A high-density navigation-obstruction-area water-depth measuring system and method

Abstract

The application discloses a high-density navigation-obstructing-area water depth measuring system, which comprises a UAV platform, a connecting and fixing device, a water depth measuring module and a water depth measuring industrial computer. The UAV platform is used for providing a carrying platform, power support, GNSS data, attitude data and a field-of-view photo in front of the flight path, and is also used for adjusting the flight path according to the identification result. The connecting and fixing device is used for connecting and fixing the UAV platform and the water depth measuring module. The water depth measuring module is used for real-time measurement of water depth data to obtain instantaneous water depth data. The water depth measuring industrial computer is used for calculating a final water depth result according to the GNSS data, the attitude data and the instantaneous water depth data. The edge computing box is used for inferences according to the water surface image in front of the flight path to output an identification result. The application also discloses a high-density navigation-obstructing-area water depth measuring method. The application drives the water depth measurement by using a multi-rotor UAV, and adopts a camera monitoring intelligent obstacle identification and obstacle crossing system design to fundamentally avoid the possibility of being disturbed by water navigation-obstructing objects.

Description

Technical Field

[0001] This invention belongs to the field of water depth measurement technology, specifically a water depth measurement system and method for high-density navigation obstruction areas. Background Technology

[0002] In current UAV-driven depth measurement technology, engineers have already combined UAVs with depth measurement. Among the published technologies, patent CN202311210957 discloses a depth measurement device including a UAV and a measuring cylinder. This invention mainly addresses the problem that when ripples appear in the sounding water area due to wind, the buoy slides up and down under the guidance of a mounting rod, causing the echo sounder to move up and down, resulting in changes in the height of the echo sounder and inaccurate measurement data. The aforementioned device uses a simple UAV, primarily serving as a limiting device for a split-type depth sounder, without detailing the layout and function of each module. Patent CN202311793483 discloses a UAV-tractioned depth measurement method, device, equipment, and readable storage medium, including a UAV carrier platform, a traction unit, and a measuring unit; the UAV carrier platform and the measuring unit are connected via the traction unit. This invention achieves depth measurement in shallow water areas or complex underwater environments at a lower cost by using a UAV to traction the depth measurement unit. The aforementioned patents employ a traction device to connect the UAV and the measurement unit, primarily focusing on using photography to locate the depth sounding device. Patent CN202411241157 discloses a depth measurement device and method based on UAV technology, including a sounding hammer, measuring rope, buoy, swing damper, depth control module, and UAV platform. This invention mainly introduces UAV technology to construct a UAV-suspended sounding hammer depth measurement system, achieving efficient, accurate, and safe depth measurement, with advantages such as simple operation, high measurement accuracy, and strong adaptability. CN202411501167 discloses a novel towed depth measurement device, including a UAV body and a towed depth measurement device; this invention, through the integrated combination of the UAV and the measurement device, compared to traditional unmanned surface vessel (USV) depth measurement, features high integration, modularity, accurate measurement, portability, and lightweight design, reducing fieldwork workload. However, this invention uses a UAV to towed depth measurement device, and the device still carries the risk of being snagged on navigational obstructions. The inventions mentioned above all use quadcopter drones combined with water depth measurement, but they still have problems such as low equipment integration, weak mobility, and low level of intelligence. They do not clearly explain the specific implementation technical route of drones combined with the depth measurement system.

[0003] Therefore, a water depth measurement system and method for high-density navigation obstruction areas are provided. Summary of the Invention

[0004] To address the aforementioned problems in the existing technology, this invention provides a water depth measurement system and method for high-density navigational obstructions. The system employs a multi-rotor UAV to drive water depth measurement and utilizes a camera-based intelligent obstacle identification and obstacle-crossing system to fundamentally avoid the possibility of interference from waterborne navigational obstructions.

[0005] The technical solution to achieve the above objectives is:

[0006] One of the present inventions provides a water depth measurement system for high-density navigation obstruction areas, comprising:

[0007] The drone platform is used to provide a mounting platform, power support, GNSS (Global Navigation Satellite System) data, attitude data, and field-of-view photos of the flight path ahead. It is also used to adjust the flight path based on the identification results.

[0008] Connection and fixing device, used to connect and fix the UAV platform and the water depth measurement module;

[0009] The water depth measurement module is used for real-time measurement of water depth data to obtain instantaneous water depth data;

[0010] The water depth measurement industrial control computer is used to calculate the final water depth result based on GNSS data, attitude data, and instantaneous water depth data;

[0011] The edge computing box is used to perform inference based on the water surface image in front of the flight, output the recognition result, and determine one of three cases based on the recognition result: easy ski-jump, difficult ski-jump, or cannot be determined.

[0012] Preferably, the drone platform includes:

[0013] The drone itself is used to carry GNSS positioning components, attitude components, cameras, connection bases and equipment mounting boxes, and also to adjust the flight path based on the identification results.

[0014] GNSS positioning components are used for real-time positioning and orientation during UAV flight and water depth measurement.

[0015] Attitude component is used to measure the roll and pitch attitude changes during UAV flight, providing attitude information for three-dimensional water depth calculation;

[0016] Cameras are used to view the water surface at different distances, both near and far, to monitor the safety status of the water surface and acquire water surface images;

[0017] Connecting base, used to secure the connecting device with screws;

[0018] Equipment mounting box, used to install the water depth measurement industrial control computer and edge computing box;

[0019] The GNSS positioning component is a dual-antenna GNSS positioning device. The main antenna is located directly above the attitude component and is used for positioning and orientation in conjunction with the secondary antenna. The antenna closer to the camera is the secondary antenna, which is mainly used to assist in orientation.

[0020] Preferably, the connecting and fixing device includes: a connecting structure frame, which is composed of a rectangular frame and a triangular frame, with one side of the rectangular frame fixedly connected to the short side of the triangular frame and forming a certain angle;

[0021] The upper rectangular frame of the connecting structure is provided with screw holes, which are connected and fixed to the connecting base by screws.

[0022] A spring retaining ring is set on each side of the rectangular frame of the connecting structure away from the fixed triangular frame.

[0023] A pull wire gauge is installed at the middle position on the same side as the rectangular frame of the connecting structure and the spring fixing ring;

[0024] One end of a fixing spring is fixed to each spring fixing ring, and the other end of the fixing spring is fixed to the water depth measurement module;

[0025] A rotation axis is provided at the top of the triangular frame of the connecting structure, away from the top of the rectangular frame;

[0026] in,

[0027] The spring retaining ring has a semi-circular structure;

[0028] The draw gauge is a small-range, high-precision draw gauge with a range of less than 1 meter, used for real-time monitoring of tension when a fixed spring is stretched.

[0029] The tops of the two sides of the triangular frame of the connecting structure are fixed to the two sides of the rotation axis.

[0030] Preferably, the water depth measurement module includes: an arc-shaped depth sounding rod, which adopts an arc-shaped design;

[0031] A locking ring is provided on the inner side of the arc-shaped depth measuring rod near the connecting and fixing device, and the other ends of the two fixing springs are fixed on the locking ring;

[0032] The inner side of the arc-shaped sounding rod is slidably connected to a circular bearing, and the rotating shaft is fixed inside the circular bearing;

[0033] A level gauge and a depth transducer are installed at the end of the arc-shaped depth sounding rod away from the connecting and fixing device;

[0034] in,

[0035] The circular bearing is a ball bearing;

[0036] The level gauge is a water pressure sensor used to measure the draft of the depth sounding transducer in real time;

[0037] The sounding transducer is a single-beam sounding transducer, used to emit sounding waves into the water bottom, and its bottom surface is flush with the bottom surface of the curved sounding rod.

[0038] The reference centers of the GNSS positioning unit, attitude unit, level gauge, and depth sounding transducer are set on the same vertical line.

[0039] Preferably, in the edge computing box, the recognition results are divided into three situations: easy to slide, difficult to slide, and impossible to determine. The dynamic window method is used to carry out path replanning according to the three situations. The path replanning determines whether to move slowly or forward and upward. An iterative dataset is established for the video data that cannot be determined or is missed, which is used for iterative training and optimization of the AI ​​recognition model. The AI ​​recognition model is selected from the YOLO series.

[0040] A second aspect of the present invention provides a method for measuring water depth in high-density navigation obstruction areas, comprising:

[0041] Step S1: Before the water depth measurement operation, after the equipment is assembled on the shore, the multi-rotor UAV carrying the equipment flies to the designated water area to carry out the operation.

[0042] Step S2: The camera acquires real-time field-of-view photos of the flight path ahead, focusing on capturing photos of the water surface ahead. The AI ​​recognition model infers the water surface images and outputs the recognition results. Based on the recognition results, the camera determines three situations: relatively easy to skid over, relatively difficult to skid over, and cannot be determined.

[0043] Step S3: The judgment is sent to the main control system of the UAV and the path replanning is carried out using the dynamic window method. Based on the path replanning, it is decided whether to move slowly or forward and upward.

[0044] Step S4: During the water depth measurement process, a safe draft line is set according to the basic information of the measurement area. The level gauge and the depth sounding transducer below maintain a stable draft. The dynamic draft is measured using the level gauge to obtain instantaneous water depth data.

[0045] Step S5: Calculate the absolute coordinate system direction using the GNSS positioning component to obtain GNSS data, and measure the roll and pitch attitude changes of the UAV during flight using the attitude component to obtain attitude data.

[0046] In step S6, the water depth measurement industrial control computer calculates the final water depth result based on GNSS data, attitude data, and instantaneous water depth data.

[0047] Preferably, in step S2, the AI ​​recognition model performs inference on the water surface image and outputs the recognition result. Based on the recognition result, it is determined into three situations: relatively easy to slide across, relatively difficult to slide across, and impossible to determine.

[0048] Step 1, Sample collection and preprocessing:

[0049] Image data of aquatic plants, fishing nets, trees, cables, plastic bags, and floating objects on the water were collected through actual collection and online search. Artificial simulation samples were added to increase the richness of the samples based on common navigational obstacles in the survey area. The collection of samples took into account factors such as different water areas, different weather, different time periods, and different angles.

[0050] Data is cleaned by removing blurry, dark, and overexposed invalid images; the original images are rotated, scaled, cropped, color-adjusted, and noise-added to expand the dataset size; the image sizes are uniformly adjusted to the fixed size of the model input and normalized to accelerate model training convergence.

[0051] The second step is obstacle annotation and dataset construction:

[0052] Accurate labels are added to obstacles using methods such as bounding box annotation, instance segmentation, and keypoint annotation; the labeled dataset is then divided into training, validation, and test sets in a 7:2:1 ratio.

[0053] The third step is the training and optimization of the AI ​​recognition model:

[0054] We selected YOLO series training, used the pre-trained model weights on a large general dataset as initial values, and then fine-tuned them on a water obstacle dataset.

[0055] Distributed training is performed using servers or cloud platforms equipped with powerful GPUs (graphics processing units).

[0056] Use loss functions and optimizers designed specifically for object detection to guide model learning;

[0057] The trained model is compressed and quantized to convert FP32 precision to INT8 precision;

[0058] Step 4: Shipborne deployment and edge computing box computation:

[0059] The edge computing box converts the AI ​​recognition model into a format optimized for specific hardware, maximizing the release of hardware computing power; it builds a lightweight application that uses OpenCV or GStreamer libraries to capture camera video streams, repeatedly calls the optimized model for inference, and outputs the recognition results, which are then classified into three categories: low difficulty of glide, high difficulty of glide, and no judgment.

[0060] Preferably, in step S3, if the identification result determines that the obstacle is relatively easy to slide over, the speed is reduced and the vehicle moves forward slowly.

[0061] If the identification result indicates that the obstacle is difficult to glide over, the emergency avoidance mechanism will be activated and the aircraft will fly forward and upward.

[0062] When it is impossible to determine the obstruction, reduce speed and move forward slowly. At this time, when the fixed spring is stretched by the force, determine whether to continue to move forward slowly or fly forward and upward based on the force measured by the pull wire gauge.

[0063] When an obstacle is missed and the curved depth sounding rod collides with the obstruction, the entire depth measurement module rotates clockwise around the circular bearing. The two fixed springs provide a certain buffering effect. As the fixed springs are stretched, the pull line gauge is also stretched. When the pull line gauge exceeds the set safety threshold, an emergency avoidance is initiated and the module flies forward and upward. If the pull line gauge does not reach the safety threshold, the module continues to move forward slowly.

[0064] Preferably, in step S4, the water depth measurement process is divided into water areas with constant water level and water areas with changing water level. If the water level is constant, the UAV platform maintains a set altitude for flight based on the real-time positioning of the main antenna of the GNSS positioning component, so that the level gauge and the depth sounding transducer below it remain near the set safe waterline for the measurement.

[0065] In waters with fluctuating water levels, when the UAV platform maintains its designed altitude by using the main antenna of the GNSS positioning component for real-time positioning, the draft measured by the level gauge will change in real time, and the UAV platform's flight altitude will be adjusted accordingly. The adjusted flight altitude is the GNSS measurement altitude plus the increase in depth measured by the level gauge.

[0066] Preferably, in step S6, during the stable operation of the water depth measurement module, the height difference between the main antenna of the GNSS positioning component and the depth transducer is H0. Then, the underwater point position will be determined by the point position measured by the main antenna of the GNSS positioning component as the plane position and the starting elevation surface. The plane position of the main antenna of the GNSS positioning component will be the plane position of the underwater point, and the elevation of the main antenna of the GNSS positioning component will be the water depth value plus H0.

[0067] During the water depth measurement process, if the instantaneous tilt causes the water depth value +H0 to become an oblique path, roll and pitch corrections are performed. These corrections include the displacement of the planar position projected onto the bottom of the water and the height difference of the water depth value +H0 oblique path projected onto the vertical line. The instantaneous roll and pitch values ​​are measured by the attitude component. The absolute coordinate system direction is calculated based on the positioning of the primary and secondary antennas of the GNSS positioning component. The GNSS data, instantaneous water depth data, and attitude data are input into the water depth measurement industrial control computer for calculation to obtain the final water depth result.

[0068] When the depth measurement module collides with an obstruction, causing the pull force of the pull gauge to exceed the threshold, the pull wire is pulled out, and the arc-shaped depth measuring rod will rotate around the circular bearing. At this time, the relative height difference h0 between the upper rectangular frame of the connecting structure and the depth measuring transducer will decrease, and there is a possibility that the level gauge and the depth measuring transducer will be discharged out of the water as they rotate. At this time, the depth measurement value is considered invalid.

[0069] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention relies on a multi-rotor drone platform, has a simple design, is easy to manufacture, and even if the arc-shaped depth measuring rod is damaged, its replacement cost is low; it is easy to install and use, and can automatically glide over or jump over common waterborne obstructions, providing great operational convenience; it adopts a camera monitoring intelligent obstacle identification and obstacle crossing system design, fundamentally avoiding the possibility of interference from waterborne obstructions. Combined with AI obstacle recognition, automatic attitude control and maintenance design, it forms a water depth measurement system suitable for high-density obstructed waters, which automatically crosses obstacles and automatically measures water depth. This solves the problem that in traditional water depth measurement processes, the hydrodynamic propeller often gets entangled in aquatic plants, cables, fishing nets, and plastic bags, losing power and causing the operation to stop, often requiring nearby vessels to assist in freeing it. Attached Figure Description

[0070] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0071] Figure 1 This is a partial structural diagram of a water depth measurement system for high-density navigation obstruction areas according to the present invention;

[0072] Figure 2 This is a block diagram of a water depth measurement system for high-density navigation obstruction areas according to the present invention;

[0073] Figure 3 This is a structural diagram of the unmanned aerial vehicle platform in this invention;

[0074] Figure 4 This is a detailed structural diagram of the connecting and fixing device in this invention;

[0075] Figure 5 This is a structural diagram of the water depth measurement module in this invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Traditional unmanned surface vessels (USVs) equipped with depth measurement devices still rely on manual handling or hoisting for underwater topography measurements. This often encounters challenges due to the surrounding reeds and trees, making it difficult to find a suitable launch path and hindering the deployment of the USV to the designated area. Operators must manually clear debris to create a launch channel. Furthermore, most traditional USVs use hydrodynamic propellers, but in special waterways such as reclamation areas, reservoirs, and inland rivers, floating obstructions like weeds, cables, fishing nets, and plastic products can easily become entangled in the underwater propeller, causing the USV to lose power. Typically, operators use shallow-draft small boats to ferry or wade to the moored USV to clear the underwater propeller, but this increases the risk of the small boat running aground again and poses a significant threat to personnel safety. Currently, existing drones combined with depth measurement devices have not been found to address obstruction and interference with navigation. The purpose of this invention is to provide a depth measurement system for high-density obstruction areas, solving the problem of depth measurement in such areas. The system is as follows:

[0078] like Figure 1-5 As shown, a water depth measurement system for a high-density navigation obstruction area includes: an unmanned aerial vehicle platform 1, a connecting and fixing device 2, a water depth measurement module 3, a water depth measurement industrial control computer 4, and an edge computing box 5;

[0079] Unmanned aerial vehicle platform 1 is used to provide a mounting platform, power support, GNSS data, attitude data, and field-of-view photos in front of the flight path, and is also used to adjust the flight path based on the identification results.

[0080] like Figure 1-3 As shown, the drone platform 1 includes:

[0081] The drone body 11 is used to carry a GNSS positioning component 12, an attitude component 13, a camera 14, a connecting base 15 and an equipment mounting box 16, and is also used to adjust the flight path according to the identification results.

[0082] GNSS positioning component 12 is used for real-time positioning and orientation during UAV flight and water depth measurement;

[0083] Attitude component 13 is used to measure the roll and pitch attitude changes during UAV flight, providing attitude information for three-dimensional water depth calculation;

[0084] Camera 14 is used to view the water surface at different distances, such as near and far, to monitor the safety status of the water surface and acquire water surface images;

[0085] Connecting base 15, used to fix the connecting fixing device 2 by screws;

[0086] Equipment mounting box 16 is used to install the water depth measurement industrial control computer 4 and the edge computing box 5;

[0087] The GNSS positioning component 12 is a dual-antenna GNSS positioning device. The main antenna is located directly above the attitude component 13 and is used for positioning and orientation in conjunction with the secondary antenna. The antenna closer to the camera 14 is the secondary antenna, which is mainly used to assist in orientation.

[0088] Connection and fixing device 2 is used to connect and fix the UAV platform and the water depth measurement module.

[0089] like Figure 1-4 As shown, the connecting and fixing device 2 includes: a connecting structure frame 21, which is composed of a rectangular frame and a triangular frame. One side of the rectangular frame is fixedly connected to the short side of the triangular frame, forming a certain angle.

[0090] The upper rectangular frame of the connecting structure 21 is provided with screw holes, which are connected and fixed to the connecting base 15 by screws.

[0091] A spring retaining ring 22 is set on each side of the rectangular frame of the connecting structure 21, away from the fixed triangular frame;

[0092] A pull wire gauge 23 is installed at the middle position on the same side as the rectangular frame of the connecting structure 21 and the spring fixing ring 22;

[0093] One end of a fixing spring 24 is fixed to each spring fixing ring 22, and the other end of the fixing spring 24 is fixed to the water depth measurement module 3.

[0094] The top of the triangular frame of the connecting structure 21, away from the rectangular frame, is provided with a rotation axis 25;

[0095] in,

[0096] The spring retaining ring 22 has a semi-circular structure;

[0097] Pull gauge 23 is a small-range, high-precision pull gauge with a range of less than 1 meter, used for real-time monitoring of the tension when the fixed spring 24 is stretched.

[0098] The tops of the two sides of the triangular frame of the connecting structure 21 are fixed to the two sides of the rotating shaft 25 respectively.

[0099] The water depth measurement module 3 is used for real-time measurement of water depth data to obtain instantaneous water depth data.

[0100] like Figure 1-5 As shown, the water depth measurement module 3 includes: an arc-shaped depth measuring rod 32. The arc-shaped depth measuring rod 32 adopts an arc-shaped design. The front side is similar to a blade, which makes it easy to break through the water and reduce resistance. The back side is wider and similar to the back of a knife, which has basic sliding ability when encountering floating objects.

[0101] A locking ring 31 is provided on the inner side of the arc-shaped depth measuring rod 32 near the connecting and fixing device 2, and the other ends of the two fixing springs 24 are fixed on the locking ring 31.

[0102] The inner side of the arc-shaped depth measuring rod 32 is slidably connected to the circular bearing 33, and the rotating shaft 25 is fixed inside the circular bearing 33;

[0103] A level gauge 34 and a depth transducer 35 are provided at the end of the arc-shaped depth sounding rod 32 away from the connecting and fixing device 2;

[0104] in,

[0105] Circular bearing 33 is a ball bearing;

[0106] The level gauge 34 is a water pressure sensor used to measure the draft of the depth sounding transducer 35 in real time.

[0107] The sounding transducer 35 is a single-beam sounding transducer used to emit sounding waves into the water bottom, and its bottom surface is flush with the bottom surface of the arc-shaped sounding rod 32.

[0108] In this embodiment, the reference centers of the GNSS positioning component 12, attitude component 13, level gauge 34, and depth sounding transducer 35 are set on the same vertical line.

[0109] The water depth measurement industrial control computer 4 is used to calculate the final water depth result based on GNSS data, attitude data, and instantaneous water depth data.

[0110] Edge computing box 5 is used to perform inference based on the water surface image in front of the flight, output the recognition result, and determine three situations based on the recognition result: low ski-jump difficulty, high ski-jump difficulty, and no judgment.

[0111] In this embodiment, the recognition results are classified into three categories: relatively easy to slide, relatively difficult to slide, and impossible to determine. Based on these three categories, a dynamic window method is used to perform path replanning. The replanning determines whether to proceed slowly or forward and upward. An iterative dataset is established for video data that cannot be determined or is missed, which is used for iterative training and optimization of the AI ​​recognition model. The AI ​​recognition model selected is the YOLO series.

[0112] A method for measuring water depth in high-density navigation obstruction areas, comprising:

[0113] Step S1: Before the water depth measurement operation, after the equipment is assembled on the shore, the multi-rotor UAV carrying the equipment flies to the designated water area to carry out the operation.

[0114] In step S2, camera 14 acquires real-time field-of-view photos of the flight path ahead, focusing on capturing photos of the water surface ahead. Based on the AI ​​recognition model, it infers from the water surface images and outputs the recognition results. Based on the recognition results, it is determined to be one of three situations: relatively easy to skid over, relatively difficult to skid over, or impossible to determine.

[0115] In this embodiment, the AI ​​recognition model performs inference on the water surface image and outputs the recognition result. Based on the recognition result, it is determined into three situations: easy to slide across, difficult to slide across, and cannot be determined.

[0116] Step 1, Sample collection and preprocessing:

[0117] Image data of aquatic plants, fishing nets, trees, cables, plastic bags, and floating objects on the water were collected through actual collection and online search. Artificial simulation samples were added to increase the richness of the samples based on common navigational obstacles in the survey area. The collection of samples took into account factors such as different water areas, different weather, different time periods, and different angles.

[0118] Data is cleaned by removing blurry, dark, and overexposed invalid images; the original images are rotated, scaled, cropped, color-adjusted, and noise-added to expand the dataset size; the image sizes are uniformly adjusted to the fixed size of the model input and normalized to accelerate model training convergence.

[0119] The second step is obstacle annotation and dataset construction:

[0120] Accurate labels are added to obstacles using methods such as bounding box annotation, instance segmentation, and keypoint annotation; the labeled dataset is then divided into training, validation, and test sets in a 7:2:1 ratio.

[0121] The third step is the training and optimization of the AI ​​recognition model:

[0122] We selected YOLO series training, used the pre-trained model weights on a large general dataset as initial values, and then fine-tuned them on a water obstacle dataset.

[0123] Distributed training is performed using servers or cloud platforms with powerful GPUs.

[0124] Use loss functions and optimizers designed specifically for object detection to guide model learning;

[0125] The trained model is compressed and quantized to convert FP32 precision to INT8 precision;

[0126] Step 4, Shipborne Deployment and Edge Computing Box 5 Calculation:

[0127] The edge computing box 5 converts the AI ​​recognition model into a format optimized for specific hardware, maximizing the release of hardware computing power; it builds a lightweight application that uses OpenCV or GStreamer libraries to capture video streams from camera 14, repeatedly calls the optimized model for inference, and outputs the recognition results, which are then classified into three categories: low difficulty of glide, high difficulty of glide, and no judgment.

[0128] Step S3: The judgment is sent to the main control system of the UAV body 11. The dynamic window method is used to carry out path replanning. Based on the path replanning, it is decided to move forward slowly or forward and upward.

[0129] In the embodiment, if the identification result determines that the obstacle is relatively easy to slide over, such as a fishing net, cable, scattered aquatic plants, scattered branches, or a small plastic bag, the speed is reduced and the vehicle moves forward slowly.

[0130] If the identification result determines that the obstacle is a vessel that is difficult to glide over, a long piece of timber, a large buoy, a large plastic bag, or other large floating objects on the water surface, the emergency avoidance will be activated and the aircraft will fly forward and upward.

[0131] When it is impossible to determine the obstruction, reduce the speed and move forward slowly. At this time, when the fixed spring 24 is stretched by the force, the force measured by the pull wire gauge 23 determines whether to continue to move forward slowly or fly forward and upward.

[0132] When an obstacle is missed and the curved depth measuring rod 32 collides with the obstruction, the water depth measuring module 3 rotates clockwise around the circular bearing 33. The two fixed springs 24 provide a certain buffering effect. As the fixed springs 24 are stretched, the pull line gauge 23 is also stretched. When the pull line gauge 23 exceeds the set safety threshold, it is determined to activate emergency avoidance and fly forward and upward. If the pull line gauge 23 does not reach the safety threshold, it continues to move forward slowly.

[0133] Step S4: During the water depth measurement process, a safe draft is set according to the basic information of the water area to keep the level gauge 34 and the depth transducer 35 below it stable in the draft. The level gauge 34 is used to measure the dynamic draft to obtain instantaneous water depth data.

[0134] In this embodiment, the water depth measurement process is divided into water areas with constant water level and water areas with changing water level. If the water level is constant, the UAV platform 1 maintains a set altitude for flight based on the real-time positioning of the main antenna of the GNSS positioning component 12, so that the level gauge 34 and the depth sounding transducer 35 below it remain near the set safe waterline for the measurement.

[0135] If the UAV platform 1 is in a water area with changing water levels, and the UAV platform 1 is maintaining its designed altitude by real-time positioning based on the main antenna of the GNSS positioning component 12, the draft measured by the level gauge 34 will change in real time, and the flight altitude of the UAV platform 1 will be adjusted in real time. The adjusted flight altitude is the GNSS measurement altitude plus the depth increase value measured by the level gauge 34.

[0136] Step S5: GNSS data is obtained by calculating the absolute coordinate system direction through GNSS positioning component 12, and attitude data is obtained by measuring the roll and pitch attitude changes of the UAV during flight through attitude component 13.

[0137] In step S6, the water depth measurement industrial control computer 4 calculates the final water depth result based on GNSS data, attitude data, and instantaneous water depth data.

[0138] In this embodiment, during the stable operation of the water depth measurement module 3, the height difference between the main antenna of the GNSS positioning component 12 and the depth sounding transducer 35 is H0. Then, the underwater point position will be determined by the point position measured by the main antenna of the GNSS positioning component 12 as the planar position and the starting elevation surface. The planar position of the main antenna of the GNSS positioning component 12 will be the underwater point planar position, and the elevation of the main antenna of the GNSS positioning component 12 will be the water depth value plus H0.

[0139] During the water depth measurement process, if the instantaneous tilt causes the water depth value +H0 to become an oblique path, roll and pitch corrections are performed, including the displacement of the planar position projected to the bottom of the water and the height difference of the water depth value +H0 oblique path projected to the vertical line. The instantaneous values ​​of roll and pitch are measured by the attitude component 13. The absolute coordinate system direction is calculated based on the positioning of the primary and secondary antennas of the GNSS positioning component 12. The GNSS data, instantaneous water depth data, and attitude data are input into the water depth measurement industrial control computer 4 for calculation to obtain the final water depth result.

[0140] When the water depth measurement module 3 collides with an obstruction, causing the tension of the pull wire gauge 23 to exceed the threshold, the pull wire is pulled out, and the arc-shaped depth measuring rod 32 will rotate around the circular bearing 33. At this time, the relative height difference h0 between the upper rectangular frame of the connecting structure frame 21 and the depth measuring transducer 35 will become smaller, and there is a possibility that the level gauge 34 and the depth measuring transducer 35 will be discharged into the water as they rotate. At this time, the water depth measurement value is considered invalid.

[0141] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A water depth measurement system for high-density navigation-obstructed areas, characterized in that, include: The UAV platform (1) is used to provide a mounting platform, power support, GNSS data, attitude data and field-of-view photos in front of the flight, and is also used to adjust the flight path according to the identification results. Connection and fixing device (2) is used to connect and fix the UAV platform and the water depth measurement module; The water depth measurement module (3) is used for real-time measurement of water depth data to obtain instantaneous water depth data; A water depth measurement industrial control computer (4) is used to calculate the final water depth result based on GNSS data, attitude data and instantaneous water depth data; The edge computing box (5) is used to reason based on the water surface image in front of the flight, output the recognition result, and determine the three situations based on the recognition result: the skid crossing difficulty is relatively small, the skid crossing difficulty is relatively large, and it cannot be determined. The drone platform (1) includes a connecting base (15) for fixing the connecting and fixing device (2) by screws; The connecting and fixing device (2) includes: a connecting structure frame (21), which is composed of a rectangular frame and a triangular frame. One side of the rectangular frame is fixedly connected to the short side of the triangular frame and forms a certain angle. The upper rectangular frame of the connecting structure frame (21) is provided with screw holes, which are fixed to the connecting base (15) by screws; A spring retaining ring (22) is set on each side of the rectangular frame away from the fixed triangular frame (21); A pull wire gauge (23) is installed at the middle position on the same side as the rectangular frame of the connecting structure (21) and the spring fixing ring (22). One end of a fixing spring (24) is fixed on each spring fixing ring (22), and the other end of the fixing spring (24) is fixed on the water depth measurement module (3); The connecting structure frame (21) has a rotation axis (25) at the top of the triangular frame away from the rectangular frame. in, The spring retaining ring (22) has a semi-circular structure; The draw gauge (23) is a small-range, high-precision draw gauge with a range of less than 1 meter, used for real-time monitoring of the tension when the fixed spring (24) is stretched. The tops of the two sides of the triangular frame of the connecting structure (21) are fixed to the two sides of the rotating shaft (25).

2. The water depth measurement system for high-density navigation-obstructed areas according to claim 1, characterized in that, The unmanned aerial vehicle platform (1) includes: The UAV body (11) is used to carry a GNSS positioning component (12), an attitude component (13), a camera (14) and an equipment mounting box (16), and is also used to adjust the flight path according to the identification results. GNSS positioning component (12) for real-time positioning and orientation during UAV flight and water depth measurement; Attitude component (13) is used to measure the roll and pitch attitude changes during UAV flight, providing attitude information for three-dimensional water depth calculation; The camera (14) is used to view the water surface at different distances, monitor the safety status of the water surface, and acquire water surface images. Equipment mounting box (16) is used to install the water depth measurement industrial control computer (4) and the edge computing box (5); The GNSS positioning component (12) is a dual-antenna GNSS positioning device. The main antenna is located directly above the attitude component (13) and is used for positioning and orientation in conjunction with the secondary antenna. The secondary antenna is located near the camera (14) and is mainly used for auxiliary orientation.

3. The water depth measurement system for high-density navigation-obstructed areas according to claim 2, characterized in that, The water depth measurement module (3) includes: an arc-shaped depth measuring rod (32), which adopts an arc-shaped design; A locking ring (31) is provided on the inner side of the arc-shaped depth measuring rod (32) near the connecting and fixing device (2), and the other ends of the two fixing springs (24) are fixed on the locking ring (31); The inner side of the arc-shaped depth measuring rod (32) is slidably connected to a circular bearing (33), and the rotating shaft (25) is fixed inside the circular bearing (33); A level gauge (34) and a depth transducer (35) are provided at the end of the arc-shaped depth sounding rod (32) away from the connecting and fixing device (2); in, The circular bearing (33) is a ball bearing; The level gauge (34) is a water pressure sensor used to measure the draft of the depth sounding transducer (35) in real time; The depth sounding transducer (35) is a single-beam depth sounding transducer used to emit depth sounding waves to the bottom of the water. Its bottom surface is flush with the bottom surface of the arc-shaped depth sounding rod (32). The reference centers of the GNSS positioning component (12), attitude component (13), level gauge (34) and depth sounding transducer (35) are set on the same vertical line.

4. The water depth measurement system for high-density navigation-obstructed areas according to claim 3, characterized in that, In the edge computing box (5), the recognition results are judged into three situations: less difficult to slide, more difficult to slide, and impossible to judge. The dynamic window method is used to carry out path replanning according to the three situations. The path replanning determines whether to move slowly or forward and upward. An iterative dataset is established for video data that cannot be judged or is missed, which is used for iterative training and optimization of the AI ​​recognition model. Among them, the AI ​​recognition model is selected from the YOLO series.

5. A water depth measurement method based on the water depth measurement system for high-density navigation obstruction areas as described in claim 4, characterized in that, include: Step S1: Before the water depth measurement operation, after the equipment is assembled on the shore, the multi-rotor UAV carrying the equipment flies to the designated water area to carry out the operation. Step S2, the camera (14) acquires the field of view photos in front of the flight in real time, focuses on the water surface in front to collect photos, and infers the water surface image according to the AI ​​recognition model, outputs the recognition results, and determines the three situations of ski jump difficulty, ski jump difficulty and no judgment based on the recognition results; Step S3: The judgment is sent to the main control system of the UAV body (11), and the path replanning is carried out using the dynamic window method. Based on the path replanning, it is decided to move slowly or move forward and upward. Step S4: During the water depth measurement process, a safe draft line is set according to the basic information of the water area to keep the level gauge (34) and the depth transducer (35) below it stable in draft. The level gauge (34) is used to measure the dynamic draft to obtain instantaneous water depth data. Step S5: Calculate the absolute coordinate system direction through the GNSS positioning component (12) to obtain GNSS data, and measure the roll and pitch attitude changes during the flight of the UAV through the attitude component (13) to obtain attitude data. Step S6, the water depth measurement industrial control computer (4) calculates the final water depth result based on GNSS data, attitude data and instantaneous water depth data.

6. The method for measuring water depth in a high-density navigation obstruction area according to claim 5, characterized in that, In step S2, the AI ​​recognition model infers from the water surface image and outputs the recognition result. Based on the recognition result, it is determined into three situations: easy to slide across, difficult to slide across, and cannot be determined. Step 1, Sample collection and preprocessing: Image data of aquatic plants, fishing nets, trees, cables, plastic bags, and floating objects on the water were collected through actual collection and online search. Artificial simulation samples were added to increase the richness of the samples based on common navigational obstacles in the survey area. The collection of samples took into account factors such as different water areas, different weather, different time periods, and different angles. Data is cleaned by removing blurry, dark, and overexposed invalid images; the original images are rotated, scaled, cropped, color-adjusted, and noise-added to expand the dataset size; the image sizes are uniformly adjusted to the fixed size of the model input and normalized to accelerate model training convergence. The second step is obstacle annotation and dataset construction: Accurate labels are added to obstacles using methods such as bounding box annotation, instance segmentation, and keypoint annotation; the labeled dataset is then divided into training, validation, and test sets in a 7:2:1 ratio. The third step is the training and optimization of the AI ​​recognition model: We selected the YOLO series for training, used the pre-trained model weights on a large general dataset as initial values, and then fine-tuned them on a water obstacle dataset. Distributed training is performed using servers or cloud platforms with powerful GPUs. Use loss functions and optimizers designed specifically for object detection to guide model learning; The trained model is compressed and quantized to convert FP32 precision to INT8 precision; Step 4, shipboard deployment and edge computing box (5) calculation: The edge computing box (5) converts the AI ​​recognition model into a format optimized for specific hardware to maximize the release of hardware computing power; builds a lightweight application, uses OpenCV or GStreamer library to capture the video stream of the camera (14), calls the optimized model in a loop for inference, and outputs the recognition results. Based on the recognition results, it is determined to be one of three situations: less difficult to slide, more difficult to slide, or cannot be determined.

7. The method for measuring water depth in a high-density navigation obstruction area according to claim 5, characterized in that, In step S3, if the recognition result determines that the obstacle is relatively easy to slide over, the speed is reduced and the vehicle moves forward slowly. If the identification result indicates that the obstacle is difficult to glide over, the emergency avoidance mechanism will be activated and the aircraft will fly forward and upward. When it is impossible to determine the obstruction, reduce the speed and move forward slowly. At this time, when the fixed spring (24) is stretched by force, the force situation measured by the pull wire gauge (23) determines whether to continue to move forward slowly or fly forward and upward. When an obstacle is missed, the arc-shaped depth measuring rod (32) hits the obstruction below, causing the water depth measuring module (3) to rotate clockwise around the circular bearing (33). The two fixed springs (24) play a certain buffering role. As the fixed springs (24) are stretched, the pull line gauge (23) is also stretched. When the pull line gauge (23) exceeds the set safety threshold, it is determined to start emergency avoidance and fly forward and upward. If the pull line gauge (23) does not reach the safety threshold, it continues to move forward slowly.

8. The method for measuring water depth in a high-density navigation obstruction area according to claim 5, characterized in that, In step S4, during the water depth measurement process, the water level is divided into water areas with constant water level and water areas with changing water level. If the water level is constant, the UAV platform (1) maintains the flight at the set altitude based on the real-time positioning of the main antenna of the GNSS positioning component (12), so that the level gauge (34) and the depth transducer (35) below it remain near the set safe draft for the measurement. If the UAV platform (1) is in a water area with changing water level, and the UAV platform (1) is flying at the designed altitude in real time according to the main antenna of the GNSS positioning component (12), the draft measured by the level gauge (34) will change in real time, and the flight altitude of the UAV platform (1) will be adjusted in real time. The adjusted flight altitude is the GNSS measurement altitude plus the depth increase value measured by the level gauge (34).

9. A method for measuring water depth in a high-density navigation obstruction area according to claim 5, characterized in that, In step S6, during the stable operation of the depth measurement module (3), the height difference between the main antenna of the GNSS positioning component (12) and the depth sounding transducer (35) is... The underwater point location will be determined by the main antenna of the GNSS positioning component (12) as the planar position and starting elevation surface. The planar position of the main antenna of the GNSS positioning component (12) will be the underwater point planar position, and the elevation of the main antenna of the GNSS positioning component (12) + water depth value + ; During the water depth measurement process, the instantaneous tilt caused the water depth value to increase. If the path becomes oblique, then roll and pitch corrections are applied, including the displacement of the planar position projected onto the bottom and the water depth value. The height difference of the oblique path projected onto the vertical line, the instantaneous values ​​of roll and pitch are measured by the attitude component (13), the absolute coordinate system direction is calculated based on the positioning of the primary and secondary antennas of the GNSS positioning component (12), and the GNSS data, instantaneous water depth data, and attitude data are input into the water depth measurement industrial control computer (4) for calculation to obtain the final water depth result; When the depth measurement module (3) collides with an obstruction, causing the tension of the wire gauge (23) to exceed the threshold, the wire is pulled out, and the arc-shaped depth measuring rod (32) will rotate around the circular bearing (33). At this time, the relative height difference between the upper rectangular frame of the connecting structure frame (21) and the depth measuring transducer (35) is... The depth will decrease, and there is a possibility that the level gauge (34) and the depth transducer (35) will rotate and discharge water. At this time, the water depth measurement value is considered invalid.

Images