multiple first feature point sets; Substitute the satellite image into the feature point detection model for feature point extraction to obtain a second feature point set; For each first feature point set, perform feature point matching on the first feature point set and the second feature point set through a feature matching algorithm; If the feature point matching result is successful, then according to the matching point pairs, convert the coordinate space of the UAV-collected image to the coordinate space of the satellite image to obtain the corresponding position of the UAV in the satellite image; the matching point pairs are the successfully matched first feature points and second feature points.

4. A method for geodetic surveying based on an unmanned aerial vehicle according to claim 1, characterized in that, Fusing all the local maps into the satellite image to obtain a mapping chart corresponding to the target area includes: Extract the corresponding area in the satellite image according to the valid areas corresponding to all the collected images to obtain valid area image pairs; the valid area image pairs include the collected images and the satellite images; For each valid area image pair, perform Gaussian pyramid decomposition on the valid area image pair to obtain a first Gaussian image set and a second Gaussian image set; Calculate the difference between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the difference between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid; For the first Laplacian pyramid and the second Laplacian pyramid, perform weighted fusion on the features at the same level to obtain a fused Laplacian pyramid; Perform image reconstruction based on the fused Laplacian pyramid to obtain a local surveying and mapping map, and obtain the surveying and mapping map corresponding to the target area by acquiring all local surveying and mapping maps.

5. An unmanned aerial vehicle-based geographical mapping system, characterized in that, The system includes: A local area determination module, configured to acquire a satellite image of a target area, and perform area division on the satellite image to obtain a plurality of local areas; A route planning module, configured to, for each local area, perform area division on the local area through a Voronoi diagram to obtain a plurality of target sub-areas, and determine the route planning of the unmanned aerial vehicle for each target sub-area; An unmanned aerial vehicle positioning module, configured to position each unmanned aerial vehicle according to the satellite image, and acquire the images collected by each unmanned aerial vehicle after positioning; An image stitching module, configured to, for the images collected by each unmanned aerial vehicle, stitch all the collected images to obtain a local map; A surveying and mapping map determination module, configured to fuse all local maps into the satellite image to obtain the surveying and mapping map corresponding to the target area; The image stitching module includes: A binarization processing module, configured to acquire the images collected by each unmanned aerial vehicle in a time series, and perform binarization processing on each image to obtain a binarized image set; A geometric analysis module, configured to, for each image in the binarized image set, perform geometric analysis on the image to obtain boundary isograms; An image alignment module, configured to align adjacent binarized images of the image according to the boundary isograms, and repeat this step until all binarized images in the binarized image set are aligned to obtain an initial local map area; An effective area stitching module, configured to perform dynamic stitching adjustment on the initial local map according to a preset region of interest to obtain a local map area, determine the effective area corresponding to each collected image according to the local map area, and stitch all effective areas to obtain a local map.

6. The geodetic surveying system based on an unmanned aerial vehicle according to claim 5, characterized in that, The local area determination module includes: A preprocessing module, configured to preprocess the satellite image to obtain an initial satellite image, and segment the initial satellite image according to a preset grid to obtain a sub-grid image set; A type extraction module, configured to, for each sub-grid image in the sub-grid image set, substitute the sub-grid image into a preset model for image type extraction to obtain a type label; A grid image fusion module, configured to fuse adjacent sub-grid images according to the type label to obtain a plurality of local areas.

7. A drone-based geographic mapping system according to claim 5, characterized in that, The unmanned aerial vehicle positioning module includes: A first feature point extraction module, configured to acquire an image set collected by the unmanned aerial vehicle at the starting point of the route planning, and substitute each image in the image set into a feature point detection model for feature point extraction to obtain a plurality of first feature point sets; A second feature point extraction module, configured to substitute the satellite image into the feature point detection model for feature point extraction to obtain a second feature point set; The feature point matching module is used to perform feature point matching on the first feature point set and the second feature point set through a feature matching algorithm for each first feature point set; The UAV position determination module is used to, if the feature point matching result is successful, convert the coordinate space of the UAV-acquired image to the coordinate space of the satellite image according to the matching point pairs to obtain the corresponding position of the UAV in the satellite image; the matching point pairs are the successfully matched first feature points and second feature points.

8. The geodetic surveying system based on an unmanned aerial vehicle according to claim 5, characterized in that, The mapping chart determination module includes: The effective area image pair determination module is used to extract the corresponding area in the satellite image according to the effective areas corresponding to all the acquired images to obtain an effective area image pair; the effective area image pair includes the acquired image and the satellite image; The Gaussian pyramid decomposition module is used to perform Gaussian pyramid decomposition on each effective area image pair to obtain a first Gaussian image set and a second Gaussian image set; The Laplacian pyramid calculation module is used to calculate the difference between adjacent-level Gaussian images for the first Gaussian image set to obtain a first Laplacian pyramid, and calculate the difference between adjacent-level Gaussian images for the second Gaussian image set to obtain a second Laplacian pyramid; The feature fusion module is used to perform weighted fusion on the features of the same level for the first Laplacian pyramid and the second Laplacian pyramid to obtain a fused Laplacian pyramid; The image reconstruction module is used to perform image reconstruction according to the fused Laplacian pyramid to obtain a local mapping chart, and obtain the mapping chart corresponding to the target area by acquiring all the local mapping charts.

Images