Unmanned aerial vehicle-based hull sub-block precision detection method

Abstract

The invention discloses a hull sub-block precision detection method based on an unmanned aerial vehicle, and the method comprises the steps: collecting the multi-view image data of a hull sub-block through the unmanned aerial vehicle after survey area investigation and initial terrain collection, reconstructing a refined three-dimensional live-action sub-block model, and carrying out the part monomer cutting of the refined three-dimensional live-action sub-block model, and metadata is made to detect the precision of the ship body sub-block. According to the method, the labor cost and the safety risk of hull sub-block precision detection are reduced, the data quality is improved, and the problems that a traditional detection method is low in detection efficiency and single in data are solved.

Description

Technical Field

[0001] The present invention relates to the field of hull sub-assembly measurement, and particularly to a method for detecting the accuracy of hull sub-assemblies based on an unmanned aerial vehicle (UAV). Background Art

[0002] At present, most hull construction adopts the sub-assembly construction method. A large or medium-sized hull is often divided into dozens to nearly a hundred sub-assemblies. The sub-assemblies are first built separately on jigs or platforms in workshops or other sites, and then the built sub-assemblies are transported to the shipbuilding berth for the overall assembly of the hull. Therefore, it is necessary to measure and analyze the accuracy of the sub-assemblies to determine whether they fully meet the accuracy requirements and match the design drawings before being lifted onto the shipbuilding berth.

[0003] Traditional detection methods mainly use contact detection methods, such as three-coordinate point marking, jigs and fixtures, etc. Although these technologies have been used for many years and are very mature, they have problems such as high labor costs, high safety risks, low detection efficiency, and single data. Moreover, total station technology is subject to foreign countries. Therefore, a new type of sub-assembly detection method is urgently needed to break through these problems.

[0004] With the rapid development of UAV measurement and computer vision technology, new digital opportunities have been brought to the hull industry. The UAV close-range photogrammetry technology has the advantage of high-precision and rapid modeling, providing a new detection method for the measurement of hull sub-assemblies during the hull construction process. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for detecting the accuracy of hull sub-assemblies based on an unmanned aerial vehicle (UAV), so as to reduce the labor cost and safety risk of hull sub-assembly accuracy detection, and solve the problems of low detection efficiency and single data of traditional detection methods.

[0006] The technical solution for achieving the above purpose is as follows:

[0007] A method for detecting the accuracy of hull sub-assemblies based on an unmanned aerial vehicle (UAV) uses the UAV to collect multi-view image data of the hull sub-assemblies, reconstructs a refined three-dimensional real-scene sub-assembly model, then performs post-processing on the model, and uses the obtained metadata to detect the accuracy of the hull sub-assemblies.

[0008] Preferably, the sub-assembly accuracy detection refers to: in hull construction using the sub-assembly construction method, performing accuracy measurement and analysis on each sub-assembly before the sub-assemblies of a large or medium-sized hull are transported to the shipbuilding berth for the overall assembly of the hull.

[0009] Preferably, the use of the UAV to collect multi-view image data of the hull sub-assemblies includes: using the UAV close-range photogrammetry technology to collect multi-view image data of the sub-assemblies, and conducting a survey area investigation and initial terrain collection before data collection.

[0010] Preferably, the initial terrain collection includes:

[0011] Use a drone remote control device to set up the mapping aerial photography area. On the premise of ensuring safety, conduct a grid flight over the survey area to obtain multi-angle aerial images.

[0012] After the acquisition of the external field data of the aerial photos is completed, first, conduct a quality inspection on the obtained images, and re-fly the unqualified areas until the quality of the obtained images meets the requirements.

[0013] Export the original images and POS (data obtained through the Position and Orientation System, abbreviated as the POS system) data, set the geographic reference system to the WGS-84 coordinate system (World Geodetic System 1984, a coordinate system established for the use of the GPS global positioning system), import the camera calibration parameters, and through aerial triangulation encryption processing, recalculate the photo azimuth and attitude at the moment of photography.

[0014] Use the results of aerial triangulation encryption to automatically output three-dimensional point cloud data.

[0015] Preferably, the collection of multi-view image data of the hull sub-assembly includes: using drone close-range photogrammetry and flight path planning to obtain multi-view image data of the sub-assembly to be measured, and combining the close-range photogrammetry and three-dimensional reconstruction algorithm to obtain a three-dimensional terrain grid of the area of the sub-assembly to be measured.

[0016] Preferably, the reconstruction of the refined three-dimensional real-scene sub-assembly model means: using the position and attribute information of the photos taken of the sub-assembly and the three-dimensional terrain grid to produce a refined three-dimensional real-scene sub-assembly model.

[0017] Preferably, the flight path planning includes: using the midpoints of the two short sides of the target outer contour rectangle (v 1 , v 2 ) and the width w of the rectangle to describe it, then the long side direction of the rectangle is the main direction of the flight path, that is, the direction of the vector (v 1 , v 2 ); when the shooting plane changes from the vertical plane to the horizontal plane, the distance between adjacent exposure points in the same flight path remains Δs, but the rotation angle φ of the drone lens changes from 0° to -90°, and the distance Δh between adjacent exposure points in the vertical direction also correspondingly becomes the distance Δw between adjacent flight paths. Then the number of flight paths required to shoot the top surface of the building in a vertically downward manner is:

[0018]

[0019] where, [a] represents rounding up a; d is the distance from the shooting plane to the measured object surface, Oy is the camera overlap degree in the y - direction; fov y is the field of view angle in the y - direction.

[0020] Preferably, the close - range photogrammetry includes: under the assistance of the initial terrain information, solving and planning the optimal unmanned close - range photography position and the photography attitude facing the terrain surface, where the endpoint coordinates of the vertically - downward shooting route can be calculated according to the following formula:

[0021]

[0022] where K represents the unit vector in the direction after rotating the vector (v 1 , v 2 ) clockwise by 90°; N 1 is the number of routes;

[0023] If the lens rotation angle of the oblique route is then the endpoint coordinates of the two oblique routes are respectively:

[0024]

[0025] The distance between adjacent exposure points in the oblique route is also Δs.

[0026] Preferably, the production of the refined three - dimensional real - scene sectional model means: submitting the result data after aerial triangulation to generate three - dimensional TIN grid construction, white - body three - dimensional model creation, automatic texture mapping, and three - dimensional scene construction.

[0027] Preferably, the model post - processing includes: performing component monomerization cutting on the generated sectional model and making it into the metadata to detect the accuracy of the hull sectional model.

[0028] The beneficial effects of the present invention are as follows: The present invention adopts the unmanned aerial vehicle (UAV) close - range photogrammetry technology to reduce manual participation, solves the problem of high labor cost in traditional detection, enables the UAV to replace humans to enter dangerous areas for measurement, and also avoids the safety risk problems brought by traditional detection. In the three - dimensional trajectory and attitude planning, according to the shape of the target outer contour, relevant parameters of the route, such as the number of routes, endpoint coordinates, etc., are calculated through specific formulas. At the same time, for different shooting surfaces and complex targets, corresponding routes are planned, reducing the time consumed by the UAV in route switching during the shooting process, improving the shooting efficiency and data quality. Using the UAV to quickly obtain multi - perspective images, combined with algorithms to produce three - dimensional terrain grids and models, improving the detection efficiency. At the same time, by obtaining multi - angle image data, establishing a refined three - dimensional model, performing component monomerization cutting on the model and making metadata, more accurate and detailed full - system data are obtained, enriching the data dimension and facilitating the subsequent refined management of the shipyard. Brief Description of the Drawings

[0029] Figure 1 It is the flow chart of the hull sub-assembly and overall assembly precision detection method based on unmanned aerial vehicle (UAV) of the present invention;

[0030] Figure 2 It is the specific flow chart of the survey area reconnaissance in step S1 of the present invention;

[0031] Figure 3 It is the specific flow chart of the initial terrain collection in step S2 of the present invention;

[0032] Figure 4 It is the specific flow chart of the field data collection in step S3 of the present invention;

[0033] Figure 5 It is the specific flow chart of the model reconstruction in step S4 of the present invention;

[0034] Figure 6 It is the specific flow chart of the model post-processing in step S5 of the present invention;

[0035] Figure 7 It is the route of the target outer contour close to the track plan and vertically downward in step S31 of the present invention;

[0036] Figure 8 It is the route of the target outer contour close to the track plan and obliquely photographed in step S31 of the present invention. Detailed implementation manners

[0037] The present invention will be further described below in conjunction with the accompanying drawings.

[0038] Please refer to Figure 1 , the hull sub-assembly and overall assembly precision detection method based on UAV of the present invention includes the following steps:

[0039] Step S1, please refer to Figure 2 , survey area reconnaissance, reconnaissance of the key areas of the shipyard sub-assembly and overall assembly, understand the key points and precautions for information collection, including:

[0040] Step S11, determine the survey area range according to the sub-assembly and overall assembly characteristics, survey area environment and measurement requirements, generally according to the terrain and buildings;

[0041] Step S12, estimate the workload, construction period, and the number of personnel for data collection and processing, and select appropriate UAVs, sensors and measurement parameters.

[0042] Step S2, please refer to Figure 3 , collect the initial terrain information and establish a rough model, including:

[0043] Step S21, use the UAV remote control device to set the mapping aerial photography area, and fly in a grid pattern over the survey area to obtain multi-angle aerial images on the premise of ensuring safety;

[0044] In step S22, after the acquisition of the field data of the aerial photos, first, the quality of the acquired images needs to be inspected, and the areas with unqualified quality are re-flown until the quality of the acquired images meets the requirements.

[0045] In step S23, the original images and POS data are exported, the geodetic reference system is set to the WGS-84 coordinate system, the camera calibration parameters are imported, and through the aerotriangulation processing, the photo orientation and attitude at the moment of photography are recalculated.

[0046] In step S24, the 3D point cloud data is automatically output using the results of the aerotriangulation.

[0047] In step S3, please refer to Figure 4 , for the field data collection, the multi-view image data of the hull sub-assemblies and the total assembly are collected using an unmanned aerial vehicle (UAV), including:

[0048] In step S31, for the 3D flight path and attitude planning, with the assistance of the initial terrain information, the optimal UAV close-range photography position and the photography attitude facing the terrain surface are solved to guide the UAV to obtain high-resolution images of a specific area.

[0049] To reduce the time consumed by the UAV in route switching during the shooting process, the long side of the target outer contour should be used as the main flight path direction, and the target top surface should be covered with a smaller number of flight paths. Please refer to Figure 7 , the midpoints (v 1 , v 2 ) of the two short sides of the target outer contour rectangle and the width w of the rectangle are used to describe it, then the long side direction of the rectangle (i.e., the direction of the vector (v 1 , v 2 ) is the main direction of the flight path.

[0050] When the shooting plane changes from the vertical plane to the horizontal plane, the distance between adjacent exposure points in the same flight path remains Δs, but the rotation angle φ of the UAV camera changes from 0° to -90°, and the distance Δh between adjacent exposure points in the vertical direction also correspondingly becomes the distance Δw between adjacent flight paths. Then, the number of flight paths required to shoot the top surface of the building in the vertically downward manner is:

[0051]

[0052] where, [a] represents rounding up a; Δh is the distance between adjacent exposure points in the vertical direction; Δw is the distance between adjacent flight paths; d is the distance from the shooting plane to the measured object surface, O y is the camera y-direction overlap; fov y is the field of view angle in the y direction.

[0053] The endpoint coordinates of these vertically downward shooting flight paths can be calculated according to the following formula:

[0054]

[0055] Among them, K represents the unit vector in the direction after the vector (v 1 , v 2 ) is rotated clockwise by 90°; N 1 is the number of flight lines; Δw is the distance between adjacent flight lines.

[0056] To ensure that the images in the inclined flight lines can capture both the top and side surfaces simultaneously, the principal optical axis of the camera needs to point to the edge of the top, that is, the lengths of the two sides of the rectangle. Please refer to Figure 8 . If the lens rotation angle of the inclined flight line is , then the endpoint coordinates of the two inclined flight lines are respectively:

[0057]

[0058] The distance between adjacent exposure points in the inclined flight line is also Δs. Finally, along the direction of the vector K, according to the distance from the straight line where the vector (v 1 , v 2 ) is located, N 1 +2 flight lines are sorted, and they are combined in a serpentine flight line manner to obtain the flight path planning result of the top surface. For other complex targets, they can be dissected into a combination of multiple rectangles, and the close flight path planning is carried out respectively according to the method described above to obtain an ideal flight path;

[0059] Step S32, flight line safety inspection. Based on the initial terrain, simulate the generated flight lines in the software to ensure the flight safety of the flight lines;

[0060] Step S33, export and import of flight lines. Export the confirmed safe flight lines from the computer software to the corresponding removable storage device, and then import the generated flight lines into the flight remote controller through this storage device;

[0061] Step S34, UAV close flight. The UAV follows the predetermined flight line and takes images of the to-be-measured sub-assembly at different angles and heights through the close-range photogrammetry method to obtain multi-view image data.

[0062] Step S35, "handheld" supplementary shooting of key areas. For sub-assembly structures with a large degree of complexity or areas where the UAV is difficult to enter, "handheld" supplementary shooting operations are carried out on them. "Handheld" supplementary shooting includes manually controlling the flight of the UAV and flying the UAV by hand.

[0063] Step S36, collection of field control points. To improve the shooting accuracy, the requirements for point layout should be fully considered, and ground feature points and target points that are well visible in topographic survey and can be clearly identified should be selected; control points need to be laid out around the survey area.

[0064] Step S4, refer to Figure 5 , and reconstruct the refined three-dimensional real-scene sectional model, including:

[0065] Step S41, data inspection, check the acquired frontal images, cross images, and "handheld" supplementary images for vibration;

[0066] Step S42, data preprocessing, adjust the color, brightness, and contrast of the original images and perform color homogenization to reduce the tonal differences between images;

[0067] Step S43, relative orientation, perform dense matching of feature points of multi-view images for camera parameters, image data, and POS data, and perform free network multi-view image joint constrained adjustment and solution for the regional network to establish a stereo model that can be moderately freely deformed in the spatial scale and complete the relative orientation;

[0068] Step S44, absolute orientation, transfer the photo control point results measured in the field (measurement and data acquisition work carried out on-site) in the indoor (data processing and analysis work carried out indoors) environment, perform constrained adjustment and solution for the existing regional network model, and incorporate the regional network into the precise geodetic coordinate system to complete the absolute orientation;

[0069] Step S45, model reconstruction, use the resulting data after aerotriangulation.

[0070] Step S5, refer to Figure 6 , and perform post-processing on the model, perform component monomerization cutting on it, and make metadata to detect the accuracy of the hull sectional model, including:

[0071] Step S51, component monomerization, cut the three-dimensional TIN grid model into individual and selectable entities, so that they can be attached with attributes and queried and statistically analyzed for subsequent management;

[0072] Step S52, metadata production, produce metadata for the cut monomerization, label important buildings, tanks, pipelines, etc. Manage the holographic information of the entire shipyard sectional model, not limited to geographical entities (points, lines, surfaces, volumes), but also including related time sequences, spatial positions, place names, addresses, attribute characteristics, and other information data required.

[0073] The above embodiments are only for illustrating the present invention and are not intended to limit the present invention. Those skilled in the relevant technical fields can also make various transformations or variations without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of the present invention and should be defined by each claim.

Claims

1. The method for detecting the accuracy of ship hull sections based on drones is characterized by: Unmanned aerial vehicles are used to collect multi-view image data of hull sections, and a refined three-dimensional real-life section model is reconstructed. The model is then post-processed, and the obtained metadata is used to detect the accuracy of the hull section.

2. The method for detecting the accuracy of hull sections based on a drone according to claim 1 is characterized in that: The sub-section accuracy inspection refers to: in the ship hull construction using the sub-section construction method, the accuracy measurement and analysis of each sub-section of the large and medium-sized hull is carried out before the sub-sections are transported to the slipway for hull assembly.

3. The method for detecting the accuracy of hull sections based on drones according to claim 1 is characterized in that: The method of collecting multi-view image data of the hull subsection by using a drone includes: collecting multi-view image data of the subsection by using a drone close-up photography technique, and conducting area survey and initial terrain collection before collecting data.

4. The method for detecting the accuracy of hull sections based on a drone according to claim 3 is characterized in that: The initial terrain collection includes: Use the drone remote control device to set the mapping aerial photography area, and perform a tic-tac-toe flight over the survey area while ensuring safety to obtain multi-angle aerial images; After the aerial photo field data is acquired, the quality of the acquired images must be checked first, and re-flights must be performed on unqualified areas until the acquired image quality meets the requirements; Export the original image and POS data, set the geographic reference system to the WGS-84 coordinate system, import the camera calibration parameters, perform aerotriangulation, and recalculate the photo orientation and posture at the moment of shooting; Automatically output 3D point cloud data using aerial triangulation results.

5. The method for detecting the accuracy of hull sections based on drones according to claim 3 is characterized in that: The method of collecting multi-view image data of the hull subsection includes: using unmanned aerial vehicle close-up photogrammetry and route planning to obtain multi-view image data of the subsection to be measured, and combining the close-up photogrammetry and three-dimensional reconstruction algorithm to obtain a three-dimensional terrain grid of the subsection area to be measured.

6. The method for detecting the accuracy of hull sections based on drones according to claim 5 is characterized in that: The reconstructing of the refined three-dimensional real scene segment model refers to: using the location and attribute information of the segment photos taken and the three-dimensional terrain grid to produce the refined three-dimensional real scene segment model.

7. The method for detecting the accuracy of hull sections based on a drone according to claim 5 is characterized in that: The route planning includes: using two sets of parameters, the midpoints of the two short sides of the target outer contour rectangle (v1, v2) and the width w of the rectangle, to describe it, then the direction of the long side of the rectangle is the main direction of the route, that is, the direction of the vector (v1, v2); when the shooting surface is transformed from the vertical plane to the horizontal plane, the distance between adjacent exposure points in the same track is still Δs, but the rotation angle φ of the drone lens changes from 0° to -90°, and the distance Δh between adjacent exposure points in the vertical direction is also correspondingly changed to the distance Δw between adjacent routes, then the number of routes required to shoot the top of the building in a vertical downward manner is: Where, [a] represents rounding up a; d is the distance from the photographic surface to the object surface, O y is the camera overlap in the y direction; fov y is the field of view angle in the y direction.

8. The method for detecting the accuracy of hull sections based on a drone according to claim 5 is characterized in that: The close-up photogrammetry includes: with the assistance of the initial terrain information, solving and planning the optimal unmanned close-up photography position and the photography posture facing the terrain surface, wherein the endpoint coordinates of the vertical downward shooting route can be calculated according to the following formula: Where K represents the unit vector of the vector (v1, v2) after rotating 90° clockwise; N1 is the number of routes; If the lens rotation angle of the inclined route is , then the endpoint coordinates of the two inclined routes are: The distance between adjacent exposure points in the inclined flight path is also Δs.

9. The method for detecting the accuracy of hull sections based on a drone according to claim 6 is characterized in that: The production of refined three-dimensional real-scene segment models refers to: submitting the result data after aerial triangulation to generate three-dimensional TIN grid construction, white body three-dimensional model creation, automatic texture mapping and three-dimensional scene construction.

10. The method for detecting the accuracy of hull sections based on a drone according to claim 9 is characterized in that: The model post-processing includes: cutting the generated sub-section model into individual components, and making the metadata to detect the accuracy of the hull sub-section.

Images