Urban three-dimensional surveying and mapping method and system based on unmanned aerial vehicle

Abstract

The invention relates to the technical field of three-dimensional surveying and mapping, in particular to a city three-dimensional surveying and mapping method and system based on an unmanned aerial vehicle, and the method comprises the steps: obtaining a surveying and mapping route of the unmanned aerial vehicle; performing data acquisition on the target city based on the surveying and mapping route, and obtaining point cloud data and a video image; obtaining a feature tag based on the video image; acquiring a key labeling position based on the video image and the feature tag; generating an initial three-dimensional model of the target city based on the point cloud data; judging whether key coordinates corresponding to the key labeling positions exist in the initial three-dimensional model or not; if the key coordinates corresponding to the key labeling positions do not exist in the initial three-dimensional model, obtaining the key coordinates; and generating a target three-dimensional model based on the feature tag, the key labeling position and the initial three-dimensional model. According to the invention, the accuracy of the urban three-dimensional surveying and mapping result can be improved.

Description

Technical Field

[0001] The present application relates to the field of three-dimensional surveying and mapping technology, and in particular to a method and system for urban three-dimensional surveying and mapping based on drones. Background Art

[0002] With the rapid development of urban construction and the growing demand for digital urban management, urban 3D mapping technology plays a vital role in many fields, including urban planning, architectural design, disaster monitoring, and smart city construction. Traditional urban 3D mapping methods, such as ground surveying and aerial photogrammetry, have certain limitations. Ground surveying is inefficient and requires a large investment of manpower, material resources, and time when mapping large urban areas. It is also difficult to implement in complex terrain or areas with heavy traffic. Although aerial photogrammetry can achieve rapid measurement of large areas, it has strict requirements on weather conditions, poor data collection flexibility, and complex post-processing of the acquired data, making it difficult to accurately capture detailed urban information.

[0003] In recent years, drone technology has been widely used in urban 3D mapping due to its flexible operation, low cost, and limited geographical restrictions. Drones are used to collect 3D point cloud data of cities and then generate 3D models based on this data, thus enabling 3D mapping of cities. However, during data collection, due to the complex urban environment with its numerous high-rise buildings and dense vegetation, some areas of the drone's flight are blocked, preventing the sensor from obtaining valid data. Furthermore, due to data anomalies or data loss during transmission, the collected point cloud data is incomplete, resulting in the loss of key coordinates. Ultimately, the generated 3D model cannot truly and accurately reflect the actual conditions of the city and cannot meet the needs of high-precision 3D urban mapping. Summary of the Invention

[0004] In order to help improve the accuracy of urban three-dimensional mapping results, the present application provides a method and system for urban three-dimensional mapping based on drones.

[0005] In the first aspect, the present application provides a method for urban 3D mapping based on drones, which adopts the following technical solutions: A method for urban three-dimensional mapping based on a drone, comprising: Get the UAV's mapping route; Collect data on the target city based on the surveying route and obtain point cloud data and video images; Based on the video image, obtaining a feature label; Based on the video image and the feature labels, obtaining key annotation positions; generating an initial three-dimensional model of a target city based on the point cloud data; Determining whether there are key coordinates corresponding to the key annotation positions in the initial three-dimensional model; If the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model, acquiring the key coordinates; A target three-dimensional model is generated based on the feature labels, the key annotation positions, and the initial three-dimensional model.

[0006] By adopting the above technical solution, the drone mapping route is first obtained, and data of the target city is collected accordingly to obtain point cloud data and video images; then, feature labels are obtained from the video images, and the key annotation positions are determined by combining the video images and feature labels; then, an initial 3D model is generated based on the point cloud data, and it is determined whether there are key coordinates corresponding to the key annotation positions in the initial 3D model. If not, it means that the key coordinates corresponding to the key annotation positions are missing, and therefore it is necessary to further obtain the key coordinates; finally, a target 3D model is generated based on the feature labels, key annotation positions and the initial 3D model; the drone is used to flexibly collect data, and combined with feature analysis of the video images, it helps to more clearly know whether there are key coordinates missing. If so, the key coordinates are supplemented, which effectively solves the problem of missing key data in drone mapping, and can generate a more realistic and accurate target 3D model, greatly improving the accuracy of urban 3D mapping results and meeting the needs of high-precision urban 3D mapping.

[0007] Optionally, obtaining key annotation positions based on the video image and the feature labels includes: Based on the feature labels, obtaining target categories of different target objects in the video image; Based on the video image, obtaining target styles of different target objects; Based on the object category and the object style, key annotation positions are obtained.

[0008] Optionally, obtaining key annotation positions based on the target category and the target style includes: Based on the target category, obtaining a special position corresponding to the target object; Based on the special position, generating a key annotation position; Analyzing the video image and determining whether different target categories and / or different target styles exist simultaneously in the video image; If different target categories exist simultaneously in the video image, obtaining a first dividing line between the different target categories; Based on the first dividing line and preset surveying and mapping rules, obtaining key annotation positions; If different target styles exist simultaneously in the video image, obtaining a second dividing line between the different target styles; Based on the second dividing line and the preset surveying and mapping rules, key annotation positions are obtained.

[0009] Optionally, if different target styles exist simultaneously in the video image, obtaining a second dividing line between different target styles includes: If different target styles exist simultaneously in the video image, dividing the video image into a plurality of sub-areas based on the target styles; Get the contour boundaries of different sub-regions; Determining whether the outline boundary is a clear boundary; If the outline boundary is the clear boundary, obtaining a second dividing line based on the outline boundary; If the outline boundary is a fuzzy boundary, obtaining image brightness and image hue corresponding to different sub-regions; Obtaining a difference position based on the image brightness and the image hue; Obtaining the change type corresponding to the difference position; Based on the change type, a second dividing line is obtained.

[0010] Optionally, obtaining a second dividing line based on the change type includes: If the change type is transient, the difference position is used as the second dividing line; If the change type is a gradual change type, obtaining a first brightness value and a first hue based on the image brightness and the image hue corresponding to different sub-regions; Get the second brightness value and second hue corresponding to the gradient area; Obtaining a brightness difference value based on the first brightness value and the second brightness value; Obtaining a hue difference based on the first hue and the second hue; A second dividing line is acquired based on the brightness difference value and the hue difference value.

[0011] Optionally, if the initial three-dimensional model does not contain key coordinates corresponding to the key annotation position, obtaining the key coordinates includes: If the key coordinates corresponding to the key annotation position do not exist in the initial three-dimensional model, obtaining a first attribute of the target object based on the video image; Based on the first attribute, determining whether a symmetric reference coordinate exists; If the symmetric reference coordinates exist, obtaining symmetric mirror coordinates based on the symmetric reference coordinates, and using the symmetric mirror coordinates as the missing key coordinates corresponding to the key marked position; If the symmetric reference coordinates do not exist, obtaining a similar object to the target object; Based on the similar objects, obtaining similar coordinates; Based on the similar coordinates, the key coordinates are calculated.

[0012] Optionally, if the symmetric reference coordinates do not exist, obtaining a similar object to the target object includes: If the symmetric reference coordinates do not exist, obtaining the coordinates of the highest point of the target object based on the video image and the initial three-dimensional model; Based on the highest point coordinates, obtaining a selectable object; Get the second attribute of the optional object; If the first attribute and the second attribute meet a preset similarity standard, obtaining an optional annotation position based on the key annotation position; Determine whether there are optional coordinates corresponding to the optional annotation position; If the optional coordinates exist, the optional object is used as the similar object.

[0013] Optionally, the calculating the key coordinates based on the similar coordinates includes: Get the reference coordinates and scale; Calculating a first distance and a second distance based on the reference coordinates, the similar coordinates, and the key annotation position; The key coordinates are calculated based on the first distance, the second distance, and the scale.

[0014] Secondly, this application also discloses a UAV-based urban 3D mapping system, which adopts the following technical solutions: A UAV-based urban 3D mapping system, comprising: The first acquisition module is used to obtain the surveying route of the UAV; A second acquisition module is used to collect data of the target city based on the surveying and mapping route, and obtain point cloud data and video images; A third acquisition module is used to acquire a feature label based on the video image; A fourth acquisition module, configured to acquire key annotation positions based on the video image and the feature labels; A first generating module, configured to generate an initial three-dimensional model of a target city based on the point cloud data; A judgment module, configured to judge whether there are key coordinates corresponding to the key marked positions in the initial three-dimensional model; a fifth acquisition module, configured to acquire the key coordinates if the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model; The second generating module is configured to generate a target three-dimensional model based on the feature labels, the key annotation positions, and the initial three-dimensional model.

[0015] By adopting the above technical solution, the drone mapping route is first obtained, and data of the target city is collected accordingly to obtain point cloud data and video images; then, feature labels are obtained from the video images, and the key annotation positions are determined by combining the video images and feature labels; then, an initial 3D model is generated based on the point cloud data, and it is determined whether there are key coordinates corresponding to the key annotation positions in the initial 3D model. If not, it means that the key coordinates corresponding to the key annotation positions are missing, and therefore it is necessary to further obtain the key coordinates; finally, a target 3D model is generated based on the feature labels, key annotation positions and the initial 3D model; the drone is used to flexibly collect data, and combined with feature analysis of the video images, it helps to more clearly know whether there are key coordinates missing. If so, the key coordinates are supplemented, which effectively solves the problem of missing key data in drone mapping, and can generate a more realistic and accurate target 3D model, greatly improving the accuracy of urban 3D mapping results and meeting the needs of high-precision urban 3D mapping.

[0016] In summary, this application has the following beneficial technical effects: First, the drone mapping route is obtained, and data is collected for the target city based on it to obtain point cloud data and video images; then, feature labels are obtained from the video images, and the key annotation positions are determined by combining the video images and feature labels; then, an initial 3D model is generated based on the point cloud data, and it is determined whether there are key coordinates corresponding to the key annotation positions in the initial 3D model. If not, it means that the key coordinates corresponding to the key annotation positions are missing, and therefore the key coordinates need to be further obtained; finally, a target 3D model is generated based on the feature labels, key annotation positions, and the initial 3D model; the drone is used to flexibly collect data, and combined with feature analysis of the video images, it helps to more clearly know whether there are key coordinates missing. If so, the key coordinates are supplemented, which effectively solves the problem of missing key data in drone mapping, and can generate a more realistic and accurate target 3D model, greatly improving the accuracy of urban 3D mapping results and meeting the needs of high-precision urban 3D mapping. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1This is a main flow chart of a method for urban three-dimensional mapping based on a drone in an embodiment of the present application; Figure 2 is a flowchart of steps S201 to S203; Figure 3 is a flowchart of steps S301 to S307; Figure 4 is a flowchart of steps S401 to S408; Figure 5 is a flowchart of steps S501 to S506; Figure 6 is a flowchart of steps S601 to S606; Figure 7 is a flowchart of steps S701 to S705; Figure 8 is a flowchart of steps S801 to S803; Figure 9 This is a module diagram of a drone-based urban three-dimensional mapping system in an embodiment of the present application.

[0018] Description of reference numerals: 1. First acquisition module; 2. Second acquisition module; 3. Third acquisition module; 4. Fourth acquisition module; 5. First generation module; 6. Judgment module; 7. Fifth acquisition module; 8. Second generation module. DETAILED DESCRIPTION

[0019] In a first aspect, the present application discloses a method for urban three-dimensional mapping based on drones.

[0020] Reference Figure 1 , a method for urban three-dimensional mapping based on a drone, comprising steps S101 to S108: Step S101: Obtain the surveying route of the UAV.

[0021] Specifically, in this embodiment, the surveying and mapping route is the flight path pre-planned when the UAV performs a surveying and mapping task, which usually includes parameters such as the starting point, end point, waypoint coordinates, flight altitude, and speed. In this embodiment, the surveying and mapping route planning needs to ensure as much as possible that when the UAV flies according to the predetermined trajectory, its data collection range can cover the target area, thereby ensuring the integrity and systematicness of the data collection.

[0022] Step S102: Collect data on the target city based on the surveying route, and obtain point cloud data and video images.

[0023] Specifically, in this embodiment, point cloud data is a set of three-dimensional spatial coordinate points obtained by devices such as lidar. Each point contains X, Y, Z coordinates and intensity information, which can reflect the spatial distribution of the object surface. It is the basic data for generating a three-dimensional model and is used to construct the geometric shape of the object; video images are continuous image frames taken by a camera mounted on a drone, which contain visual information such as texture, color, and shape of the target area, and are used to assist point cloud data in feature recognition and scene understanding, and provide texture mapping and detail supplementation for the three-dimensional model.

[0024] Step S103: Acquire feature labels based on the video image.

[0025] Specifically, in this embodiment, the feature label is a classification identifier of the target object in the video image, such as "building", "road", "vegetation", "water body", etc., which is used to distinguish different types of ground objects.

[0026] Step S104: Acquire key annotation positions based on the video image and feature labels.

[0027] Specifically, key annotation positions are spatial positions with special significance determined based on feature labels and video image analysis, such as the dividing lines between different categories of targets, characteristic points of target objects (such as building corners and road intersections), etc. Key annotation positions serve as key reference points for optimizing the three-dimensional model to ensure the accuracy of the positions and boundaries of important features in the model. In this embodiment, key annotation positions can be divided into coordinate points and dividing lines, etc., according to different objects and different situations.

[0028] Step S105: Generate an initial three-dimensional model of the target city based on the point cloud data.

[0029] Specifically, in this embodiment, the initial three-dimensional model is a three-dimensional model preliminarily generated based on point cloud data.

[0030] Step S106: Determine whether there are key coordinates corresponding to the key annotation positions in the initial three-dimensional model.

[0031] Specifically, in this embodiment, the key coordinates are the three-dimensional space coordinates (X, Y, Z) corresponding to the key annotation position, which are used to accurately locate the key annotation position in the three-dimensional model.

[0032] Step S107: If there are no key coordinates corresponding to the key annotation positions in the initial three-dimensional model, the key coordinates are obtained.

[0033] Specifically, in this embodiment, if there are no key coordinates corresponding to the key annotation position in the initial three-dimensional model, it means that the key coordinates corresponding to the key annotation position are missing. In order to improve the precision and accuracy of urban three-dimensional surveying and mapping, it is necessary to further obtain the key coordinates.

[0034] Step S108: Generate a target three-dimensional model based on the feature labels, key annotation positions and the initial three-dimensional model.

[0035] Specifically, in this embodiment, the key annotation position of the target object is obtained based on the feature label, and then the corresponding key coordinates are obtained based on the key annotation position. Then, based on the initial three-dimensional model, the key coordinates are fused to generate a three-dimensional model with higher accuracy and completeness. This three-dimensional model is the target three-dimensional model.

[0036] The drone-based urban three-dimensional mapping method provided in this embodiment first obtains the drone mapping route, and collects data on the target city based on it to obtain point cloud data and video images; then obtains feature tags from the video images, and then determines the key annotation positions in combination with the video images and feature tags; then generates an initial three-dimensional model based on the point cloud data, and determines whether there are key coordinates corresponding to the key annotation positions in the initial three-dimensional model. If not, it means that the key coordinates corresponding to the key annotation positions are missing, and therefore it is necessary to further obtain the key coordinates; finally, a target three-dimensional model is generated based on the feature tags, key annotation positions and the initial three-dimensional model; the drone is used to flexibly collect data, and combined with feature analysis of the video images, it helps to more clearly know whether there are key coordinates missing. If so, the key coordinates are supplemented, which effectively solves the problem of missing key data in drone mapping, can generate a more realistic and accurate target three-dimensional model, greatly improves the accuracy of urban three-dimensional mapping results, and meets the needs of high-precision urban three-dimensional mapping.

[0037] Reference Figure 2 In one implementation of this embodiment, step S104 obtains key annotation positions based on the video image and feature tags, including steps S201 to S203: Step S201: Based on the feature labels, the target categories of different target objects in the video image are obtained.

[0038] Specifically, in this embodiment, the target object is an entity in the video image with independent geometric features and semantic meaning, such as a building, natural terrain (mountain, river), artificial facility (road surface, pond), etc.

[0039] Step S202: Acquire target styles of different target objects based on the video image.

[0040] Specifically, in this embodiment, the target style is a classification based on the physical properties or functional attributes of the target object, such as "house" (residential building), "tower" (high-rise structure), "pavilion" (small landscape building), "mountain" (natural terrain), "river" (natural water body), "pond" (artificial / natural water body), "road surface" (transportation facility), etc.; the target style is the characteristics or type of the target object in appearance, structure or design, such as architectural style (modern simplicity, classical Chinese, European), terrain features (gentle hills, steep mountains), water form (winding river, regular pond), road surface material (asphalt, concrete, brick and stone), etc.

[0041] Step S203: Acquire key annotation positions based on the target category and target style.

[0042] The drone-based urban three-dimensional mapping method provided in this embodiment first identifies the target categories of different target objects in the video image based on feature labels and clarifies their properties, such as distinguishing different categories such as houses, towers, and pavilions. Then, the target styles of different target objects are obtained based on the video image, such as the appearance of the building and the undulating shape of the terrain. Finally, the target category and target style are combined to accurately locate the key annotation positions. Through the dual analysis of target category and target style, the information in the video image can be comprehensively and meticulously grasped, and the omission of key areas can be effectively avoided. It ensures that the truly important areas in the city that need to be highlighted are marked, so that the subsequently generated three-dimensional model can more realistically and accurately reflect the actual situation of the city, greatly improving the accuracy of the urban three-dimensional mapping results, and better meeting the needs of high-precision urban three-dimensional mapping in urban planning, disaster monitoring and other fields.

[0043] Reference Figure 3 In one implementation of this embodiment, step S203: obtaining key annotation positions based on target category and target style includes steps S301 to S307: Step S301: Based on the target category, obtain the special position corresponding to the target object.

[0044] Specifically, in this embodiment, the special position is a location point in the target object that has special geometric or functional significance, such as the corner of a building, the top of a tower, the center of a pavilion, the ridge line of a mountain, the turning point of a river, the boundary point of a pond, the intersection of a road, etc.

[0045] Step S302: Generate key annotation positions based on the special positions.

[0046] Step S303: Analyze the video image and determine whether different target categories and / or different target styles exist simultaneously in the video image.

[0047] Step S304: If different target categories exist in the video image at the same time, a first dividing line between different target categories is obtained.

[0048] Specifically, in this embodiment, if different target categories exist in the video image at the same time, it means that objects of different target categories exist in the frame of video image at the same time, such as buildings and streets, etc., so it is necessary to distinguish objects of different target categories; the first dividing line is the boundary line between different target categories in the video image, such as the boundary line between "buildings" and "road surface", the boundary line between "mountains" and "rivers", the boundary line between "ponds" and "vegetation", etc.

[0049] Step S305: Based on the first dividing line and preset surveying and mapping rules, key annotation positions are obtained.

[0050] Specifically, the preset surveying and mapping rules are pre-set principles or standards for guiding the determination of key annotation positions. In this embodiment, according to the preset surveying and mapping rules, the first dividing line is used as the key annotation position.

[0051] Step S306: If different target styles exist simultaneously in the video image, a second dividing line between the different target styles is obtained.

[0052] Specifically, in this embodiment, the second dividing line is the boundary line between the same target category but different target styles in the video image, such as the boundary line between "modern style buildings" and "classical style buildings", the boundary line between "asphalt pavement" and "brick and stone pavement", the boundary line between "natural river" and "artificial river channel", etc.

[0053] Step S307: Based on the second dividing line and the preset surveying and mapping rules, obtain the key annotation position.

[0054] Specifically, in this embodiment, according to the preset surveying and mapping rules, the second dividing line is used as the key marking position.

[0055] The drone-based urban three-dimensional mapping method provided in this embodiment, when obtaining key annotation positions based on target categories and target styles, first determines special positions corresponding to target objects based on target categories, such as building corners and road intersections, and generates key annotation positions based on these special positions; then analyzes the video image to determine whether different target categories and / or different target styles exist at the same time. If different target categories exist at the same time, a first dividing line between them is obtained, such as the boundary between a building and a road surface, and the key annotation positions are determined in combination with preset mapping rules; if different target styles exist at the same time, a second dividing line between them is obtained, such as the boundary between a modern building and a classical building, and the key annotation positions are also determined based on the preset mapping rules.

[0056] By conducting targeted analysis of the boundaries of different target categories and target styles, the spatial boundaries of various types of landforms in the urban environment can be identified more accurately; the determination of special locations and key annotation points provides precise coordinate references for boundary annotation, avoiding boundary deviations caused by data ambiguity; the application of preset surveying and mapping rules ensures the standardization and consistency of the annotation process, thereby significantly improving the accuracy and detail expression capabilities of key position annotations in the three-dimensional model, and ultimately improving the overall accuracy and reliability of urban three-dimensional mapping results, better meeting the needs of urban planning, architectural design and other fields for high-precision three-dimensional models.

[0057] Reference Figure 4 In one implementation of this embodiment, if different target styles exist simultaneously in the video image, step S306, obtaining a second dividing line between different target styles includes steps S401 to S408: Step S401: If different target styles exist in the video image at the same time, the video image is divided into several sub-areas based on the target styles.

[0058] Specifically, in this embodiment, after dividing the video image according to the target style, several local areas with similar visual features, i.e., sub-areas, are obtained. For example, a video image containing "modern architecture" and "classical architecture" is divided into "modern architecture area" and "classical architecture area".

[0059] Step S402: Obtain the contour boundaries of different sub-regions.

[0060] Specifically, in this embodiment, the contour boundary, i.e., the outer edge line of the sub-region, is usually manifested as a location where the pixel value (such as brightness, color) in the image changes suddenly, such as the boundary between the exterior wall of a building and the sky, the boundary between a river and its bank, etc.

[0061] Step S403: Determine whether the outline boundary is a clear boundary.

[0062] Specifically, in this embodiment, a clear boundary is a boundary where the pixel value changes significantly (mutated) at the outline boundary and is visually easy to distinguish, such as the junction of buildings of different materials (such as a glass curtain wall and a brick wall), the junction of a river and a hard road surface, etc., and the image features (such as brightness, color, and texture) on both sides of the boundary are obviously different.

[0063] Step S404: If the outline boundary is a clear boundary, a second dividing line is obtained based on the outline boundary.

[0064] Specifically, in this embodiment, if the outline boundary is a clear boundary, the outline boundary can be directly used as the second dividing line.

[0065] Step S405: If the outline boundary is a fuzzy boundary, the image brightness and image hue corresponding to different sub-regions are obtained.

[0066] Specifically, in this embodiment, a fuzzy boundary is a boundary where the pixel value changes slowly (gradually) at the outline boundary and is difficult to distinguish visually, such as the transition area between buildings of different styles (such as the mixed style area between modern and classical buildings), the smooth transition section between natural rivers and artificial waterways, etc. The image features on both sides of the boundary are not significantly different and there is a gradual process; image brightness is the brightness of the pixels in the image, usually expressed by grayscale values (0-255), where larger values indicate brighter (such as white) and smaller values indicate darker (such as black); image hue is the color feature of the image, usually expressed by HSV (hue, saturation, value).

[0067] Step S406: Obtain the difference position based on the image brightness and image hue.

[0068] Specifically, in this embodiment, the difference position is a set of position points where the image brightness or image hue changes in the fuzzy boundary area.

[0069] Step S407: Obtain the change type corresponding to the difference position.

[0070] Specifically, in this embodiment, the change type, that is, the change pattern of the image features (brightness, hue) at the difference position, is mainly divided into a transient type and a gradual type.

[0071] Step S408: Based on the change type, obtain a second dividing line.

[0072] The drone-based urban 3D mapping method provided in this embodiment, when different target styles exist simultaneously in a video image, first divides the video image into several sub-areas based on the target style, then obtains the contour boundary of each sub-area, and determines whether it is a clear boundary. If it is a clear boundary, the contour boundary is directly used as the second dividing line; if it is a fuzzy boundary, the image brightness and image tone of different sub-areas are further analyzed, and the difference positions with significant changes are determined by comparing the differences in image brightness and image tone. Then, the second dividing line is determined according to the change type (transient or gradual) of the difference position.

[0073] In order to address the possible ambiguity of boundaries between different target styles, quantitative analysis of image features such as image brightness and image tone is used to accurately identify and locate fuzzy boundaries, avoiding boundary labeling deviations caused by visual judgment errors; at the same time, the hierarchical processing mechanism of sub-area division and boundary type judgment helps to adapt to boundary characteristics in different scenarios, enhance the universality and flexibility of the technical solution, and thus help improve the expression accuracy of boundaries of different styles of land objects in urban three-dimensional surveying and mapping.

[0074] Reference Figure 5 In one implementation of this embodiment, step S408 obtains the second dividing line based on the change type, including steps S501 to S506: Step S501: If the change type is transient, the difference position is used as the second dividing line.

[0075] Specifically, in this embodiment, the transient type is a type of change in which the image features (image brightness, image hue) at the difference position undergo a significant mutation within a short distance, for example, a sudden change from a "modern building area (high brightness)" to a "classical building area (low brightness)" at the boundary. The image features at the difference position corresponding to this type have a large change amplitude and a short change distance.

[0076] Step S502: If the change type is a gradual change type, a first brightness value and a first hue are obtained based on the image brightness and image hue corresponding to different sub-regions.

[0077] Specifically, in this embodiment, the first brightness value is the average brightness value of the typical target style area in different sub-areas; the first hue is the average hue value of the typical target style area in different sub-areas, which is usually represented by the hue component in the HSV color space, such as the main hue of the exterior wall of the "modern building area" (such as blue, H=210°), or the main hue of the exterior wall of the "classical building area" (such as red, H=0°).

[0078] Step S503: obtaining a second brightness value and a second hue corresponding to the gradient area.

[0079] Specifically, the gradient area is a transition area between two sub-areas of different target styles, where image features (image brightness, image hue) show continuous changes, such as the mixed style area at the junction of "modern architecture" and "classical architecture", or the buffer zone between "natural river" and "artificial river"; the second brightness value is the average brightness value of different units in the gradient area. In this embodiment, the gradient area can be divided into several units, for example, evenly divided according to a specific number, or the number of divisions can be set according to the area of the gradient area; the second hue is the average hue value of different units in the gradient area.

[0080] Step S504: obtaining a brightness difference value based on the first brightness value and the second brightness value.

[0081] Specifically, in this embodiment, the brightness difference value is the absolute value of the difference between the first brightness value and the second brightness value.

[0082] Step S505: obtaining a hue difference based on the first hue and the second hue.

[0083] Specifically, in this embodiment, the hue difference is the absolute value of the difference between the first hue and the second hue.

[0084] Step S506: Obtain a second dividing line based on the brightness difference and the hue difference.

[0085] Specifically, in this embodiment, the comprehensive differences between different units and sub-regions are first calculated using the following formula: Wherein, L1 is the first brightness value; L(x,y) is the second brightness value; is the hue difference; in this embodiment, the hue difference is represented by a hue value, which is usually represented by an angle (ranging from 0° to 360°); H1 is the first hue, and H(x, y) is the second hue.

[0086] Then according to the spatial gradient vector of the difference Calculate the gradient modulus , when the gradient modulus is the maximum module length corresponding to all units, and When it exceeds a preset threshold (eg, 0.5), the unit is determined to be a second dividing line.

[0087] The drone-based urban three-dimensional mapping method provided in this embodiment directly uses the difference position as the second dividing line if the change type is transient, and uses the characteristic of significant short-range mutation of its image features to quickly locate it; if it is gradual, first obtain the first brightness value and first hue of the typical area of the target style in different sub-areas, then determine the second brightness value and second hue of the gradual area, then calculate the brightness difference between the first and second brightness, and the hue difference between the first and second hues, and finally, based on these differences, obtain the second dividing line by constructing a comprehensive difference model; the transient type uses feature mutation to quickly determine the boundary, and the gradual type uses multi-feature fusion and quantitative calculation to adapt to complex scenes such as architectural style transitions, normalization processing eliminates feature magnitude differences, and gradient analysis and threshold screening accurately locate the boundary, thereby improving the accuracy of boundary extraction.

[0088] Reference Figure 6 In one implementation of this embodiment, if there is no key coordinate corresponding to the key annotation position in the initial three-dimensional model in step S107, obtaining the key coordinate includes steps S601 to S606: Step S601: If there are no key coordinates corresponding to the key annotation positions in the initial three-dimensional model, a first attribute of the target object is obtained based on the video image.

[0089] Specifically, in this embodiment, the first attribute is a characteristic attribute of the target object, which is used to describe key information such as the nature, shape, and structure of the target object.

[0090] Step S602: Based on the first attribute, determine whether there is a symmetric reference coordinate.

[0091] Specifically, in this embodiment, the symmetry reference coordinates are coordinate points in three-dimensional space that can be used as symmetry references.

[0092] Step S603: If the symmetric reference coordinates exist, the symmetric mirror coordinates are obtained based on the symmetric reference coordinates, and the symmetric mirror coordinates are used as the missing key coordinates corresponding to the key annotation position.

[0093] Specifically, in this embodiment, coordinates obtained through mirror transformation based on symmetric reference coordinates are used to complete missing coordinates corresponding to key annotation positions.

[0094] Step S604: If the symmetric reference coordinates do not exist, obtain similar objects to the target object.

[0095] Specifically, in this embodiment, similar objects are other features that are similar to the target object in category, shape, structure or attributes.

[0096] Step S605: Based on similar objects, similar coordinates are obtained.

[0097] Specifically, in this embodiment, similar coordinates, namely the coordinates corresponding to similar objects in the three-dimensional model, can be used as a reference for calculating the key coordinates of the target object.

[0098] Step S606: Calculate key coordinates based on the similar coordinates.

[0099] The drone-based urban three-dimensional mapping method provided in this embodiment, when it is determined that there are no key coordinates corresponding to the key annotation position in the initial three-dimensional model, first obtains the first attribute of the target object based on the video image, and determines whether there are symmetric reference coordinates based on the first attribute; if so, obtains symmetric mirror coordinates based on the symmetric reference coordinates, and uses them as the missing key coordinates; if not, obtains similar objects of the target object, obtains similar coordinates based on the similar objects, and then calculates the key coordinates; to address the problem of missing key coordinates in the initial three-dimensional model, this method determines the symmetric reference coordinates and uses the symmetric mirror coordinates to supplement them, or when there are no symmetric reference coordinates, uses the similar coordinates of similar objects to calculate the key coordinates, which can effectively solve the problem of missing key coordinates caused by incomplete data collection, improves the integrity and accuracy of the three-dimensional model, and thus improves the accuracy of the drone-based urban three-dimensional mapping results, so that the generated three-dimensional model more realistically and accurately reflects the actual situation of the city, and meets the needs of high-precision urban three-dimensional mapping.

[0100] Reference Figure 7In one implementation of this embodiment, if the symmetric reference coordinates do not exist in step S604, obtaining similar objects of the target object includes steps S701 to S705: Step S701: If the symmetrical reference coordinates do not exist, the coordinates of the highest point of the target object are obtained based on the video image and the initial three-dimensional model.

[0101] Specifically, in this embodiment, the highest point coordinates are the vertex coordinates of the target object in the altitude or height direction in the three-dimensional space.

[0102] Step S702: Based on the highest point coordinates, obtain the selectable object.

[0103] Specifically, in this embodiment, the selectable object is a similar object whose height is lower than the highest point coordinate of the target object.

[0104] Step S703: Obtain the second attribute of the optional object.

[0105] Specifically, in this embodiment, the second attribute is a characteristic attribute of the selectable object, corresponds to the first attribute of the target object, and is used to measure the similarity between the two.

[0106] Step S704: If the first attribute and the second attribute meet the preset similarity standard, then obtain an optional annotation position based on the key annotation position.

[0107] Specifically, in this embodiment, the preset similarity standard is a pre-set rule or threshold for determining whether two objects are similar, which is usually set based on the similarity of attribute parameters; the optional annotation position is based on the key annotation position, the corresponding annotation position on the optional object.

[0108] Step S705: Determine whether there are optional coordinates corresponding to the optional annotation position.

[0109] Specifically, in this embodiment, the optional coordinates are coordinates corresponding to the optional marked positions.

[0110] Step S706: If there are optional coordinates, the optional objects are treated as similar objects.

[0111] The drone-based urban three-dimensional mapping method provided in this embodiment first obtains the coordinates of the highest point of the target object based on the video image and the initial three-dimensional model when the target object does not have symmetrical reference coordinates, and uses this as a basis to screen out optional objects that are spatially adjacent and have heights higher than the highest coordinates; then, by comparing the second attribute of the optional object with the first attribute of the target object, the optional objects that meet the preset similarity criteria are determined as similar objects; then, it is determined whether there are optional coordinates at the corresponding optional annotation position on the similar object, and if so, the key coordinates of the target object are calculated using the optional coordinates.

[0112] By screening the highest point coordinates and comparing attribute similarities, the problem of missing key coordinates due to occlusion or data transmission anomalies is effectively solved. The missing coordinates can be supplemented with existing similar object information without re-collecting data, significantly improving the efficiency of urban three-dimensional surveying and mapping, the integrity of the model, and the accuracy of the surveying and mapping results.

[0113] Reference Figure 8 In one implementation of this embodiment, step S605 obtains similar coordinates based on similar objects, including steps S801 to S803: Step S801: Obtain reference coordinates and scale.

[0114] Specifically, in this embodiment, the reference coordinates are coordinate points in three-dimensional space that serve as a reference origin or a known accurate position, which are used to establish a coordinate system or measure the relative positions of other coordinates; the scale is the proportional relationship between the video image and the real urban space.

[0115] Step S802: Calculate a first distance and a second distance based on the reference coordinates, the similar coordinates, and the key annotation positions.

[0116] Specifically, in this embodiment, the first distance is the X-axis distance between the key annotation position in the video image and the similar coordinates; the second distance is the Y-axis distance between the key annotation position in the video image and the similar coordinates.

[0117] Step S803: Calculate key coordinates based on the first distance, the second distance, and the scale.

[0118] The drone-based urban three-dimensional mapping method provided in this embodiment, when the key coordinates of the target object are missing and similar coordinates are obtained through similar objects, first determines the base coordinates as the reference origin and combines them with the known scale to calculate the first distance from the key annotation position to the similar coordinates and the second distance from the key annotation position to the similar coordinates, respectively. Then, using the proportional relationship between these two distances and the scale, the relative distance in the video image is converted into actual geographic coordinates, thereby inferring the missing key coordinates; by introducing the base coordinates and the scale to establish a proportional relationship, the known coordinate information of similar objects is cleverly used to complete the missing coordinates of the target object, without relying on additional complex measuring equipment or re-collecting data, thereby significantly improving the efficiency and accuracy of three-dimensional model construction.

[0119] Secondly, this application also discloses a three-dimensional urban mapping system based on drones.

[0120] Reference Figure 9 , a UAV-based urban 3D mapping system, including: The first acquisition module is used to obtain the surveying route of the UAV; The second acquisition module is used to collect data on the target city based on the surveying and mapping route, and obtain point cloud data and video images; A third acquisition module is used to acquire feature labels based on the video image; A fourth acquisition module is used to obtain key annotation positions based on the video image and feature labels; The first generation module is used to generate an initial three-dimensional model of the target city based on the point cloud data; A judgment module is used to judge whether there are key coordinates corresponding to the key annotation positions in the initial three-dimensional model; a fifth acquisition module, configured to acquire the key coordinates if the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model; The second generation module is used to generate a target three-dimensional model based on the feature labels, key annotation positions and the initial three-dimensional model.

[0121] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for urban three-dimensional mapping based on drones, characterized in that: include: Get the UAV's mapping route; Collect data on the target city based on the surveying route and obtain point cloud data and video images; Based on the video image, obtaining a feature label; Based on the video image and the feature labels, obtaining key annotation positions; generating an initial three-dimensional model of a target city based on the point cloud data; Determining whether there are key coordinates corresponding to the key annotation positions in the initial three-dimensional model; If the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model, acquiring the key coordinates; A target three-dimensional model is generated based on the feature labels, the key annotation positions, and the initial three-dimensional model.

2. The method for urban three-dimensional mapping based on drone according to claim 1, characterized in that: The acquiring of key annotation positions based on the video image and the feature labels includes: Based on the feature labels, obtaining target categories of different target objects in the video image; Based on the video image, obtaining target styles of different target objects; Based on the object category and the object style, key annotation positions are obtained.

3. The method for urban three-dimensional mapping based on drone according to claim 2, characterized in that: The acquiring of key annotation positions based on the target category and the target style includes: Based on the target category, obtaining a special position corresponding to the target object; Based on the special position, generating a key annotation position; Analyzing the video image and determining whether different target categories and / or different target styles exist simultaneously in the video image; If different target categories exist simultaneously in the video image, obtaining a first dividing line between the different target categories; Based on the first dividing line and preset surveying and mapping rules, obtaining key annotation positions; If different target styles exist simultaneously in the video image, obtaining a second dividing line between the different target styles; Based on the second dividing line and the preset surveying and mapping rules, key annotation positions are obtained.

4. The method for urban three-dimensional mapping based on drone according to claim 3, characterized in that: If different target styles exist simultaneously in the video image, obtaining a second dividing line between the different target styles includes: If different target styles exist simultaneously in the video image, dividing the video image into a plurality of sub-areas based on the target styles; Get the contour boundaries of different sub-regions; Determining whether the outline boundary is a clear boundary; If the outline boundary is the clear boundary, obtaining a second dividing line based on the outline boundary; If the outline boundary is a fuzzy boundary, obtaining image brightness and image hue corresponding to different sub-regions; Obtaining a difference position based on the image brightness and the image hue; Obtaining the change type corresponding to the difference position; Based on the change type, a second dividing line is obtained.

5. The method for urban three-dimensional mapping based on drone according to claim 4, characterized in that: The acquiring of a second dividing line based on the change type includes: If the change type is transient, the difference position is used as the second dividing line; If the change type is a gradual change type, obtaining a first brightness value and a first hue based on the image brightness and the image hue corresponding to different sub-regions; Get the second brightness value and second hue corresponding to the gradient area; Obtaining a brightness difference value based on the first brightness value and the second brightness value; Obtaining a hue difference based on the first hue and the second hue; A second dividing line is acquired based on the brightness difference value and the hue difference value.

6. The method for urban three-dimensional mapping based on drone according to claim 1, characterized in that: If the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model, obtaining the key coordinates includes: If the key coordinates corresponding to the key annotation position do not exist in the initial three-dimensional model, obtaining a first attribute of the target object based on the video image; Based on the first attribute, determining whether a symmetric reference coordinate exists; If the symmetric reference coordinates exist, obtaining symmetric mirror coordinates based on the symmetric reference coordinates, and using the symmetric mirror coordinates as the missing key coordinates corresponding to the key marked position; If the symmetric reference coordinates do not exist, obtaining a similar object to the target object; Based on the similar objects, obtaining similar coordinates; Based on the similar coordinates, the key coordinates are calculated.

7. The method for urban three-dimensional mapping based on drone according to claim 6, characterized in that: If the symmetric reference coordinates do not exist, obtaining a similar object to the target object includes: If the symmetric reference coordinates do not exist, obtaining the coordinates of the highest point of the target object based on the video image and the initial three-dimensional model; Based on the highest point coordinates, obtaining a selectable object; Get the second attribute of the optional object; If the first attribute and the second attribute meet a preset similarity standard, obtaining an optional annotation position based on the key annotation position; Determine whether there are optional coordinates corresponding to the optional annotation position; If the optional coordinates exist, the optional object is used as the similar object.

8. The method for urban three-dimensional mapping based on drone according to claim 6, characterized in that: The calculating the key coordinates based on the similar coordinates includes: Get the reference coordinates and scale; Calculating a first distance and a second distance based on the reference coordinates, the similar coordinates, and the key annotation position; The key coordinates are calculated based on the first distance, the second distance, and the scale.

9. A three-dimensional urban mapping system based on drones, characterized in that: include: The first acquisition module is used to obtain the surveying route of the UAV; A second acquisition module is used to collect data of the target city based on the surveying and mapping route, and obtain point cloud data and video images; A third acquisition module is used to acquire a feature label based on the video image; A fourth acquisition module, configured to acquire key annotation positions based on the video image and the feature labels; A first generating module, configured to generate an initial three-dimensional model of a target city based on the point cloud data; A judgment module, configured to judge whether there are key coordinates corresponding to the key marked positions in the initial three-dimensional model; a fifth acquisition module, configured to acquire the key coordinates if the key coordinates corresponding to the key annotation positions do not exist in the initial three-dimensional model; The second generating module is configured to generate a target three-dimensional model based on the feature labels, the key annotation positions, and the initial three-dimensional model.

Images