Automatic engineering investigation method and system based on unmanned aerial vehicle

Abstract

The invention provides an automatic engineering investigation method and system based on an unmanned aerial vehicle. The method comprises the following steps: shooting first image data of a to-be-surveyed area, and identifying an engineering area and a non-engineering area corresponding to an engineering drawing from the first image data; extracting region features of each region adjacent to the engineering region based on the image part of the non-engineering region, and determining an adjacent region corresponding to the engineering region based on the region features and the engineering type; and determining an investigation track of the unmanned aerial vehicle based on the outer boundary line of the adjacent area, and controlling the unmanned aerial vehicle to perform engineering automatic investigation according to the investigation track. According to the automatic engineering investigation method based on the unmanned aerial vehicle, the engineering area and the adjacent area can be automatically identified and the investigation track can be planned by means of image identification, feature extraction and path planning technologies.

Description

Technical Field

[0001] This invention relates to the field of engineering surveying technology, and more specifically, to an automated engineering surveying method and system based on unmanned aerial vehicles (UAVs). Background Technology

[0002] While drone technology has been applied in the current engineering survey field, it still has significant shortcomings in determining and implementing survey areas. Existing engineering survey methods rely entirely on the operator's subjective judgment based on experience. When controlling the drone, the operator must manually adjust the flight path, altitude, and angle to cover the target area. This method severely limits survey efficiency. When facing large areas with complex terrain, operators need to frequently plan routes manually and repeatedly confirm the survey area, which not only consumes a lot of time and energy but also makes it difficult to guarantee the comprehensiveness and accuracy of the survey. Furthermore, manual operation is prone to errors, leading to repeated surveys or omissions of certain areas, further reducing efficiency and failing to meet the urgent demand for high-efficiency surveys in modern engineering construction.

[0003] Therefore, there is an urgent need for a method that can automatically determine the scope of the survey and realize automated surveying in order to improve the overall efficiency of engineering surveying. Summary of the Invention

[0004] In order to at least solve the technical problems existing in the background art, the present invention provides an engineering automatic survey method, system, electronic device, computer storage medium and computer program product based on unmanned aerial vehicles (UAVs).

[0005] The first aspect of the present invention provides an automated engineering survey method based on unmanned aerial vehicles (UAVs), comprising the following steps: First image data of the area to be surveyed is captured, and engineering areas and non-engineering areas corresponding to engineering drawings are identified from the first image data; Based on the image portion of the non-engineering area, the regional features of each region adjacent to the engineering area are extracted, and based on the regional features and the engineering type, the adjacent regions corresponding to the engineering area are determined. The UAV's survey trajectory is determined based on the outer boundary line of the adjacent area, and the UAV is controlled to conduct automatic engineering surveys according to the survey trajectory.

[0006] A second aspect of the present invention provides an automated engineering survey system based on unmanned aerial vehicles (UAVs), including a processor and a memory. The processor calls and executes instructions stored in the memory to achieve the following steps: First image data of the area to be surveyed is captured, and engineering areas and non-engineering areas corresponding to engineering drawings are identified from the first image data; Based on the image portion of the non-engineering area, the regional features of each region adjacent to the engineering area are extracted, and based on the regional features and the engineering type, the adjacent regions corresponding to the engineering area are determined. The UAV's survey trajectory is determined based on the outer boundary line of the adjacent area, and the UAV is controlled to conduct automatic engineering surveys according to the survey trajectory.

[0007] A third aspect of the present invention provides an electronic device comprising: a memory storing executable program code; a processor coupled to the memory; the processor invoking the executable program code stored in the memory to perform the method as described in any of the preceding claims.

[0008] A fourth aspect of the present invention provides a computer storage medium storing a computer program that, when executed by a processor, performs the method described in any of the preceding claims.

[0009] A fifth aspect of the invention provides a computer program product comprising a computer program executable by a processor to implement the method as described in any of the preceding claims.

[0010] The UAV-based automated engineering survey method of this invention can automatically identify the engineering area and adjacent areas and plan the survey trajectory by using image recognition, feature extraction, and path planning technologies. Compared with traditional manual surveying, it effectively avoids human error and makes the survey scope more accurate and comprehensive. Moreover, UAVs can automatically perform survey tasks, greatly improving survey efficiency and significantly reducing survey costs and time investment. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating an automated engineering survey method based on unmanned aerial vehicles (UAVs) disclosed in an embodiment of the present invention.

[0012] Figure 2 This is a schematic diagram of the structure of an unmanned aerial vehicle (UAV)-based automated engineering survey system disclosed in an embodiment of the present invention. Detailed Implementation

[0013] The following specific embodiments illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] like Figure 1 As shown in the figure, an automated engineering survey method based on unmanned aerial vehicles (UAVs) according to an embodiment of the present invention includes the following steps: S100: Capture first image data of the area to be surveyed, and identify the engineering area and non-engineering area corresponding to the engineering drawings from the first image data.

[0015] Using image acquisition equipment such as high-definition cameras and multispectral cameras mounted on drones, images are taken of the target area to be surveyed, obtaining first image data containing rich information such as topography, buildings, and vegetation. The design information in engineering drawings (such as building outlines and road plans) is converted into digital recognition features and compared and analyzed with the first image data. For example, edge detection algorithms are used to identify lines in the image and match them with lines in the engineering drawings. Combined with deep learning models (such as convolutional neural networks, CNNs), objects in the image are classified, thereby accurately distinguishing between engineering areas that belong to the engineering construction scope and non-engineering areas outside the engineering areas.

[0016] S200, based on the image portion of the non-engineering area, extract the regional features of each region adjacent to the engineering area, and determine the adjacent region corresponding to the engineering area based on the regional features and the engineering type.

[0017] For the non-engineering areas (i.e., the outer perimeter of the engineering area) adjacent to the engineering area in the first image data, feature extraction algorithms are used for analysis to obtain regional features. These regional features include, but are not limited to, topographic relief features (slope and aspect analyzed using digital elevation models), land cover type features (such as vegetation cover and water distribution), and geological features (if the image contains relevant clues). For example, using remote sensing image processing technology, vegetation indices of adjacent areas are extracted to determine the degree of vegetation cover; elevation changes of adjacent areas are extracted using topographic data to analyze topographic complexity. Different types of engineering projects (such as road engineering, building engineering, and water conservancy projects) have different requirements and impacts on adjacent areas. Based on the extracted regional characteristics and the features of each engineering type, corresponding judgment rules or models are established. Taking road engineering as an example, to ensure that road construction is not affected by geological disasters such as landslides and debris flows in the surrounding terrain, areas with unstable terrain features (such as steep slopes) within a certain range around the project area are identified as adjacent areas. For building engineering, areas within a certain distance that may affect construction safety or the later use of the building (such as areas near high-voltage power line corridors or underground pipeline distribution areas) may be identified as adjacent areas based on the scale and design of the building.

[0018] S300, determine the UAV's survey trajectory based on the outer boundary line of the adjacent area, and control the UAV to perform automatic engineering survey according to the survey trajectory.

[0019] After obtaining the outer boundary lines of adjacent areas, a path planning algorithm is used, combined with the UAV's performance parameters (such as flight speed, endurance, turning radius, etc.) and survey accuracy requirements, to generate the optimal survey trajectory. For example, the adjacent area can be divided into multiple grid cells using a grid method, and the survey trajectory covering all areas and having the shortest path can be searched in the grid map based on the Dijkstra algorithm or A* algorithm; alternatively, intelligent optimization algorithms such as genetic algorithms can be used to comprehensively consider factors such as terrain and obstacles to generate a safe and efficient flight path. By receiving and executing the generated survey trajectory commands through the drone's flight control system, the drone can automatically fly along the planned path. During flight, it collects images or data of the survey area according to preset shooting parameters (such as shooting interval, shooting angle, and shooting altitude), eliminating the need for frequent manual adjustments. The entire survey process is automated, which not only significantly improves survey efficiency and reduces the risk of human error, but also ensures the integrity and consistency of the survey data, providing reliable data support for subsequent engineering analysis and design.

[0020] Further, the step of identifying the engineering area and non-engineering area corresponding to the engineering drawing from the first image data includes: The engineering drawings are converted into electronic drawings with the same resolution as the first image data according to the scale, and the electronic drawings are then subjected to noise reduction processing. The first image data is semantically segmented using a deep learning model to obtain the category information of each pixel in the first image data. By spatially matching the outline of the engineering design area in the electronic drawing with the semantically segmented pixel area in the first image data, the engineering area and non-engineering area are identified.

[0021] First, the engineering drawings (e.g., images of engineering drawings pre-stored in the UAV's memory) are converted to electronic drawings with the same resolution as the first image data, ensuring consistency in spatial scale and detail between the drawings and the image data. For example, if the resolution of the first image data is 0.5 meters per pixel on the actual ground, the engineering drawings also need to be converted to the same corresponding scale to ensure that matching errors do not occur due to scale differences during subsequent comparisons.

[0022] Meanwhile, since noise may be introduced during the conversion or storage of drawings (such as stray lines and data errors generated by digitization), noise reduction algorithms such as median filtering and Gaussian filtering are used to remove these interference factors, making the outline of the design area in the drawing clearer and more accurate. This provides reliable basic data for subsequent area matching and avoids the recognition accuracy of engineering areas and non-engineering areas being affected by problems with the drawing itself.

[0023] Next, the powerful feature extraction capabilities of deep learning models are used to perform semantic segmentation on the first image data. Deep learning models (such as U-Net and DeepLab semantic segmentation models) are trained on large amounts of labeled image data, enabling them to learn pixel feature patterns of different features (such as buildings, roads, and vegetation) in images. After inputting the first image data into the model, it analyzes each pixel and determines its category based on the learned feature patterns. For example, pixels in building areas are labeled "buildings," and pixels in road areas are labeled "roads," thus achieving semantic segmentation of the first image data and obtaining the category information of each pixel. This processing method can meticulously divide various regions in the image, and compared to traditional image recognition methods, it can more accurately capture the boundaries and features of different features in complex scenes.

[0024] After obtaining the outline of the engineering design area in the electronic drawing and the category information of each pixel after semantic segmentation of the first image data, spatial location matching algorithms (such as feature-based matching algorithms SIFT and SURF) are used to find the positional relationship between the outline of the engineering design area in the drawing and the corresponding category of pixel regions in the image data. For example, if there is a road design outline in the drawing, the matching algorithm finds the corresponding position in the semantically segmented "road" pixel region in the image data, and these matched regions in the image are identified as engineering areas; other regions that do not match the outline of the engineering design area in the drawing are determined as non-engineering areas. It can be understood that non-engineering areas can simply be a certain width of the area adjacent to the engineering area.

[0025] The above-mentioned spatial location matching method can accurately map the design intent in engineering drawings to the actual image scene, and achieve accurate identification of engineering areas and non-engineering areas.

[0026] Further, the extraction of regional features of each region adjacent to the engineered region based on the image portion of the non-engineered region includes: Multi-scale pyramid decomposition is performed on the image portion of the non-engineering area to obtain hierarchical images of different resolutions; For images at each level, texture features are extracted using the local binary mode algorithm, and principal component analysis is used to reduce the dimensionality of the extracted texture features. Topographic relief features are obtained by overlaying digital elevation model data with images at each level. Based on the spectral feature extraction algorithm, the spectral reflectance data of land features are extracted from the image portion to obtain land feature type features. The texture features, terrain undulation features, and land cover type features after dimensionality reduction are fused to obtain the regional features of each region adjacent to the engineering area.

[0027] In current engineering surveys, determining the adjacent boundaries of a project area often relies on manual experience or simple distance delineation, lacking a scientific and comprehensive analysis of the characteristics of adjacent areas. This leads to inaccurate delineation of the survey scope, failing to effectively meet the needs of engineering design and construction. This embodiment extracts the characteristics of adjacent areas through a multi-technology fusion approach to support accurate subsequent determination of adjacent areas.

[0028] First, non-engineering area images contain rich details. High-resolution images can capture subtle features, but they require large amounts of data and are computationally complex. Low-resolution images, while having less data, lose details. By using multi-scale pyramid decomposition, images are constructed into a multi-level structure from low to high resolution, with each level preserving information at different scales. During subsequent feature extraction, an appropriate level can be selected based on requirements, ensuring comprehensive feature extraction while reducing computational load and improving processing efficiency. For example, low-resolution levels can quickly identify large-scale terrain trends, while high-resolution levels can accurately analyze local terrain details. Then, for each level of image, the following feature extraction is performed: (1) The Local Binary Pattern (LBP) algorithm is specifically designed for extracting image texture information. By comparing the grayscale relationship between the center pixel and its neighboring pixels, it transforms the image texture into quantifiable numerical features, effectively describing the texture differences of ground features in the image, such as the density of vegetation and the roughness of rock surfaces. However, the extracted texture features have high dimensionality and large data redundancy. To address this, this invention further employs Principal Component Analysis (PCA) to reduce the dimensionality of the extracted texture features. That is, by using linear transformation, the high-dimensional texture features are mapped to a low-dimensional space, removing redundant information and retaining the most representative feature components. The dimensionality-reduced texture features not only reduce data storage and computation costs but also highlight key texture features, improving feature recognition and usability. (2) Digital Elevation Model (DEM) records the elevation information of the regional terrain. By overlaying DEM data with images of different levels of non-engineering areas and using the spatial analysis function of Geographic Information System (GIS), the terrain elevation information can be intuitively correlated with the features in the image. For example, by marking the range of areas with different elevations in the image, terrain parameters such as slope and aspect can be calculated to obtain the terrain undulation characteristics.

[0029] The aforementioned topographic relief features can be used to determine the topographic stability and construction difficulty of the adjacent area of ​​the project. For example, high-slope areas may have landslide risks and should be included in the key investigation of the adjacent area to provide key basis for engineering design and construction safety assessment. (3) The multispectral camera carried by the UAV can acquire image data of multiple spectral bands. Since different land features have different reflectance in different spectral bands, the spectral reflectance curves of each land feature in different bands can be extracted by analyzing these data through spectral feature extraction algorithms. By comparing these spectral reflectance curves with the spectral library of known land features, the land feature type can be accurately identified, such as distinguishing water bodies, vegetation, buildings, etc.

[0030] The aforementioned land feature characteristics clearly define the environmental composition of the adjacent area, which can be used to assess the impact of engineering construction on the surrounding environment and plan construction schemes. For example, in engineering areas near water bodies, it is necessary to focus on investigating hydrogeological conditions to avoid pollution of water bodies during construction. Next, since a single feature cannot fully reflect the true situation of adjacent areas, this invention performs feature fusion on the above-mentioned feature information to integrate multiple aspects of information. Specifically, the dimensionality-reduced texture features, terrain relief features, and land cover type features are fused using methods such as weighted fusion and decision-level fusion to integrate different types of feature data into a comprehensive feature vector or dataset.

[0031] The integrated regional features include multi-dimensional information such as landform texture, topography, and land feature type, which more comprehensively and accurately describes the characteristics of adjacent areas. This makes the subsequent delineation of adjacent areas more scientific and reasonable, ensuring the integrity and effectiveness of engineering surveys. Further, determining the adjacent regions corresponding to the project area based on the regional characteristics and project type includes: The regional features are converted into feature vectors, and a type vector containing key attributes of the project type is constructed. Using a multi-head attention mechanism, the attention weight between the feature vector and the type vector is calculated to highlight the regional features that are highly correlated with the current project type; the regional features are then weighted and fused based on the attention weight to obtain a fused feature vector. The fused feature vector is matched with feature templates at different preset distance thresholds for similarity. When the similarity exceeds the judgment threshold, the corresponding region is determined to be the adjacent region corresponding to the engineering region.

[0032] This embodiment vectorizes regional features and project types, and utilizes a multi-head attention mechanism to calculate the correlation weights between feature vectors and type vectors from multiple dimensions. This accurately highlights regional features highly correlated with project types, overcoming the limitations of manual judgment in systematically analyzing complex relationships. Furthermore, weighted fusion further strengthens key features, making regional features more aligned with project requirements. Finally, it uses preset feature templates and similarity matching to determine adjacent regions, replacing subjective judgment with quantitative standards to ensure the consistency and accuracy of the determination results. Specifically: First, the abstract information of regional features and project types is transformed into quantifiable and computable vector data. For regional features such as texture features, topographic relief features, and land cover type features, these are arranged and combined in a specific order to form a multidimensional array. For example, the dimensionality reduction results of texture features, topographic slope and aspect values, and land cover spectral reflectance data are sequentially concatenated to form a regional feature vector. Similarly, key attributes of project types, including the length, width, and orientation of road projects, and the building type, number of floors, and foundation type of building projects, are also transformed into numerical forms to construct a type vector.

[0033] Then, the region feature vector and type vector are input into a multi-head attention mechanism, with each "head" calculating the similarity of each element in the two vectors. For example, using a dot product method, each element of the region feature vector is multiplied by the corresponding element of the type vector, the sums are obtained, and then the result is scaled by the square root of the vector dimension to obtain a similarity score. These scores are converted into probability values, i.e., attention weights, using a softmax function. These attention weights reflect the degree of correlation between each element in the region feature vector and the key attributes of the project type. The higher the weight, the greater the importance of the region feature to the current project type. The attention weights calculated by multiple "heads" are concatenated or weighted averaged to ultimately highlight the region features with a high correlation to the current project type, thereby achieving the filtering and strengthening of region features to focus on key information.

[0034] Based on the calculated attention weights, the regional feature vectors are weighted. Each element of the regional feature vector is multiplied by its corresponding attention weight, and then the weighted elements are summed to obtain a new vector, the fused feature vector. This process allows regional features highly correlated with the project type to occupy a larger proportion in the fused feature vector, while features with low correlation are weakened. For example, for road engineering, terrain slope features have a high correlation. After weighted fusion, slope-related feature information is enhanced in the fused feature vector, thus making the fused feature vector more reflective of regional features closely related to the current project type, providing more targeted data for adjacent area determination.

[0035] Next, multiple feature templates with different distance thresholds are pre-set according to different project types and common project scenarios. These feature templates contain typical feature vectors that adjacent regions should possess within the corresponding distance range. Cosine similarity, Euclidean distance, and other metrics are used to calculate the similarity between the fused feature vector and each feature template. For example, cosine similarity measures similarity by calculating the cosine of the angle between two vectors; the closer the value is to 1, the higher the similarity.

[0036] The calculated similarity is compared with a pre-set threshold. When the similarity exceeds the threshold, it indicates that the characteristics of the region match the typical adjacent region characteristics of the current project type at the corresponding distance, thus identifying the region as the adjacent region corresponding to the project area. This method achieves automated and accurate determination of adjacent regions, avoiding the subjectivity and uncertainty of human experience-based judgment.

[0037] Furthermore, determining the UAV's reconnaissance trajectory based on the outer boundary line of the adjacent region includes: The outer boundary line of the adjacent area is imported into the geographic information system for vectorization processing to generate three-dimensional spatial boundary data with elevation information. A genetic algorithm is used to perform path planning on the three-dimensional spatial boundary data, with the objective function of covering all adjacent areas and minimizing the flight path. This is combined with the performance parameters of the UAV and the limiting factors in the constraint database to generate the UAV's preliminary reconnaissance trajectory. The generated preliminary exploration trajectory is smoothed to obtain the exploration trajectory.

[0038] In GIS, vectorization operations transform boundary information, which may originally be in image format or unstructured, into vector data composed of points, lines, and polygons. This vector data has explicit coordinate information and can accurately represent the shape and location of the boundary.

[0039] Simultaneously, the GIS system, combined with digital elevation model (DEM) data, assigns elevation information to the vectorized boundary data, thereby generating three-dimensional spatial boundary data containing X and Y plane coordinates and Z elevation coordinates. For example, in engineering surveys in mountainous areas, the generated three-dimensional spatial boundary data can accurately reflect the topographic relief of adjacent areas, avoiding the planning of infeasible flight paths such as those traversing mountains. The objective function is to cover all adjacent areas and minimize the flight path, aiming to ensure that the UAV completes the reconnaissance task efficiently, reducing flight time and energy consumption. Simultaneously, performance parameters such as the UAV's maximum flight speed, range, and minimum turning radius are considered to ensure that the planned path is within the UAV's actual capabilities. For example, if the UAV's range is limited, the algorithm will avoid planning long-distance paths exceeding its range. Furthermore, the constraint database contains limiting factors such as obstacle locations (e.g., buildings, high-voltage power lines) and no-fly zone information. These constraints are taken into account during path planning to prevent the planned path from crossing obstacles or no-fly zones. For example, when encountering buildings, the algorithm will automatically adjust the path to bypass them, ensuring the UAV's flight safety.

[0040] Through continuous iteration and optimization, under the premise of meeting various conditions, path planning is performed on the three-dimensional spatial boundary data, and finally the preliminary reconnaissance trajectory of the UAV is generated. The aforementioned preliminary survey trajectory is a relatively optimal solution after comprehensively considering various factors, but it may have problems such as excessively frequent path turns. For example, to meet the requirements of coverage area and obstacle avoidance, there may be many inflection points and sharp turns. Such a trajectory is not conducive to the stable flight of the UAV and may also increase the difficulty of flight control and energy consumption. This embodiment further optimizes the inflection points in the preliminary trajectory through smoothing processing, such as Bézier curve fitting and spline curve interpolation. Taking Bézier curve fitting as an example, by selecting key nodes on the trajectory, Bézier curves are constructed, and smooth curves are used to connect these nodes, replacing the original abrupt broken lines and making the trajectory smoother.

[0041] After smoothing, redundant inflection points are removed, reducing the complexity of the trajectory. The generated survey trajectory not only meets the flight control precision requirements of the UAV, but also enables the UAV to remain stable during flight, reducing the number of attitude adjustments, reducing energy consumption, improving survey efficiency and data acquisition quality, and ensuring that the UAV can safely and efficiently complete engineering survey tasks.

[0042] It is understandable that some areas among the multiple adjacent areas of the project area may not belong to the aforementioned adjacent areas. In this case, the outline of the project area in that area can be used as the outer boundary line.

[0043] like Figure 2 As shown in the figure, an embodiment of the present invention provides an automated engineering survey system based on unmanned aerial vehicles (UAVs), comprising a processor 101 and a memory 102. The processor 101 calls and executes instructions stored in the memory 102 to achieve the following steps: First image data of the area to be surveyed is captured, and engineering areas and non-engineering areas corresponding to engineering drawings are identified from the first image data; Based on the image portion of the non-engineering area, the regional features of each region adjacent to the engineering area are extracted, and based on the regional features and the engineering type, the adjacent regions corresponding to the engineering area are determined. The UAV's survey trajectory is determined based on the outer boundary line of the adjacent area, and the UAV is controlled to conduct automatic engineering surveys according to the survey trajectory.

[0044] Further, the engineering area and non-engineering area corresponding to the engineering drawing are identified from the first image data, including: The engineering drawings are converted into electronic drawings with the same resolution as the first image data according to the scale, and the electronic drawings are then subjected to noise reduction processing. The first image data is semantically segmented using a deep learning model to obtain the category information of each pixel in the first image data. By spatially matching the outline of the engineering design area in the electronic drawing with the semantically segmented pixel area in the first image data, the engineering area and non-engineering area are identified.

[0045] Furthermore, based on the image portion of the non-engineered area, regional features of each region adjacent to the engineered area are extracted, including: Multi-scale pyramid decomposition is performed on the image portion of the non-engineering area to obtain hierarchical images of different resolutions; For images at each level, texture features are extracted using the local binary mode algorithm, and principal component analysis is used to reduce the dimensionality of the extracted texture features. Topographic relief features are obtained by overlaying digital elevation model data with images at each level. Based on the spectral feature extraction algorithm, the spectral reflectance data of land features are extracted from the image portion to obtain land feature type features. The texture features, terrain undulation features, and land cover type features after dimensionality reduction are fused to obtain the regional features of each region adjacent to the engineering area.

[0046] Further, based on the regional characteristics and project type, the adjacent regions corresponding to the project region are determined, including: The regional features are converted into feature vectors, and a type vector containing key attributes of the project type is constructed. Using a multi-head attention mechanism, the attention weight between the feature vector and the type vector is calculated to highlight the regional features that are highly correlated with the current project type; the regional features are then weighted and fused based on the attention weight to obtain a fused feature vector. The fused feature vector is matched with feature templates at different preset distance thresholds for similarity. When the similarity exceeds the judgment threshold, the corresponding region is determined to be the adjacent region corresponding to the engineering region.

[0047] Furthermore, determining the UAV's reconnaissance trajectory based on the outer boundary line of the adjacent region includes: The outer boundary line of the adjacent area is imported into the geographic information system for vectorization processing to generate three-dimensional spatial boundary data with elevation information. A genetic algorithm is used to perform path planning on the three-dimensional spatial boundary data, with the objective function of covering all adjacent areas and minimizing the flight path. This is combined with the performance parameters of the UAV and the limiting factors in the constraint database to generate the UAV's preliminary reconnaissance trajectory. The generated preliminary exploration trajectory is smoothed to obtain the exploration trajectory.

[0048] This invention also discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calling the executable program code stored in the memory to execute the method described in Embodiment 1.

[0049] This invention also discloses a computer storage medium storing a computer program, which is executed by a processor as described in Embodiment 1.

[0050] This invention also discloses a computer program product comprising a computer program that can be executed by a processor to implement the method as described in any of the preceding embodiments.

[0051] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, all of which fall within the scope of protection of the present invention.

Claims

1. A method for automated engineering surveying based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: First image data of the area to be surveyed is captured, and engineering areas and non-engineering areas corresponding to engineering drawings are identified from the first image data; Based on the image portion of the non-engineering area, the regional features of each region adjacent to the engineering area are extracted, and based on the regional features and the engineering type, the adjacent regions corresponding to the engineering area are determined. The UAV's survey trajectory is determined based on the outer boundary line of the adjacent area, and the UAV is controlled to conduct automatic engineering surveys according to the survey trajectory.

2. The method for automated engineering surveying based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: The engineering area and non-engineering area corresponding to the engineering drawing are identified from the first image data, including: The engineering drawings are converted into electronic drawings with the same resolution as the first image data according to the scale, and the electronic drawings are then subjected to noise reduction processing. The first image data is semantically segmented using a deep learning model to obtain the category information of each pixel in the first image data. By spatially matching the outline of the engineering design area in the electronic drawing with the semantically segmented pixel area in the first image data, the engineering area and non-engineering area are identified.

3. The method for automated engineering surveying based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that: Based on the image portion of the non-engineered area, the regional features of each region adjacent to the engineered area are extracted, including: Multi-scale pyramid decomposition is performed on the image portion of the non-engineering area to obtain hierarchical images of different resolutions; For images at each level, texture features are extracted using the local binary mode algorithm, and principal component analysis is used to reduce the dimensionality of the extracted texture features. Topographic relief features are obtained by overlaying digital elevation model data with images at each level. Based on the spectral feature extraction algorithm, the spectral reflectance data of land features are extracted from the image portion to obtain land feature type features. The texture features, terrain undulation features, and land cover type features after dimensionality reduction are fused to obtain the regional features of each region adjacent to the engineering area.

4. The method for automated engineering surveying based on unmanned aerial vehicles (UAVs) according to claim 3, characterized in that: Based on the regional characteristics and project type, the adjacent regions corresponding to the project region are determined, including: The regional features are converted into feature vectors, and a type vector containing key attributes of the project type is constructed. Using a multi-head attention mechanism, the attention weight between the feature vector and the type vector is calculated to highlight the regional features that are highly correlated with the current project type; the regional features are then weighted and fused based on the attention weight to obtain a fused feature vector. The fused feature vector is matched with feature templates at different preset distance thresholds for similarity. When the similarity exceeds the judgment threshold, the corresponding region is determined to be the adjacent region corresponding to the engineering region.

5. A method for automated engineering surveying based on unmanned aerial vehicles (UAVs) according to any one of claims 1-4, characterized in that: Determining the UAV's reconnaissance trajectory based on the outer boundary line of the adjacent region includes: The outer boundary line of the adjacent area is imported into the geographic information system for vectorization processing to generate three-dimensional spatial boundary data with elevation information. A genetic algorithm is used to perform path planning on the three-dimensional spatial boundary data, with the objective function of covering all adjacent areas and minimizing the flight path. This is combined with the performance parameters of the UAV and the limiting factors in the constraint database to generate the UAV's preliminary reconnaissance trajectory. The generated preliminary exploration trajectory is smoothed to obtain the exploration trajectory.

6. An automated engineering survey system based on unmanned aerial vehicles (UAVs), comprising a processor and a memory, characterized in that: The processor calls and executes the instructions stored in the memory to achieve the following steps: First image data of the area to be surveyed is captured, and engineering areas and non-engineering areas corresponding to engineering drawings are identified from the first image data; Based on the image portion of the non-engineering area, the regional features of each region adjacent to the engineering area are extracted, and based on the regional features and the engineering type, the adjacent regions corresponding to the engineering area are determined. The UAV's survey trajectory is determined based on the outer boundary line of the adjacent area, and the UAV is controlled to conduct automatic engineering surveys according to the survey trajectory.

7. The unmanned aerial vehicle (UAV)-based automated engineering survey system according to claim 6, characterized in that: The engineering area and non-engineering area corresponding to the engineering drawing are identified from the first image data, including: The engineering drawings are converted into electronic drawings with the same resolution as the first image data according to the scale, and the electronic drawings are then subjected to noise reduction processing. The first image data is semantically segmented using a deep learning model to obtain the category information of each pixel in the first image data. By spatially matching the outline of the engineering design area in the electronic drawing with the semantically segmented pixel area in the first image data, the engineering area and non-engineering area are identified.

8. An electronic device, comprising: Memory containing executable program code; A processor coupled to the memory; characterized in that: the processor calls the executable program code stored in the memory to perform the method as described in any one of claims 1-5.

9. A computer storage medium storing a computer program, characterized in that: The computer program is executed by the processor to perform the method as described in any one of claims 1-5.

10. A computer program product, characterized in that: The computer program product includes a computer program that can be executed by a processor to implement the method as described in any one of claims 1-5.

Images