Overwater and underwater integrated three-dimensional model construction method based on image processing

Abstract

The invention discloses an overwater and underwater integrated three-dimensional model construction method based on image processing. According to the method, an unmanned aerial vehicle is adopted to carry a photogrammetry system to obtain aerial remote sensing image data; carrying a laser radar measurement system to obtain point cloud data of a navigable river channel and an overwater part 200 meters away from the two sides of a bank; an unmanned surveying ship is matched with a manned ship and underwater surveying equipment to obtain the water depth and underwater topographic data of a navigable river channel, and finally, underwater and overwater data are fused. According to the method, the constructed overwater data and underwater data can be fused, so that the overwater and underwater integrated three-dimensional model is obtained, the model is conveniently utilized for evaluation, and the evaluation accuracy and the utilization rate are improved; by adopting a mode of manned ship power guarantee and unmanned ship water area measurement, the problems of torrential water flow and insufficient horsepower, high energy consumption and short endurance time of an unmanned ship which sails against water in an inner navigation river channel are solved, and the working efficiency is greatly improved.

Description

Technical Field

[0001] The invention relates to an above-water and underwater integrated three-dimensional model construction method based on image processing, belonging to the technical field of above-water and underwater integrated three-dimensional model construction. Background Art

[0002] The navigable mileage of water transport is 3,900 kilometers, but there is only a handful of underwater terrain information, resulting in a generally low level of water management informationization. Safety accidents in the fields of water transport and water use occur from time to time, which seriously restricts the protection and utilization of water resources in our province; carry out underwater terrain mapping of major rivers, lakes, reservoirs and other water bodies, realize integrated mapping of water and underwater, and fill in the gaps in underground and underwater space data. During the testing process, only the establishment of water models and underwater models has been achieved separately, and there is no literature recording the construction method of the overall water and underwater models, which leads to a lack of comprehensive evaluation results during the evaluation, and the results after the evaluation are not scientific and accurate enough. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a method for constructing an integrated above-water and underwater three-dimensional model based on image processing, which can realize the fusion of above-water and underwater models and obtain an integrated above-water and underwater three-dimensional model, which is more conducive to using the model for river channel assessment, making the assessment more accurate and improving the utilization rate.

[0004] The technical solution adopted by the present invention is: a method for constructing an integrated above-water and underwater three-dimensional model based on image processing, the method comprising the following steps: (1) Use an unmanned aerial vehicle (UAV) equipped with a photogrammetry system to obtain aerial remote sensing image data with a resolution better than 0.05 m for 200 m on both sides of the navigable river and the coast; (2) Use a UAV equipped with a laser radar measurement system to obtain point cloud data of the navigable river and the 200-meter-wide water section on both sides of the coast; (3) Use an unmanned survey vessel equipped with a highly integrated high-frequency multi-beam echo sounder, GNSS receiver, integrated deck unit, and industrial computer to conduct underwater measurements and obtain data on the water depth and underwater topography of navigable rivers; During the measurement, it is necessary to cooperate with manned ships, and adopt the method of manned ship power support + unmanned ship water area measurement. The manned ship is mainly responsible for towing the unmanned ship equipment and logistics support. The unmanned ship is specifically responsible for the measurement. The unmanned ship is towed from the rapids channel to the upstream, and then underwater measurement is carried out. At the same time, the manned ship uses the unmanned ship remote control function to control the unmanned ship to reach the shallow areas on both sides for measurement, and go back and forth many times until the measurement range completely covers the water area; Choose to measure the water area when the river is at flood level to obtain the river depth and elevation data to the greatest extent possible. Use drones equipped with airborne laser radar measurement systems and traditional engineering measurement methods to measure both sides of the river and some areas of the river shoal during low water levels, and finally fuse the data; (4) Deploy a set number of image control points according to the regional network point layout requirements, and use GPS (RTK) combined with Beidou satellite navigation positioning GZCORS to measure image control points; (5) Use ContextCapture Center or Huace Copre software to perform aerial triangulation to generate a real-life 3D model of the above-water part and a true orthogonal image TDOM; (6) Using the Huace Copre software, the data collected by the airborne laser radar measurement system is solved by the base station and the point cloud color matching is performed to generate the above-water point cloud data; (7) Use CMS multi-beam post-processing software to process the data collected by the multi-beam echo sounder to generate navigable river depth and underwater topography data; (8) Use the airborne LiDAR point cloud data processing system to filter the integrated above-water and underwater point cloud data to generate a digital elevation model (DEM); (9) Use Global Mapper and OSGBLab Enterprise system to fuse the above-water real-life 3D model and the underwater digital elevation model (DEM) to generate an integrated above-water and underwater 3D model; (10) In DasViewer, use relevant functional modules to convert the integrated 3D model of water and water to generate a digital surface model (DSM); (11) In Southern CASS 3D, the generated integrated 3D model of water and water is used for data collection to generate a digital linear graph (DLG).

[0005] Furthermore, when drones are used for flight as mentioned above, aerial photography zoning is required. The aerial photography zoning method is: determine the reference surface elevation based on the minimum, maximum, median, mean and majority values ​​of the elevation of the DEM statistical photography area. When the height difference of the test area is too large to simultaneously meet the minimum overlap, flight safety and minimum ground resolution, aerial photography zones should be divided. The aerial photography zones are divided based on the terrain undulations and flight safety conditions of the test area to ensure that there are no mountains higher than the flight height within 1 km in front of the end of each zone route. After zoning, each zone determines the average reference surface elevation of the zone according to the aerial photography specifications and designs the route.

[0006] Furthermore, the relative altitude calculation formula of the drone is as follows: (1) Where:h - flight altitude; f —Focal length of the lens (35 mm); a—pixel size (3.9 μm); GSD — ground resolution; When planning the phased flight path, the ground resolution GSD is set in combination with the camera and aircraft parameters to ensure that the ground resolution of the survey area is better than 5 cm. The relative altitude is calculated according to formula (1). For example, when the design GSD-ground resolution is 0.03 m, the relative altitude of the terrain simulation flight is 269 meters. The absolute altitude is the relative altitude plus the reference surface elevation.

[0007] Furthermore, the above-mentioned UAV aerial survey adopts a terrain-imitating flight mode and multi-angle aerial photography, with a heading overlap of ≥80%, a lateral overlap of ≥70%, and a route curvature of no more than 1%. When the route length is less than 5km, the maximum route curvature is no more than 3%.

[0008] Furthermore, the supplementary photography and re-photography during the above-mentioned drone aerial survey process include: 1) For absolute holes, relative holes and other serious defects that appear during aerial photography, retake photos of the corresponding positions of absolute holes, relative holes and other serious defects; 2) When the GNSS signal of the ground base station is lost, interrupted or other reasons cause the azimuth elements outside the scan line to be unable to be solved and the solution accuracy of the single-point positioning technology cannot meet the set requirements, supplementary or re-shooting should be carried out according to the size of the solution range; 3) When the IMU data recording is interrupted or the initial alignment flight is insufficient, the entire flight data cannot be correctly calculated; 4) The loophole re-shooting is carried out according to the original design track, and the length of the re-shooting route meets the requirements of the user's regional network density.

[0009] 5) For relative loopholes and local defects (such as clouds, cloud shadows, spots, etc.) that do not affect the encrypted point selection and model connection, additional photography is only required at the loopholes, and the length of the additional photography route exceeds the four baselines outside the loopholes.

[0010] Furthermore, the above underwater mapping steps: a) Survey line planning: The planning method is: 1) The survey line is planned based on the water depth and passability of the main rivers, and the strip width is three to four times the water depth; 2) Survey line planning in unobstructed areas is conducive to base station differential data transmission, and acquisition route planning should take into account the overlap rate of beam width; 3) After completing the survey area, select the location where the underwater terrain has a slope change as the correction line. The round-trip survey line on flat ground is used to correct the roll, the round-trip survey line on sloping ground is used to correct the pitch, and the parallel survey line on sloping ground is used to correct the heading; b) Multi-beam equipment assembly: The underwater probe of the multi-beam echo sounder is connected and fixed on the hull of the unmanned ship through the hull mounting plate, the three antennas are installed at the corresponding positions of the unmanned ship, and the batteries are placed in the corresponding battery slots and connected to power; c) Positioning base station installation: Choose an open area to set up the GNSS base station, use the China Measurement X10 receiver and DL5 external radio, set the correct engineering parameters and appropriate radio channels, and use known points to set up the station to avoid the inconvenience of parameter calibration caused by the small number of control points in the survey area; d) Launch and start-up inspection: Start the unmanned boat, check whether the power is normal, connect all sensors to the acquisition terminal, check whether the GNSS signal, attitude signal, and surface sound speed are normal; open the sonar acquisition software to check whether the echo image is normal; adjust the upper and lower thresholds, beam opening angle and working mode according to the operating environment, and turn on the roll and pitch stabilization; e) Multi-beam data acquisition: 1) Before the acquisition operation, the multi-beam echo sounder needs to be calibrated. Open the Norbit GUI multi-beam control software, follow the software prompts, activate and adjust the inertial navigation system, and sail along the figure-8 route upstream and downstream until the attitude accuracy meets the established requirements, usually 0.5°. After the calibration is completed, the formal acquisition work can be carried out. During this period, the GPS signal is guaranteed to be a fixed solution; 2) Use a ruler to measure the relative position of each hardware phase center to the center of the mounting flange and input it into the software. Set the opening angle, depth threshold, frequency and beam mode of the multi-beam echo sounder according to the actual situation of the survey area; 3) Open the CMS multi-beam data acquisition software, select the engineering parameters corresponding to the base station, add the Norbit iWBMS hardware (multi-beam echo sounder), edit the survey area matrix graphics, add the planned route, and set the display resolution and other parameters; 4) After starting to record data, the ship will sail along the planned route and correct the route in real time according to the sweep width until the entire survey area is scanned.

[0011] 5) Check whether the designed collection routes have been completed, and the routes that have been missed should be supplemented in time.

[0012] Furthermore, the above-mentioned UAV image control measurement method includes: The point distribution method is: (1) The layout range of control points should meet the mapping range of the project area, and all control points should be arranged at level and high points; (2) Adopt the corner point and inflection point layout method, that is, to arrange flat height points at the turning points of convex corners and concave corners of the regional network; (3) Each encryption unit should have 2-3 redundant control points as encryption check points. The check points should be selected in places with clear images and clear targets; (4) Since RTK technology is mainly used to determine the plane position of the flat high point, the feasibility and convenience of RTK measurement must also be considered when selecting the point. The selected point should be convenient for the placement of receiving equipment and operation, and the field of view should be wide; (5) The image control point target is selected at the intersection of the image and the ground object with an intersection angle between 30° and 150°; (6) Image control points should be selected from obvious features that are clear on adjacent images and meet the requirements for measuring plane position and elevation, intersection points of thin linear features that are nearly orthogonal, obvious corner points of features, or fixed features’ centers. Arc-shaped features, places with abrupt changes in local elevation, shadows, and features that are easily variable should not be selected. (7) Control points are arranged at the junction of aerial photography areas: Control points are arranged in the overlapping areas of adjacent flight routes. Adjacent areas should be shared as much as possible. If the sharing requirement cannot be met, points should be arranged separately. (8) The city level meets the 1:500 topographic mapping accuracy, the spacing between image control points is 400-500 meters, and each point evenly covers the survey area; The selection and trimming methods of image control points are as follows: (1) The targets of the image control points are selected from obvious objects with clear images, intersections of linear objects with good intersection angles, and corner vertices of obvious objects; (2) The main reference objects for selecting image control points include: the corners of flat-roofed houses, the top corners of fences, the top corners of gate piers, the top corners of gate covers, and other piers and columns that are convenient for measuring height and distance. The corners of gabled houses and those with irregular local elevation changes should not be used as puncture point targets; (3) For maps with fewer objects, some tiny landforms, such as the corners of ridges, small corners, and the center of small pits with a diameter of less than 0.2 mm on the map can be used. When using such objects, care should be taken to compare them with the actual field repeatedly to ensure that there is no change in the actual field after photography. (4) Select points and mark them on the digital image, accurately indicating the location of the selected points; (5) Check and organize the measurement results of image control points; (6) Prepare image control point results table and point information table.

[0013] The image control point measurement method is: The image control points are measured using network RTK, and the normal height of the image control points is obtained by using the elevation anomaly correction of the quasi-geoid refined model. The network RTK observation should meet the following requirements: (1) The instrument should be initialized before observation; (2) First, check the known points. The test results are obtained and the accuracy meets the requirements. The observation values ​​should be recorded after the RTK fixed solution is obtained and the convergence is stable. (3) Repeat the observation for no less than three times. When the difference between the measurements is less than 0.05 m, the average value is taken as the final result of the point.

[0014] Furthermore, the above-mentioned method for generating the above-mentioned 3D real-scene model on water is as follows: using 3D real-scene modeling software (ContextCapture Center software) to densely match the camera parameters, image data, and POS data with multi-view image feature points, and thereby perform a joint constraint adjustment solution of the free network multi-view image of the regional network, and establish a stereo model that can be appropriately freely deformed at a spatial scale to complete relative orientation; the measured photo control points are constrained to solve the existing regional network model, and the regional network is incorporated into an accurate geodetic coordinate system to complete absolute orientation. Fully automatic 3D modeling is performed using a 3D real-scene modeling software (ContextCapture Center software) with multi-machine and multi-node parallel computing. The modeling process is as follows Figure 4 That is, the aerial triangulation data is directly submitted to generate 3D TIN grid construction, white body 3D model creation, self-service texture mapping and 3D scene construction. In principle, the model modification only repairs the water area gaps or model holes. The model is regenerated after constraint intervention using 3D real scene modeling software (ContextCapture Center software) and water surface or supplementary flight data to ensure that there are no loopholes in the model. (1) The number of photos in the survey area is 20,000 to 30,000 photos per partition. There must be at least three overlapping routes between the partitions. There must be common image control points to facilitate aerial triangulation. After the aerial triangulation encrypted partitions are divided, they are divided according to the flight data and image control conditions, which may not be consistent with the aerial photography partitions; (2) Each encrypted partition of the triangulation is modeled separately. Each partition must have a unified origin (coordinate origin and tile origin) and size. Finally, merge the triangulation and select only the edge tiles to output the model. Then copy the edge tiles of each encrypted partition to a Data folder. After the edge tiles are copied to the Data file, replace the files with the same name, and then create a total index file.

[0015] Furthermore, the above-mentioned water point cloud data processing method has the following steps: (1) Data acquisition: Generate mission areas based on LiDAR data and obtain high-precision LiDAR point cloud data; (2) Trajectory solution: the original real-time GNSS / IMU trajectory data and CORS station data are jointly solved to obtain accurate trajectory information; (3) Laser point cloud solution, combined with accurate trajectory information, to obtain laser point cloud data in LAS format; (4) Noise removal: performing operations such as gross error elimination on laser point cloud data to remove erroneous points and abnormal elevation points; (5) Matching between flight strips: matching adjacent flight strips of laser point clouds to reduce the system error between flight strips; (6) Point cloud output: finally, the processed point cloud data of the above-water part is output.

[0016] Furthermore, the above underwater data processing method comprises the following steps: (1) Installation deviation calibration First, calibrate the roll deviation: select a pair of round-trip survey lines with flat underwater terrain, select a calibration line in the direction perpendicular to the route at the overlapped part of the survey lines, adjust the fitting step size and step, calculate the roll deviation through the software, exit the interface, reload the route file, and fill in the roll deviation value at the equipment deviation. Then calibrate the pitch deviation: select a pair of round-trip survey lines with a sloped underwater terrain, select a calibration line in the direction of the route at the overlapped part of the survey lines, and record the pitch deviation value; exit the software, reload the file, and enter all the roll and pitch deviations. Finally, calibrate the heading deviation: select a pair of parallel survey lines with a sloped underwater terrain, select a calibration line in the direction of the route at the overlapped part of the survey lines, record the heading deviation, reload all files, and enter the recorded installation deviation; (2) First-level editor Use CMS software to post-process the collected raw data. First, perform primary editing, open the raw data file, select the survey line to be processed, edit the view according to the plane, and perform cropping and noise editing as needed; (3) Secondary Editor Select the survey line that has been edited in the first level to enter the second level editing, mainly to delete the noise and flying points generated by the multi-beam edge clutter, and perform detailed processing frame by frame and line by line; (4) Level 3 Editor After the secondary editing, the data enters the tertiary editing mode, where the noise and flying spots generated by the multi-beam edge clutter are deleted in the 3D editing mode; (5) Data output The processed data file is compressed and saved in a specified format, usually .las format.

[0017] Furthermore, the above-water and underwater data fusion method has the following steps: 6.1) The steps of generating the integrated digital elevation model (DEM) above and below water are as follows: After merging the underwater data (las format) scanned by the multi-beam unmanned ship with the pre-processed surface point cloud data, the data was imported into the point cloud data processing software for point cloud data processing. Through point cloud segmentation, point cloud filtering, manual fine editing, TIN construction, and interpolation, a DEM model was generated. (1) Point cloud segmentation DEM generation often involves a wide range of areas, a huge amount of point cloud data, high requirements on computer configuration, and a large workload of later human-computer interaction and many people involved. Based on this, point cloud data is divided into blocks using standard map sheets to facilitate later collaborative work; (2) Point cloud filtering: using algorithms or algorithm combinations based on emission intensity, echo number, ground shape, etc. to filter laser point cloud data, classify data of objects such as buildings, vegetation, and power facilities into non-ground point data, extract surface data, and separate non-ground points; (3) Manual fine editing: Use laser point cloud classification software to construct TIN inspection for automatically classified data and correct misclassified data. Usually, the software automatic filtering results are retrieved in the point cloud data processing software to generate a surface triangulated network. The elevation information of the point cloud is checked by pulling sections of the triangulated network in different areas to check whether there are discontinuous, uneven, or leaky objects or terrain in the area, or whether there are large elevation differences of the same objects (i.e., obvious "pits" or "protruding points"), so as to determine whether the point cloud here is misclassified or underclassified. In order to check the misclassification of the point cloud in the surface details more carefully, the DOM or DLG data of the same period should be added to the point cloud triangulated network of the same area for terrain comparison and auxiliary judgment during the operation. Different terrains and objects are identified piece by piece. The point cloud manual classification tool is used to locally separate the misclassified or underclassified points into the corresponding point cloud categories according to the correct object type using an iterative algorithm to improve the accuracy of the final results of manual classification. During project generation, the main manual fine editing involved includes manual classification, manual point addition, addition of fracture lines, regional smoothing, foundation setting and other related operations.

[0018] 1) Small-area buildings and structures are easily misclassified as ground points. When manually classifying, these point clouds should be separated from the ground points and placed in the building point layer; 2) Due to vegetation coverage, there is a lack of point cloud data in some small areas. The real landform can be restored by adding appropriate points; 3) The ground points on the steep slopes may be misjudged as other land points (such as low vegetation), resulting in a lack of ground points and discontinuous terrain. This part needs to be manually filled in, usually by adding breaklines; 4) The noise points left after vegetation filtering and the temporarily piled garbage on the road surface may cause the ground to be uneven (i.e. obvious "pits" or "protruding points"). This part needs to be smoothed by the regional smoothing tool; 5) For houses built on the mountain or against the mountain, there is a certain height difference between the foundation of the house and the slope or steep slope behind the house. At the same time, such houses are often blocked by trees, resulting in no clear boundary ground points around the house. If the actual point cloud shows different height differences around the house, the foundation surface of the house will be a slope after the DEM is generated, which obviously does not conform to the actual situation of the house. The solution is: when the height difference of the ground points around the foundation of such houses is large, add a breakline to level the foundation. The elevation of the breakline is based on the correct foundation elevation around the house; (4) Construct TIN and interpolate to generate DEM. For the point cloud data that has been manually edited, construct an irregular triangulated network, set the network spacing, and interpolate to output DEM data. (5) Generation of digital elevation model products: completing edge processing with existing model data and metadata generation to form a digital elevation model.

[0019] 6.2) Generation of integrated 3D models above and below water (1) Underwater model generation The generated integrated underwater and above-water DEM model is cut according to the measured shoreline range on site to generate the underwater DEM model. The underwater DEM model data is opened in Globalmapper23, the water depth color gradient information is configured, and the data is stored in the *.FBX 3D model data format. Then, in the OSGBLab software, the *.FBX 3D model data format is converted into the *.OSGB 3D model data format. (2) Cutting of 3D models on water Cut out the model data within the water shoreline range from the generated water 3D model to generate a water 3D model (*.OSGB 3D model data format); (3) Fusion of above-water and underwater models Copy the above-water model partition tiles and underwater model tiles to a Data folder, and then make a general index file to generate an above-water and underwater integrated 3D model; 6.3) Digital Surface Model (DSM) Generation Combined with the generated above-water and underwater integrated 3D model, the digital surface model (DSM) is directly generated in DasViewer through the relevant module functions; 6.4) Digital Linear Graph (DLG) Generation CASS 3D supports loading and browsing of integrated above-water and underwater 3D models in the CASS environment, and supports direct acquisition of digital linear graph (DLG) data from 3D models; The digital linear map (DLG) of the water part mainly uses CASS 3D to load the inclined three-dimensional model, and collects various element data according to the "National Basic Scale Map Diagram Part 1: 1:500 1:1000 1:2000 Topographic Map Diagram" (GB / T 20257.1-2017); The underwater digital linear map (DLG) mainly uses the generated DEM model data to generate the navigable river contour (depth) lines; The above-water and underwater digital linear graph (DLG) data are merged to form an integrated above-water and underwater digital linear graph (DLG).

[0020] Beneficial effects of the present invention: Compared with the prior art, the effects of the present invention are as follows: (1) The present invention can fuse the constructed above-water data and underwater data to obtain an integrated above-water and underwater three-dimensional model, which is convenient for evaluation using the model and improves the evaluation accuracy and utilization rate; (2) The aerial survey of the present invention adopts a terrain-simulating flight method and multi-angle aerial photography to better ensure the consistency and overlap of resolution; supplementary photography and re-photography can effectively improve the accuracy and avoid the problem of being unable to solve or inaccurately solving due to omissions.

[0021] (3) The number of photos in the survey area is 20,000 to 30,000 photos per partition. There should be at least three overlapping routes between the partitions, and there should be common image control points, which is conducive to the edge connection of aerial triangulation. The completed aerial triangulation can be submitted for modeling to improve efficiency. Finally, all models are output and merged to create a general index file; match between flight strips. Match adjacent flight strips of laser point clouds to reduce the system error between flight strips.

[0022] (4) The present invention can greatly improve the measurement accuracy by calibrating the roll deviation; (5) DEM generation often involves a wide range of areas, a large amount of point cloud data, and high requirements for computer configuration. In addition, the workload of later human-computer interaction processing is large and involves many participants. Based on this, it is often used to divide the point cloud data into blocks using standard map sheets to facilitate later collaborative work; (6) In order to check the misclassification of point clouds in the surface details more carefully, the DOM or DLG data of the same period should be added to the point cloud triangulation network of the same area for terrain comparison and auxiliary judgment during the operation. Different terrains and objects should be identified piece by piece. The point cloud manual classification tool should be used to locally separate the misclassified or underclassified points into the corresponding point cloud categories according to the correct object type using an iterative algorithm to improve the accuracy of the final results of manual classification; (7) Rapid and accurate fusion of surface and underwater data; (8) During underwater measurement, the method of manned ship power support + unmanned ship water area measurement is adopted. The manned ship is mainly responsible for towing the unmanned ship equipment and logistics support, and the unmanned ship is responsible for the specific measurement. The specific mode is to tow the unmanned ship from the rapids channel to the upstream, and then conduct underwater measurement. At the same time, on the manned ship, through the unmanned ship remote control function, the unmanned ship is controlled to reach the shallow area on both sides for measurement, and it goes back and forth many times until the measurement range completely covers the water area. This measurement method solves the problem of insufficient horsepower, high energy consumption and short endurance of unmanned ships sailing against the current in inland rivers, and greatly improves work efficiency; (9) During underwater surveying and mapping, water level differences occur due to the influence of time differences. A series of problems such as data gaps and data overlaps may occur during aerial photography and water operations due to water level differences. In addition, the phenomenon of irregular water storage and flood discharge by the upstream hydropower station in the pilot area together constitutes the generation of measurement blank areas. It is possible to measure the water area when the river is flooded to obtain the river depth and elevation data to the greatest extent. During low water levels, drones equipped with airborne laser radar measurement systems and traditional engineering measurement methods are used to measure both sides of the river and some areas of the river shoal. Finally, the data is fused to ensure that the data fully covers the river. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart for underwater and underwater integration; Figure 2 This is a schematic diagram of flight altitude calculation; Figure 3 This is the underwater measurement flow chart of multi-beam echo sounder; Figure 4 Modeling a flow chart for ContextCapture Center; Figure 5 This is a flow chart of water point cloud data processing based on LiDAR; Figure 6 This is the flow chart of DLG data collection operation. DETAILED DESCRIPTION

[0024] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0025] Example 1: Figure 1-6 As shown, a method for constructing an integrated above-water and underwater three-dimensional model based on image processing, the method comprises the following steps: (1) Use drone-mounted photogrammetry systems to obtain aerial remote sensing image data with a resolution better than 0.05 m for 200 m on both sides of the navigable river and the coast; (2) Use a UAV equipped with a laser radar measurement system to obtain point cloud data of the navigable river and the 200-meter-wide water section on both sides of the coast; (3) Use an unmanned survey vessel equipped with a highly integrated high-frequency multi-beam echo sounder, GNSS receiver, integrated deck unit, and industrial computer to conduct underwater measurements and obtain data on the water depth and underwater topography of navigable rivers; The underwater terrain of the test area is the Wujiang waters, which are densely covered with reefs and have a large shallow area. The current is turbulent but the water level is shallow. The width of the navigable waters in the river is less than 20% of the width of the entire area. Based on the actual situation, it is necessary to cooperate with manned ships during measurement, and adopt the method of manned ship power guarantee + unmanned ship water area measurement. The manned ship is mainly responsible for towing the unmanned ship equipment and logistics support. The unmanned ship specifically conducts the measurement, dragging the unmanned ship from the rapids channel to the upstream, and then conducting underwater measurements. At the same time, on the manned ship, the unmanned ship remote control function is used to control the unmanned ship to reach the shallow areas on both sides for measurement, and go back and forth many times until the measurement range completely covers the water area; In the process of underwater surveying and mapping, water level differences occur due to the influence of time differences. A series of problems such as data gaps and data overlaps will occur in aerial photography and water operations due to water level differences. In addition, the phenomenon of irregular water storage and flood discharge by upstream hydropower stations in the pilot area together constitutes the generation of blank measurement areas. It is possible to measure the water area when the river is flooded to obtain the river depth and elevation data to the greatest extent. In low water levels, drones are used, equipped with airborne lidar measurement systems, and traditional engineering measurement methods are used to measure both sides of the river and some areas of the river shoals. The final data is fused to ensure that the data fully covers the river.

[0026] Similarly, when the river is at a low water level for a long time, manned ships cannot pass through, and unmanned ships are underpowered, drones equipped with airborne laser radar measurement systems or traditional engineering measurement methods are used to measure the river banks and some areas of the river shoals. When the river is at a high water level, the water area is measured to obtain the river depth and elevation data to the greatest extent. Finally, the data is fused to ensure that the data fully covers the river.

[0027] (4) Deploy a set number of image control points according to the regional network point layout requirements, and use GPS (RTK) combined with Beidou satellite navigation positioning GZCORS to measure image control points; (5) Use ContextCapture Center or Huace Copre software to perform aerial triangulation to generate a real-life 3D model of the above-water part and a true orthogonal image TDOM; (6) Using the Huace Copre software, the data collected by the airborne laser radar measurement system is solved by the base station and the point cloud color matching is performed to generate the above-water point cloud data; (7) Use CMS multi-beam post-processing software to process the data collected by the multi-beam echo sounder to generate navigable river depth and underwater topography data; (8) Use the airborne LiDAR point cloud data processing system to filter the integrated above-water and underwater point cloud data to generate a digital elevation model (DEM); (9) Use Global Mapper and OSGBLab Enterprise system to fuse the above-water real-life 3D model and the underwater digital elevation model (DEM) to generate an integrated above-water and underwater 3D model; (10) In DasViewer, use relevant functional modules to convert the integrated 3D model of water and water to generate a digital surface model (DSM); (11) In Southern CASS 3D, the generated integrated 3D model of water and water is used for data collection to generate a digital linear graph (DLG).

[0028] Furthermore, when drones are used for flight as mentioned above, aerial photography zoning is required. The aerial photography zoning method is: determine the reference surface elevation based on the minimum, maximum, median, mean and majority values ​​of the elevation of the DEM statistical photography area. When the height difference of the test area is too large to simultaneously meet the minimum overlap, flight safety and minimum ground resolution, aerial photography zones should be divided. The aerial photography zones are divided based on the terrain undulations and flight safety conditions of the test area to ensure that there are no mountains higher than the flight height within 1 km in front of the end of each zone route. After zoning, each zone determines the average reference surface elevation of the zone according to the aerial photography specifications and designs the route.

[0029] Furthermore, the relative altitude calculation formula of the drone is as follows: (1) Where: h - flight altitude; f —Focal length of the lens (35 mm); a—pixel size (3.9 μm); GSD — ground resolution; When planning the phased flight path, the GSD is set in combination with the camera and aircraft parameters to ensure that the ground resolution of the survey area is better than 5 cm. The relative altitude is calculated according to formula (1). For example, when the design GSD-ground resolution is 0.03 m, the relative altitude of the terrain simulation flight is 269 meters. The absolute altitude is the relative altitude plus the reference surface elevation.

[0030] Furthermore, the above-mentioned UAV aerial survey adopts a terrain-imitating flight mode and multi-angle aerial photography, with a heading overlap of ≥80%, a lateral overlap of ≥70%, and a route curvature of no more than 1%. When the route length is less than 5km, the maximum route curvature is no more than 3%.

[0031] Furthermore, the supplementary photography and re-photography during the above-mentioned drone aerial survey process include: 1) For absolute holes, relative holes and other serious defects that appear during aerial photography, retake photos of the corresponding positions of absolute holes, relative holes and other serious defects; 2) When the GNSS signal of the ground base station is lost, interrupted or other reasons cause the azimuth elements outside the scan line to be unable to be solved and the solution accuracy of the single-point positioning technology cannot meet the set requirements, supplementary or re-shooting should be carried out according to the size of the solution range; 3) When the IMU data recording is interrupted or the initial alignment flight is insufficient, the entire flight data cannot be correctly calculated; 4) The loophole re-shooting is carried out according to the original design track, and the length of the re-shooting route meets the requirements of the user's regional network density.

[0032] 5) For relative loopholes and local defects (such as clouds, cloud shadows, spots, etc.) that do not affect the encrypted point selection and model connection, additional photography is only required at the loopholes, and the length of the additional photography route exceeds the four baselines outside the loopholes.

[0033] Furthermore, the above underwater mapping steps: a) Survey line planning: The planning method is: 1) The survey line is planned based on the water depth and passability of the main rivers, and the strip width is three to four times the water depth; 2) Survey line planning in unobstructed areas is conducive to base station differential data transmission, and acquisition route planning should take into account the overlap rate of beam width; 3) After completing the survey area, select the location where the underwater terrain has a slope change as the correction line. The round-trip survey line on flat ground is used to correct the roll, the round-trip survey line on sloping ground is used to correct the pitch, and the parallel survey line on sloping ground is used to correct the heading; b) Multi-beam equipment assembly: The underwater probe of the multi-beam echo sounder is connected and fixed on the hull of the unmanned ship through the hull mounting plate, the three antennas are installed at the corresponding positions of the unmanned ship, and the batteries are placed in the corresponding battery slots and connected to power; c) Positioning base station installation: Choose an open area to set up the GNSS base station, use the China Measurement X10 receiver and DL5 external radio, set the correct engineering parameters and appropriate radio channels, and use known points to set up the station to avoid the inconvenience of parameter calibration caused by the small number of control points in the survey area; d) Launch and start-up inspection: Start the unmanned boat, check whether the power is normal, connect all sensors to the acquisition terminal, check whether the GNSS signal, attitude signal, and surface sound speed are normal; open the sonar acquisition software to check whether the echo image is normal; adjust the upper and lower thresholds, beam opening angle and working mode according to the operating environment, and turn on the roll and pitch stabilization; e) Multi-beam data acquisition: 1) Before the acquisition operation, the multi-beam echo sounder needs to be calibrated. Open the Norbit GUI multi-beam control software, follow the software prompts, activate and adjust the inertial navigation system, and sail along the figure-8 route upstream and downstream until the attitude accuracy meets the established requirements, usually 0.5°. After the calibration is completed, the formal acquisition work can be carried out. During this period, the GPS signal is guaranteed to be a fixed solution; 2) Use a ruler to measure the relative position of each hardware phase center to the center of the mounting flange and input it into the software. Set the opening angle, depth threshold, frequency and beam mode of the multi-beam echo sounder according to the actual situation of the survey area; 3) Open the CMS multi-beam data acquisition software, select the engineering parameters corresponding to the base station, add the Norbit iWBMS hardware (multi-beam echo sounder), edit the survey area matrix graphics, add the planned route, and set the display resolution and other parameters; 4) After starting to record data, the ship will sail along the planned route and correct the route in real time according to the sweep width until the entire survey area is scanned.

[0034] 5) Check whether the designed collection routes have been completed, and the routes that have been missed should be supplemented in time.

[0035] Furthermore, the above-mentioned UAV image control measurement method includes: The point distribution method is: (1) The layout range of control points should meet the mapping range of the project area, and all control points should be arranged at level and high points; (2) Adopt the corner point and inflection point layout method, that is, to arrange flat height points at the turning points of convex corners and concave corners of the regional network; (3) Each encryption unit should have 2-3 redundant control points as encryption check points. The check points should be selected in places with clear images and clear targets; (4) Since RTK technology is mainly used to determine the plane position of the flat high point, the feasibility and convenience of RTK measurement must also be considered when selecting the point. The selected point should be convenient for the placement of receiving equipment and operation, and the field of view should be wide; (5) The image control point target is selected at the intersection of the image and the ground object with an intersection angle between 30° and 150°; (6) Image control points should be selected from obvious features that are clear on adjacent images and meet the requirements for measuring plane position and elevation, intersection points of thin linear features that are nearly orthogonal, obvious corner points of features, or fixed features’ centers. Arc-shaped features, places with abrupt changes in local elevation, shadows, and features that are easily variable should not be selected. (7) Control points are arranged at the junction of aerial photography areas: Control points are arranged in the overlapping areas of adjacent flight routes. Adjacent areas should be shared as much as possible. If the sharing requirement cannot be met, points should be arranged separately. (8) The city level meets the 1:500 topographic mapping accuracy, the spacing between image control points is 400-500 meters, and each point evenly covers the survey area; The selection and trimming methods of image control points are as follows: (1) The targets of the image control points are selected from obvious objects with clear images, intersections of linear objects with good intersection angles, and corner vertices of obvious objects; (2) The main reference objects for selecting image control points include: the corners of flat-roofed houses, the top corners of fences, the top corners of gate piers, the top corners of gate covers, and other piers and columns that are convenient for measuring height and distance. The corners of gabled houses and those with irregular local elevation changes should not be used as puncture point targets; (3) For maps with fewer objects, some tiny landforms, such as the corners of ridges, small corners, and the center of small pits with a diameter of less than 0.2 mm on the map can be used. When using such objects, care should be taken to compare them with the actual field repeatedly to ensure that there is no change in the actual field after photography. (4) Select points and mark them on the digital image, accurately indicating the location of the selected points; (5) Check and organize the measurement results of image control points; (6) Prepare image control point results table and point information table.

[0036] The image control point measurement method is: The image control points are measured using network RTK, and the normal height of the image control points is obtained by using the elevation anomaly correction of the quasi-geoid refined model. The network RTK observation should meet the following requirements: (1) The instrument should be initialized before observation; (2) First, check the known points. The test results are obtained and the accuracy meets the requirements. The observation values ​​should be recorded after the RTK fixed solution is obtained and the convergence is stable. (3) Repeat the observation for no less than three times. When the difference between the measurements is less than 0.05 m, the average value is taken as the final result of the point.

[0037] Furthermore, the above-mentioned method for generating the above-mentioned 3D real-scene model on water is as follows: using 3D real-scene modeling software (ContextCapture Center software) to densely match the camera parameters, image data, and POS data with multi-view image feature points, and thereby perform a joint constraint adjustment solution of the free network multi-view image of the regional network, and establish a stereo model that can be appropriately freely deformed at a spatial scale to complete relative orientation; the measured photo control points are constrained to solve the existing regional network model, and the regional network is incorporated into an accurate geodetic coordinate system to complete absolute orientation. Fully automatic 3D modeling is performed using a 3D real-scene modeling software (ContextCapture Center software) with multi-machine and multi-node parallel computing. The modeling process is as follows Figure 4That is, the aerial triangulation data is directly submitted to generate 3D TIN grid construction, white body 3D model creation, self-service texture mapping and 3D scene construction. In principle, the model modification only repairs the water area gaps or model holes. The model is regenerated after constraint intervention using 3D real scene modeling software (ContextCapture Center software) and water surface or supplementary flight data to ensure that there are no loopholes in the model. (1) The number of photos in the survey area is 20,000 to 30,000 photos per partition. There must be at least three overlapping routes between the partitions. There must be common image control points to facilitate aerial triangulation. After the aerial triangulation encrypted partitions are divided, they are divided according to the flight data and image control conditions, which may not be consistent with the aerial photography partitions; (2) Each encrypted partition of the triangulation is modeled separately. Each partition must have a unified origin (coordinate origin and tile origin) and size. Finally, merge the triangulation and select only the edge tiles to output the model. Then copy the edge tiles of each encrypted partition to a Data folder. After the edge tiles are copied to the Data file, replace the files with the same name, and then create a total index file.

[0038] Furthermore, the above-mentioned water point cloud data processing method has the following steps: (1) Data acquisition: Generate mission areas based on LiDAR data and obtain high-precision LiDAR point cloud data; (2) Trajectory solution: the original real-time GNSS / IMU trajectory data and CORS station data are jointly solved to obtain accurate trajectory information; (3) Laser point cloud solution, combined with accurate trajectory information, to obtain laser point cloud data in LAS format; (4) Noise removal: performing operations such as gross error elimination on laser point cloud data to remove erroneous points and abnormal elevation points; (5) Matching between flight strips: matching adjacent flight strips of laser point clouds to reduce the system error between flight strips; (6) Point cloud output: finally, the processed point cloud data of the above-water part is output.

[0039] Furthermore, the above underwater data processing method comprises the following steps: (1) Installation deviation calibration First, calibrate the roll deviation: select a pair of round-trip survey lines with flat underwater terrain, select a calibration line in the direction perpendicular to the route at the overlapped part of the survey lines, adjust the fitting step size and step, calculate the roll deviation through the software, exit the interface, reload the route file, and fill in the roll deviation value at the equipment deviation. Then calibrate the pitch deviation: select a pair of round-trip survey lines with a sloped underwater terrain, select a calibration line in the direction of the route at the overlapped part of the survey lines, and record the pitch deviation value; exit the software, reload the file, and enter all the roll and pitch deviations. Finally, calibrate the heading deviation: select a pair of parallel survey lines with a sloped underwater terrain, select a calibration line in the direction of the route at the overlapped part of the survey lines, record the heading deviation, reload all files, and enter the recorded installation deviation; (2) First-level editor Use CMS software to post-process the collected raw data. First, perform primary editing, open the raw data file, select the survey line to be processed, edit the view according to the plane, and perform cropping and noise editing as needed; (3) Secondary Editor Select the survey line that has been edited in the first level to enter the second level editing, mainly to delete the noise and flying points generated by the multi-beam edge clutter, and perform detailed processing frame by frame and line by line; (4) Level 3 Editor After the secondary editing, the data enters the tertiary editing mode, where the noise and flying spots generated by the multi-beam edge clutter are deleted in the 3D editing mode; (5) Data output The processed data file is compressed and saved in a specified format, usually .las format.

[0040] Furthermore, the above-water and underwater data fusion method has the following steps: 6.1) The steps of generating the integrated digital elevation model (DEM) above and below water are as follows: After merging the underwater data (las format) scanned by the multi-beam unmanned ship with the pre-processed surface point cloud data, the data was imported into the point cloud data processing software for point cloud data processing. Through point cloud segmentation, point cloud filtering, manual fine editing, TIN construction, and interpolation, a DEM model was generated. (1) Point cloud segmentation DEM generation often involves a wide range of areas, a huge amount of point cloud data, high requirements on computer configuration, and a large workload of later human-computer interaction and many people involved. Based on this, point cloud data is divided into blocks using standard map sheets to facilitate later collaborative work; (2) Point cloud filtering: using algorithms or algorithm combinations based on emission intensity, echo number, ground shape, etc. to filter laser point cloud data, classify data of objects such as buildings, vegetation, and power facilities into non-ground point data, extract surface data, and separate non-ground points; (3) Manual fine editing: Use laser point cloud classification software to construct TIN inspection for automatically classified data and correct misclassified data. Usually, the software automatic filtering results are retrieved in the point cloud data processing software to generate a surface triangulated network. The elevation information of the point cloud is checked by pulling sections of the triangulated network in different areas to check whether there are discontinuous, uneven, or leaky objects or terrain in the area, or whether there are large elevation differences of the same objects (i.e., obvious "pits" or "protruding points"), so as to determine whether the point cloud here is misclassified or underclassified. In order to check the misclassification of the point cloud in the surface details more carefully, the DOM or DLG data of the same period should be added to the point cloud triangulated network of the same area for terrain comparison and auxiliary judgment during the operation. Different terrains and objects are identified piece by piece. The point cloud manual classification tool is used to locally separate the misclassified or underclassified points into the corresponding point cloud categories according to the correct object type using an iterative algorithm to improve the accuracy of the final results of manual classification. During project generation, the main manual fine editing involved includes manual classification, manual point addition, addition of fracture lines, regional smoothing, foundation setting and other related operations.

[0041] 1) Small-area buildings and structures are easily misclassified as ground points. When manually classifying, these point clouds should be separated from the ground points and placed in the building point layer; 2) Due to vegetation coverage, there is a lack of point cloud data in some small areas. The real landform can be restored by adding appropriate points; 3) The ground points on the steep slopes may be misjudged as other land points (such as low vegetation), resulting in a lack of ground points and discontinuous terrain. This part needs to be manually filled in, usually by adding breaklines; 4) The noise points left after vegetation filtering and the temporarily piled garbage on the road surface may cause the ground to be uneven (i.e. obvious "pits" or "protruding points"). This part needs to be smoothed by the regional smoothing tool; 5) For houses built on the mountain or against the mountain, there is a certain height difference between the foundation of the house and the slope or steep slope behind the house. At the same time, such houses are often blocked by trees, resulting in no clear boundary ground points around the house. If the actual point cloud shows different height differences around the house, the foundation surface of the house will be a slope after the DEM is generated, which obviously does not conform to the actual situation of the house. The solution is: when the height difference of the ground points around the foundation of such houses is large, add a breakline to level the foundation. The elevation of the breakline is based on the correct foundation elevation around the house; (4) Construct TIN and interpolate to generate DEM. For the point cloud data that has been manually edited, construct an irregular triangulated network, set the network spacing, and interpolate to output DEM data. (5) Generation of digital elevation model products: completing edge processing with existing model data and metadata generation to form a digital elevation model.

[0042] 6.2) Generation of integrated 3D models above and below water (1) Underwater model generation The generated integrated underwater and above-water DEM model is cut according to the measured shoreline range on site to generate the underwater DEM model. The underwater DEM model data is opened in Globalmapper23, the water depth color gradient information is configured, and the data is stored in the *.FBX 3D model data format. Then, in the OSGBLab software, the *.FBX 3D model data format is converted into the *.OSGB 3D model data format. (2) Cutting of 3D models on water Cut out the model data within the water shoreline range from the generated water 3D model to generate a water 3D model (*.OSGB 3D model data format); (3) Fusion of above-water and underwater models Copy the above-water model partition tiles and underwater model tiles to a Data folder, and then make a general index file to generate an above-water and underwater integrated 3D model; 6.3) Digital Surface Model (DSM) Generation Combined with the generated above-water and underwater integrated 3D model, the digital surface model (DSM) is directly generated in DasViewer through the relevant module functions; 6.4) Digital Linear Graph (DLG) Generation CASS 3D supports loading and browsing of integrated above-water and underwater 3D models in the CASS environment, and supports direct acquisition of digital linear graph (DLG) data from 3D models; The digital linear map (DLG) of the water part mainly uses CASS 3D to load the inclined three-dimensional model, and collects various element data according to the "National Basic Scale Map Diagram Part 1: 1:500 1:1000 1:2000 Topographic Map Diagram" (GB / T 20257.1-2017); The underwater digital linear map (DLG) mainly uses the generated DEM model data to generate the navigable river contour (depth) lines; The above-water and underwater digital linear graph (DLG) data are merged to form an integrated above-water and underwater digital linear graph (DLG).

[0043] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for constructing an integrated above-water and underwater three-dimensional model based on image processing, characterized in that: The method comprises the following steps: (1) Use drone-mounted photogrammetry systems to obtain aerial remote sensing image data with a resolution better than 0.05 m for 200 m on both sides of the navigable river and the coast; (2) Use a UAV equipped with a laser radar measurement system to obtain point cloud data of the navigable river and the 200-meter-high water section on both sides of the coast; (3) Use an unmanned survey vessel equipped with a device that integrates a high-frequency multi-beam echo sounder, a GNSS receiver, an integrated deck unit, and an industrial computer to conduct underwater measurements and obtain data on the water depth and underwater topography of navigable rivers; During the measurement, it is necessary to cooperate with manned ships, and adopt the method of manned ship power support + unmanned ship water area measurement. The manned ship is mainly responsible for towing the unmanned ship equipment and logistics support. The unmanned ship is specifically responsible for the measurement. The unmanned ship is towed from the rapids channel to the upstream, and then underwater measurement is carried out. At the same time, the manned ship uses the unmanned ship remote control function to control the unmanned ship to reach the shallow areas on both sides for measurement, and go back and forth many times until the measurement range completely covers the water area; Choose to measure the water area when the river is at flood level to obtain the river depth and elevation data to the greatest extent possible. Use drones equipped with airborne laser radar measurement systems to measure both sides of the river and some areas of the river shoal during low water levels, and finally fuse the data; (4) Deploy the set number of image control points according to the regional network point layout requirements, and use GPS to measure the image control points in combination with the Beidou satellite navigation and positioning base station in the area to be measured; (5) Carry out aerial triangulation to generate a real-life 3D model of the above-water part and a true orthogonal image TDOM; (6) The data collected by the airborne laser radar measurement system is solved by the base station and the point cloud color matching is performed to generate the above-water point cloud data; (7) Processing the data collected by the multi-beam echo sounder to generate data on the water depth and underwater topography of navigable rivers; (8) Filter the integrated above-water and underwater point cloud data to generate a digital elevation model; (9) Perform image model fusion on the above-water real-scene 3D model and the underwater digital elevation model to generate an above-water and underwater integrated 3D model; (10) Convert the integrated above-water and underwater three-dimensional model to generate a digital surface model; (11) Use the generated integrated three-dimensional model of water and water to collect data and generate a digital linear diagram.

2. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 1, characterized in that: When using drones for flight, aerial photography zoning is required. The aerial photography zoning method is: determine the reference surface elevation based on the minimum, maximum, median, mean and majority values ​​of the statistical photography area. When the height difference in the test area is too large to simultaneously meet the minimum overlap, flight safety and minimum ground resolution, the aerial photography zones are divided. The aerial photography zones are divided based on the terrain undulations and flight safety conditions of the test area to ensure that there are no mountains higher than the flight height within 1 km in front of the end of each zone route. After zoning, each zone determines the average reference surface elevation of the zone according to the aerial photography specifications and designs the route.

3. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 2, characterized in that: The relative altitude calculation formula of the drone is: （1） Where: h - flight altitude; f —Focal length of the lens; a—pixel size; GSD — ground resolution; When planning the phased flight path, the ground resolution GSD is set in combination with the camera and aircraft parameters to ensure that the ground resolution of the survey area is better than 5 cm. The relative altitude is calculated according to formula (1), and the absolute altitude is the relative altitude plus the reference surface elevation.

4. A method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to any one of claims 1 to 3, characterized in that: The drone aerial survey adopts the terrain-simulating flight mode and multi-angle aerial photography, with the heading overlap ≥80%, the lateral overlap ≥70%, and the route curvature not exceeding 1%. When the route length is less than 5km, the maximum route curvature is not more than 3%; The supplementary and re-shooting during the UAV aerial survey process include: 1) For absolute holes, relative holes and other serious defects that appear during aerial photography, retake photos of the corresponding positions of absolute holes, relative holes and other serious defects; 2) When the GNSS signal of the ground base station is lost, interrupted or other reasons cause the azimuth elements outside the scan line to be unable to be solved and the solution accuracy of the single-point positioning technology cannot meet the set requirements, supplementary or re-shooting should be carried out according to the size of the solution range; 3) When the IMU data recording is interrupted or the initial alignment flight is insufficient, the entire flight data cannot be correctly calculated; 4) The loophole re-shooting is carried out according to the original design track, and the length of the re-shooting route meets the requirements of the user's regional network density distribution; 5) For relative loopholes and local defects that do not affect the encrypted point selection and model connection, additional photography is only conducted at the loopholes, and the length of the additional photography route exceeds the four baselines outside the loopholes.

5. A method for constructing an above-water and underwater integrated three-dimensional model based on image processing according to any one of claims 1 to 3, characterized in that: Underwater mapping steps: a) Survey line planning, the planning method is: 1) The survey line is planned based on the water depth and passability of the main rivers, and the strip width is three to four times the water depth; 2) The survey line is planned in an unobstructed area, and the acquisition route planning should take into account the overlap rate of the beam width; 3) After completing the survey area, select the location where the underwater terrain has a slope change as the correction line. The round-trip survey line on flat ground is used to correct the roll, the round-trip survey line on sloping ground is used to correct the pitch, and the parallel survey line on sloping ground is used to correct the heading; b) Multi-beam equipment assembly: The underwater probe of the multi-beam echo sounder is connected and fixed on the hull of the unmanned ship through the hull mounting plate, the three antennas are installed at the corresponding positions of the unmanned ship, and the batteries are placed in the corresponding battery slots and connected to power; c) Positioning base station installation: Choose an open area to install the GNSS base station, use the Huace receiver and external radio, set the correct engineering parameters and appropriate radio channels, and use known points to install the base station; d) Launch and start-up inspection: Start the unmanned boat, check whether the power is normal, connect all sensors to the acquisition terminal, check whether the GNSS signal, attitude signal, and surface sound speed are normal; open the sonar acquisition software to check whether the echo image is normal; adjust the upper and lower thresholds, beam opening angle and working mode according to the operating environment, and turn on the roll and pitch stabilization; e) Multi-beam data acquisition: 1) Before the acquisition operation, the multi-beam echo sounder needs to be calibrated. Open the multi-beam control software, activate and adjust the inertial navigation system according to the software prompts, and sail along the figure-8 route upstream and downstream until the attitude accuracy meets the established requirements, usually 0.5°. After the calibration is completed, the formal acquisition work will be carried out, and the GPS signal will be guaranteed to be a fixed solution during this period; 2) Use a ruler to measure the relative position of each hardware phase center to the center of the mounting flange, input the data into the software, and set the opening angle, depth threshold, frequency and beam mode of the multi-beam echo sounder according to the actual situation of the survey area; 3) Open the CMS multi-beam data acquisition software, select the engineering parameters corresponding to the base station, add the multi-beam echo sounder hardware, edit the survey area matrix graphics, add the planned route, and set the display resolution; 4) After starting to record data, the ship will sail along the planned route and correct the route in real time according to the sweep width until the entire survey area is scanned; 5) Check whether the designed collection routes have been completed, and the routes that have been missed should be supplemented in time.

6. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 5, characterized in that: UAV image control measurement methods include: The point distribution method is: (1) The layout range of control points should meet the mapping range of the project area, and all control points should be arranged at level and high points; (2) Adopt the corner point and inflection point layout method, that is, to arrange flat height points at the turning points of convex corners and concave corners of the regional network; (3) Each encryption unit should have 2-3 redundant control points as encryption check points. The check points should be selected in places with clear images and clear targets; (4) The selected location should be convenient for placement of receiving equipment and operation, and the field of view should be wide; (5) The image control point target is selected at the intersection of the image and the ground object with an intersection angle between 30° and 150°; (6) The image control points are selected from obvious ground objects, intersection points of thin and nearly orthogonal linear ground objects, obvious ground object corner points or fixed ground object centers, which are clear in the images of adjacent images and meet the requirements for measuring plane position and elevation; (7) Control points are arranged at the junction of aerial photography areas: Control points are arranged in the overlapping areas of adjacent flight routes. Adjacent areas should be shared as much as possible. If the sharing requirement cannot be met, points should be arranged separately. (8) The city level meets the 1:500 topographic mapping accuracy, the spacing between image control points is 400-500 meters, and each point evenly covers the survey area; The selection and trimming methods of image control points are as follows: (1) The targets of the image control points are selected from obvious objects with clear images, intersections of linear objects with good intersection angles, and corner vertices of obvious objects; (2) The main reference objects for selecting image control points include: the corners of flat-roofed houses, the top corners of walls, the top corners of gate piers, the top corners of gate covers, and the top corners of piers and columns for measuring height and distance; (3) For maps with few objects, some tiny landforms, such as the corners of ridges, small corners, and the center of small pits with a diameter of less than 0.2 mm on the map, when using such objects, care should be taken to compare them with the actual field repeatedly to ensure that there is no change in the actual field after photography. (4) Select points and mark them on the digital image, accurately indicating the location of the selected points; (5) Check and organize the measurement results of image control points; (6) Prepare image control point result table and point information table; The image control point measurement method is: The image control points are measured using network RTK, and the normal height of the image control points is obtained by using the elevation anomaly correction of the quasi-geoid refined model. The following conditions should be met when using network RTK observation: (1) The instrument should be initialized before observation; (2) First, check the known points. The test results are obtained and the accuracy meets the requirements. The observation values ​​should be recorded after the RTK fixed solution is obtained and the convergence is stable. (3) Repeat the observation for no less than three times. When the difference between the measurements is less than 0.05 m, the average value is taken as the final result of the point.

7. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 5, characterized in that: The method for generating a 3D real-scene model on water is as follows: using 3D real-scene modeling software, densely matching the camera parameters, image data and POS data with multi-view image feature points, and using this to perform joint constraint adjustment and solution of the free network multi-view image of the regional network, and establishing a stereo model that can be moderately freely deformed on a spatial scale to complete relative orientation; using the measured photo control points to perform constraint adjustment and solution on the existing regional network model, and incorporating the regional network into the geodetic coordinate system to complete absolute orientation; fully automatic 3D modeling is performed using 3D real-scene modeling software with multi-machine and multi-node parallel computing, and the result data after aerial triangulation is directly submitted to generate 3D TIN grid construction, white body 3D model creation, self-service texture mapping and 3D scene construction. In principle, model modification only repairs water vacancies or model loopholes, and the model is regenerated after constraint intervention using 3D real-scene modeling software and water surface or supplementary flight data. The specific operations are as follows: (1) The number of photos in the survey area is 20,000 to 30,000 photos per zone, and at least three flight routes overlap between zones, and public image control points are set; (2) Each encrypted partition of the three-dimensional model is demolded separately. Each partition must have a unified origin and size. Finally, the three-dimensional model is merged and only the edge tiles are selected to output the model. Then the edge tiles of each encrypted partition are copied to a Data folder. The edge tiles can be copied to the Data file to replace the files with the same name, and then a total index file is created.

8. The testing method of the method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 7, characterized in that: The steps of water point cloud data processing method are as follows: (1) Data acquisition: Generate mission areas based on LiDAR data and obtain high-precision LiDAR point cloud data; (2) Trajectory calculation: The trajectory information is obtained by jointly calculating the original real-time GNSS / IMU trajectory data and the Beidou satellite navigation and positioning base station (CORS station) data in the test area; (3) Laser point cloud solution, combined with trajectory information, to obtain laser point cloud data in LAS format; (4) Noise removal: Perform gross error elimination on the laser point cloud data to remove erroneous points and abnormal elevation points; (5) Matching between flight strips: matching adjacent flight strips of laser point clouds; (6) Point cloud output: finally, the processed point cloud data of the above-water part is output.

9. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 8, characterized in that: The steps of underwater data processing method are as follows: (1) Installation deviation calibration First, calibrate the roll deviation: select a pair of round-trip survey lines with flat underwater terrain, select a calibration line in the overlapped part of the survey lines in the direction perpendicular to the route, adjust the fitting step size and step, calculate the roll deviation through the software, exit the interface, reload the route file, and fill in the roll deviation value at the device deviation; Then, calibrate the pitch deviation: select a pair of round-trip survey lines with a slope in the underwater terrain, select a calibration line along the route direction at the overlapping part of the survey lines, and record the pitch deviation value; exit the software, reload the file, and enter all the roll and pitch deviations; finally, calibrate the heading deviation: select a pair of parallel survey lines with a slope in the underwater terrain, select a calibration line along the route direction at the overlapping part of the survey lines, record the heading deviation, reload all files, and enter the recorded installation deviation; (2) First-level editor Use CMS software to post-process the collected raw data, perform primary editing, open the raw data file, select the survey line to be processed, edit the view according to the plane, and perform cropping and noise editing as needed; (3) Secondary Editor Select the survey line that has been edited in the first level to enter the second level editing, mainly to delete the noise and flying points generated by the multi-beam edge clutter, and perform detailed processing frame by frame and line by line; (4) Level 3 Editor After the secondary editing, the data enters the tertiary editing mode, where the noise and flying spots generated by the multi-beam edge clutter are deleted in the 3D editing mode; (5) Data output Compress the data file processed in step (4) and save the processed corrected data file in a specified format, selecting the .las format.

10. The method for constructing an integrated above-water and underwater three-dimensional model based on image processing according to claim 9, characterized in that: The steps of the above-water and underwater data fusion method are as follows: 6.1) The steps of generating the integrated digital elevation model (DEM) above and below water are as follows: After merging the underwater data in las. format scanned by the multi-beam unmanned ship with the pre-processed surface point cloud data, the data was imported into the point cloud data processing software for point cloud data processing. Through point cloud segmentation, point cloud filtering, manual fine editing, TIN construction, DEM model was generated by interpolation. (1) Point cloud segmentation Use standard map frames to divide point cloud data into blocks; (2) Point cloud filtering: using an algorithm or combination of algorithms based on emission intensity, echo number, and ground shape to filter the laser point cloud data, classify data on objects such as buildings, vegetation, and power facilities into non-ground point data, extract surface data, and separate non-ground points; (3) Manual fine editing: Use laser point cloud classification software to construct TIN inspection for automatically classified data and correct misclassified data. Usually, the software automatic filtering results are retrieved in the point cloud data processing software to generate a surface triangulated network. The elevation information of the point cloud is checked by pulling sections of the triangulated network in different areas to check whether there are discontinuous, uneven, or leaky objects or terrain in the area, or whether there are large height differences between the same objects. This is to determine whether the point cloud here is misclassified or underclassified. During the operation, DOM or DLG data from the same period should be added to the point cloud triangulated network of the same area for terrain comparison and auxiliary judgment. Different terrains and objects are identified piece by piece. The point cloud manual classification tool is used to locally separate the misclassified or underclassified points into the corresponding point cloud categories according to the correct object type using an iterative algorithm; 1) Small-area buildings and structures are easily misclassified as ground points. During manual classification, these point clouds should be separated from the ground points and placed in the building point layer; 2) When a small area is covered by vegetation, the real landform is restored by adding points; 3) When the ground points of the steep slope line are misjudged as other land points, they can be manually filled in and processed by adding break lines; 4) When there are noise points and temporary garbage piled up on the road surface after vegetation filtering, this part should be smoothed by the regional smoothing tool; 5) For houses built on the mountain or against the mountain, the treatment method is: when the ground point height difference around the foundation of such houses is large, add a breakline to level the foundation. The elevation of the breakline is based on the correct foundation elevation around the house; (4) Construct TIN and interpolate to generate DEM model. For the point cloud data that has been manually edited, construct irregular triangulated network, set the network spacing, and interpolate to output DEM model data. (5) Generate digital elevation model products, complete edge processing with existing model data and generate metadata to form a digital elevation model; 6.2) Generation of integrated 3D models above and below water (1) Underwater model generation The generated integrated underwater and above-water DEM model is cut according to the measured shoreline range on site to generate the underwater DEM model. The underwater DEM model data is opened in the Global mapper23 software, the water depth color gradient information is configured, and the data is stored in the *.FBX 3D model data format. Then, in the OSGBLab software, the *.FBX 3D model data format is converted into the *.OSGB 3D model data format. (2) Cutting of 3D models on water The generated water 3D model results are cut off from the model data within the water shoreline range to generate a water 3D model in the format of *.OSGB 3D model data format; (3) Fusion of above-water and underwater models Copy the above-water model partition tiles and underwater model tiles to a Data folder, and then make a general index file to generate an above-water and underwater integrated 3D model; 6.3) Digital surface model generation Combined with the generated above-water and underwater integrated 3D model, the digital surface model (DSM) is directly generated in the DasViewer software through the relevant module functions; 6.4) Digital Linear Graph Generation Digital linear map of the water part, using CASS 3D software to load the inclined 3D model, according to the "National Basic Scale Map Diagram Part 1: 1:500 1:1000 1:2000 Topographic Map Diagram" (GB / T 20257.1-2017), collect various element data; Digital linear map of underwater part, using the generated DEM model data to generate navigable river contour lines / depth lines; The digital linear map data of the above-water and underwater parts are merged to form an integrated above-water and underwater digital linear map.

Images