Karst landform karst cave detection method and underwater robot

Abstract

The invention discloses a karst landform karst cave detection method and an underwater robot. The method comprises the following steps: 1) water entry and initial deployment; the invention relates to a three-dimensional karst cave navigation system which comprises the following components: 1) a three-dimensional karst cave construction module, 2) unpowered drifting navigation, 3) karst cave internal observation and data acquisition, 4) resistance meeting judgment and self-rescue control, 5) continuous drifting and path recording and 6) a three-dimensional karst cave construction module.The three-dimensional karst cave navigation system is compact in structure, waterproof, reliable and high in navigation precision, has the resistance meeting self-rescue capacity and the post-processing three-dimensional modeling capacity and is suitable for narrow and complex underwater environment detection.

Description

Technical Field

[0001] This invention belongs to the field of underwater exploration technology, specifically relating to a method for exploring karst caves and an underwater robot. Background Technology

[0002] Karst landforms are formed by the dissolution and precipitation of soluble rocks by groundwater and surface water, erosion and deposition, as well as gravity collapse, deposition, and other processes. Karst conduits are important storage and transport channels for groundwater. Efficient and accurate detection and characterization of these conduits are crucial for water exploration and groundwater pollution control in karst mountain areas.

[0003] Due to the intricate distribution of karst conduit networks and underground river networks, often accompanied by narrow and enclosed spaces, karst conduit exploration is still in its early stages. The acquisition of raw data mainly relies on traditional geophysical and engineering methods, including drilling, ground-penetrating radar, electrical methods (such as resistivity methods), seismic wave methods, and ground laser scanning. Although these methods are relatively mature in shallow structure exploration, they suffer from problems such as low resolution, limited detection range, and inability to enter complex underwater environments.

[0004] Therefore, providing an efficient, accurate, and controllable method for detecting karst caves and an underwater robot is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for detecting karst caves and an underwater robot to overcome the shortcomings of traditional detection devices and methods and improve detection efficiency and accuracy.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for detecting karst caves includes the following steps:

[0008] 1) Water entry and initial deployment

[0009] The underwater robot was deployed into the water at the entrance of the karst cave.

[0010] 2) Non-powered drifting navigation

[0011] The underwater robot relies on the natural flow of underground water to drift freely, forming a non-powered drifting working mode. The navigation module uses a three-source fusion algorithm to calculate the relative positioning.

[0012] 3) Observation and data collection inside the cave

[0013] The binocular camera captures video images of the forward direction, while the searchlight provides active illumination for the forward field of view. All images and raw data from the navigation module are packaged and encoded, and then saved through the storage module in the control and power supply module, or transmitted back in real time or accessed remotely through the communication module interface.

[0014] 4) Obstacle assessment and self-rescue control

[0015] The Doppler velocimeter detects abnormal speed and identifies drifting and stopping. The inertial measurement unit detects abnormal rotation or floating and overturning by detecting angular velocity. Combined with optical flow anomalies and image recognition, it is determined whether it is in a blind spot or dead zone. Once it is determined that it has entered a distress state, the control and power supply module controls the start of the walking thruster, so that the underwater robot can get out of trouble and resume its natural drift state.

[0016] 5) Continuous drifting and path recording

[0017] After escaping, it continued to drift with the water, repeatedly completing image acquisition, optical flow positioning, and inertial navigation calculation, recording the cave structure, flow characteristics, and path morphology segment by segment;

[0018] 6) Three-dimensional cave construction module

[0019] After completing a full cave rafting mission, the recorded binocular images, velocity information from the inertial measurement unit, optical flow sensor, and Doppler velocimeter, as well as magnetic field-aided information, are exported to a host computer for processing. First, a lightweight visual SLAM framework is used to perform feature matching and inter-frame pose estimation on the binocular images. The data detected by the inertial measurement unit and Doppler velocimeter are then combined to perform trajectory correction and time synchronization. When some visual frames fail, displacement estimation by the optical flow sensor is introduced to maintain trajectory continuity. Subsequently, using the synchronized trajectory and image data, a dense 3D point cloud model of the cave is generated through structured light reconstruction, dense matching, or depth map fusion methods.

[0020] Furthermore, the specific process of the three-source fusion algorithm in step 2) is as follows:

[0021] 1) Optical flow visual perception

[0022] The optical flow sensor continuously captures visual features of the cave ceiling, water surface light and shadow, and skylight structure, calculating the pixel displacement of feature points between adjacent frames. and sampling time interval The relative velocity vector is obtained. :

[0023]

[0024] in, This is the camera imaging scaling factor;

[0025] 2) Inertial Measurement Unit

[0026] The bottom integrates a high-sampling-rate three-axis gyroscope and accelerometer, outputting angular velocity. With linear acceleration attitude angle Obtained by integrating angular velocity:

[0027]

[0028] Among them, linear acceleration Describes the linear acceleration of an object in three-dimensional space. For along Acceleration in the axial direction, For along Acceleration in the axial direction, For along Acceleration in the axial direction,

[0029] Indicates the current moment attitude angle Its three components This is the roll angle. The pitch angle, Yaw angle

[0030] Indicates the initial time. The attitude angle is the initial condition or reference value for integration.

[0031] This represents the lower limit of integration. Indicates the upper limit of the integral.

[0032] : Represents the three-axis angular velocity vector measured by the gyroscope in the inertial measurement unit. A function that varies with time τ, where, Around angular velocity of the axis, Around angular velocity of the axis, Around Angular velocity of the axis;

[0033] 3) Doppler velocimeters

[0034] Multiple beams of sound waves are emitted downwards, and the relative velocity is measured by the reflected echoes from the bottom of the water or suspended particles, outputting a three-dimensional velocity vector. ;

[0035] The relative displacement can be obtained by integrating the velocity:

[0036] ;

[0037] in, This is relative displacement.

[0038] The three-dimensional velocity vector measured by DVL;

[0039] 4) Multi-source data fusion

[0040] The standard discrete-time Kalman filter algorithm is introduced into the data fusion process to predict and update multi-source sensor data. Its mathematical expression is as follows:

[0041] =

[0042]

[0043]

[0044]

[0045] =(I )

[0046] In the state prediction formula: for Predicted state estimation at time; for -1 hour arrives The state transition matrix at time t; For a moment -1 state estimation;

[0047] In the covariance prediction formula: covariance matrix express The covariance of the estimation error at time -1; express The prediction error covariance at time; obtained after measurement update express Error covariance after update at time step; The process noise input matrix; for -1 time step noise covariance;

[0048] In the Kalman gain formula: for The Kalman gain at time step 1 is a weight matrix; for Time-based observation matrix; for Observe the noise covariance at all times;

[0049] In the state update formula, for Posterior state estimation at time 1; for Actual measured value at any given time; To measure the residual, it is the difference between the actual observed value and the predicted observed value. The differences between them;

[0050] In the covariance update formula This is the updated error covariance.

[0051] An underwater robot for performing the methods described above includes:

[0052] The butterfly-shaped fuselage has a buoyancy material on its lower surface.

[0053] The detection unit includes a detection module and a navigation module installed inside the butterfly-shaped fuselage;

[0054] A walking propulsion device, which is mounted on the butterfly-shaped body;

[0055] The control unit includes a control and power supply module and a walking drive module installed inside the butterfly-shaped fuselage. The detection module and the navigation module are electrically connected to the control and power supply module, the control and power supply module is electrically connected to the walking drive module, and the walking drive module is electrically connected to the walking propeller.

[0056] Furthermore, the butterfly-shaped fuselage includes an upper shell, a frame, and a lower shell. The upper shell, the frame, and the lower shell are stacked sequentially from top to bottom and connected together by a plurality of circumferentially distributed screws, thereby forming an installation space between the upper shell, the frame, and the lower shell. The detection module is mounted on the frame; the navigation module is mounted on the frame, the upper shell, and the lower shell; the walking propulsion unit is mounted on the lower shell; the control and power supply module and the walking drive module are both mounted on the frame; and the buoyancy material is coated on the outer surface of the lower shell.

[0057] Furthermore, the detection module includes a binocular camera, the circumferential wall of the frame has a detection hole, and a waterproof plate is installed at the outer end or inside of the detection hole; the binocular camera is installed on the upper platform of the frame near the upper housing, and the camera direction of the binocular camera corresponds to the position of the detection hole; the binocular camera is electrically connected to the control and power supply module.

[0058] Furthermore, the detection module also includes a searchlight, which is mounted on the top of the upper housing, and the direction of illumination of the searchlight is the same as the imaging direction of the binocular camera.

[0059] Furthermore, the navigation module includes an optical flow sensor, an inertial measurement unit, and a Doppler velocimeter. The optical flow sensor is vertically mounted in the middle of the upper housing through a waterproof casing; the inertial measurement unit is mounted on the frame; and the Doppler velocimeter is mounted in the middle of the lower housing. The optical flow sensor, the inertial measurement unit, and the Doppler velocimeter are electrically connected to the control and power supply module, respectively.

[0060] Furthermore, the navigation module also includes LED fill lights, which are mounted on the upper housing and located on both sides of the optical flow sensor.

[0061] Furthermore, sealing rings are provided between the upper housing and the frame, and between the lower housing and the frame; sealing covers are installed around the inertial measurement unit, the control and power supply module, and the walking drive module.

[0062] Therefore, this invention provides a method for detecting karst caves and an underwater robot. Compared with the prior art, this invention has the following advantages:

[0063] (1) Breakthrough in hydrodynamic and structural performance

[0064] The disc-shaped body design significantly reduces water flow resistance and improves the ability to pass through narrow and winding caves (minimum passage diameter ≤30cm); the skeleton-type layered sealing structure (sealing ring + sealing cover for secondary protection) achieves IP68 waterproof rating, ensuring the long-term reliability of the core sensor in high-pressure and turbid water.

[0065] (2) High-precision navigation in GPS-free environments

[0066] It adopts a three-source fusion navigation architecture of optical flow sensor (observing cave top / water surface features from above) + MEMS-IMU + DVL, and suppresses sensor drift in real time through Kalman filtering. The position estimation error is less than 5% of the range, and the attitude angle accuracy reaches ±0.1°, solving the positioning and orientation problem in caves with no light, no map, and no communication.

[0067] (3) Intelligent escape

[0068] The non-powered drifting mode reduces energy consumption (improving range by more than 40%), and the walking thruster pulse propulsion is triggered only when encountering obstacles, avoiding disturbing the sediment; the self-rescue system based on multi-sensor status monitoring (DVL stagnation alarm, IMU attitude anomaly, optical flow feature loss) can respond within 200ms and achieve a jamming escape rate of more than 90%.

[0069] (4) Three-dimensional spatial reconstruction capability

[0070] Simultaneously record binocular vision, IMU, optical flow displacement and DVL velocity data. In the post-processing stage, use vision-inertial tightly coupled SLAM and multi-source data fusion to generate a three-dimensional point cloud model of the cave with centimeter-level accuracy, making up for the deficiency of traditional geophysical exploration methods in obtaining three-dimensional structure.

[0071] (5) Enhanced engineering application value

[0072] Compared to drilling / manual exploration, the cost is reduced by more than 60%, and a single mission can explore underground river channels of ≥3km; the modular design supports quick battery / sensor replacement with a reuse rate of >95%; the multi-machine collaborative mode can realize parallel exploration of karst networks, improving efficiency by more than 3 times. Attached Figure Description

[0073] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0074] Figure 1 An isometric drawing of an underwater robot suitable for exploring karst caves, provided by the present invention;

[0075] Figure 2 A left view of an underwater robot suitable for exploring karst caves, provided by the present invention;

[0076] Figure 3 A bottom view of an underwater robot suitable for exploring karst caves, provided by the present invention;

[0077] Figure 4 An isometric view of an underwater robot with its upper shell removed, suitable for exploring karst caves, provided by this invention;

[0078] Figure 5 The present invention provides an isometric view of an underwater robot suitable for exploring karst caves after removing its lower shell. Detailed Implementation

[0079] 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.

[0080] like Figures 1-5 As shown in the figure, an embodiment of the present invention discloses a method for detecting karst caves, including the following steps:

[0081] 1) Water entry and initial deployment

[0082] The underwater robot was deployed into the water at the entrance of the karst cave. Because the lower shell 13 was coated with buoyancy material, the underwater robot was able to quickly achieve automatic horizontal attitude fixation and float stably on the water surface or in the middle layer of the water body, ready to start the drifting exploration mission.

[0083] 2) Non-powered rafting navigation

[0084] The underwater robot drifts freely using the natural currents of underground water bodies, forming a non-powered drifting operation mode. Its navigation module employs a three-source fusion algorithm for relative positioning calculation.

[0085] S1 Optical Flow Visual Perception

[0086] The optical flow sensor 6 continuously captures visual features of the cave ceiling, water surface light and shadow, and skylight structure, calculating the pixel displacement of feature points between adjacent frames. and sampling time interval The relative velocity vector is obtained. :

[0087]

[0088] in, The camera imaging ratio is used; in low-light environments, the optical flow sensor 6 can be combined with the built-in LED fill light 10 to improve imaging contrast and feature recognition accuracy.

[0089] S2 Inertial Measurement Unit 7

[0090] The bottom integrates a high-sampling-rate three-axis gyroscope and accelerometer, outputting angular velocity. With linear acceleration attitude angle Obtained by integrating angular velocity:

[0091]

[0092] Among them, linear acceleration Describes the linear acceleration of an object in three-dimensional space. For the Acceleration in the axial direction, For the Acceleration in the axial direction, For the Acceleration in the axial direction,

[0093] Indicates the current moment attitude angle Its three components This is the roll angle. The pitch angle, Yaw angle

[0094] Indicates the initial time. The attitude angle is the initial condition or reference value for integration.

[0095] This represents the lower limit of integration. Indicates the upper limit of the integral.

[0096] : Represents the three-axis angular velocity vector measured by the gyroscope in the inertial measurement unit. A function that varies with time τ, where, Around angular velocity of the axis, Around angular velocity of the axis, Around Angular velocity of the axis;

[0097] By using Kalman filtering to model and suppress sensor noise and drift during the integration process, the stability and accuracy of attitude calculation can be significantly improved.

[0098] S3 Doppler Velocimeter 8

[0099] Multiple beams of sound waves are emitted downwards, and the relative velocity is measured by the reflected echoes from the bottom of the water or suspended particles, outputting a three-dimensional velocity vector. ;

[0100] The relative displacement can be obtained by integrating the velocity:

[0101] ;

[0102] in, This is relative displacement.

[0103] The three-dimensional velocity vector measured by DVL;

[0104] During long-term operation, DVL measurements exhibit low-frequency drift, which can be corrected by using optical flow velocity and MEMS predicted velocity, thereby reducing integration error.

[0105] S4 Multi-Source Data Fusion

[0106] In the data fusion process, the standard discrete-time Kalman filter algorithm is introduced to predict and update multi-source sensor data. Kalman filtering calculates the covariance between the prior state and the prediction error based on the system state transition model during the prediction phase, and corrects the prior estimate using observation information during the update phase. This allows for the suppression of sensor noise and drift while obtaining a stable and highly accurate state estimate. Its mathematical expression is as follows:

[0107] =

[0108]

[0109]

[0110]

[0111] =(I )

[0112] In the state prediction formula: for Predicted state estimation at time; for -1 hour arrives The state transition matrix at time t; For a moment -1 state estimation;

[0113] In the covariance prediction formula: covariance matrix express The covariance of the estimation error at time -1; express The prediction error covariance at time; obtained after measurement update express Error covariance after update at time step; The process noise input matrix; for -1 time step noise covariance;

[0114] In the Kalman gain formula: for The Kalman gain at time step 1 is a weight matrix; for Time-based observation matrix; for Observe the noise covariance at all times;

[0115] In the state update formula, for Posterior state estimation at time 1; for Actual measured value at any given time; To measure the residual, it is the difference between the actual observed value and the predicted observed value. The differences between them;

[0116] In the covariance update formula This is the updated error covariance;

[0117] 3) Observation and data collection inside the cave

[0118] The binocular camera 31 captures video images in the forward direction, while the searchlight 32 provides active illumination for the forward field of view. All images and the original data from the navigation module are packaged and encoded and saved through the storage module in the control and power supply module 4, or the data can be transmitted back in real time or accessed remotely through the communication module interface.

[0119] 4) Obstacle assessment and self-rescue control

[0120] The Doppler velocimeter 8 detects abnormal speed and identifies drifting stop. The inertial measurement unit 7 detects abnormal rotation or floating and overturning by detecting angular velocity. Combined with optical flow anomalies and image recognition, it is determined whether it is in a dead angle or blind spot. Once it is determined that it has entered a distress state (stuck, water flow changes drastically, or stopped rotating), the control and power supply module 4 controls the activation of the walking thruster 2, so that the underwater robot can get out of trouble and return to a natural drift state. The water pump nozzle is arranged backward at a preset angle to generate forward directional thrust. The thrust direction is automatically determined and selected according to the current attitude, or a time-sharing trial switching method is used. The thrust magnitude is output according to a predetermined pulse mode, such as first a short pulse from the left pump, if ineffective, then the right pump, and then both pumps; the pulse duration is 100-500ms. The walking thruster 2 equipped in this invention ensures that the equipment has the ability to get out of trouble autonomously in abnormal situations, and enhances the adaptability of the system in extreme environments such as narrow, winding, and unknown environments.

[0121] 5) Continuous drifting and path recording

[0122] After escaping, it continues to drift with the water, repeatedly completing image acquisition, optical flow positioning, and inertial navigation calculation, recording the cave structure, flow direction characteristics, and path morphology segment by segment; optionally, positioning calibration or trigger mode switching is performed when passing through specific identification points (such as infrared markers, manual tags, flow velocity boundaries, etc.).

[0123] 6) Three-dimensional cave construction module

[0124] After completing a full cave rafting mission, the recorded binocular images, velocity information from the inertial measurement unit 7, optical flow sensor 6, and Doppler velocimeter 8, along with magnetic field-aided information, are exported to a host computer for processing. First, a lightweight visual SLAM framework is used to perform feature matching and inter-frame pose estimation on the binocular images. Then, data detected by the inertial measurement unit 7 and Doppler velocimeter 8 are combined for trajectory correction and time synchronization. When some visual frames fail, displacement estimation is introduced using the optical flow sensor 6 to maintain trajectory continuity. Subsequently, using the synchronized trajectory and image data, a dense 3D point cloud model of the cave is generated through structured light reconstruction, dense matching, or depth map fusion methods. The 3D model can be used for subsequent scientific analysis, path planning, and cave structure modeling. This method does not rely on real-time computing power and is suitable for embedded platforms with limited equipment resources. This invention achieves stable positioning and three-dimensional spatial structure reconstruction of karst caves (karst pipes) without human intervention by integrating a multi-source navigation module consisting of an optical flow sensor 6, an inertial measurement unit 7, and a Doppler velocimeter 8, combined with a post-processing modeling method based on three-dimensional point cloud construction.

[0125] This invention also discloses an underwater robot for performing the method described above, capable of detecting underground karst caves (karst pipes) with a diameter greater than 50 cm. The robot includes a hydrodynamically optimized butterfly-shaped body 1, a detection unit, a walking propeller 2, and a control unit. The diameter-to-depth ratio of the butterfly-shaped body 1 is 3:2. The lower surface of the butterfly-shaped body 1 is provided with buoyancy material to provide positive buoyancy and build a stable torque for automatic self-alignment and self-buoyancy in case of failure. The detection unit includes a detection module 3 and a navigation module installed inside the butterfly-shaped body 1. The walking propeller 2 is mounted on the butterfly-shaped body 1. In this embodiment, the walking propeller... Device 2 is selected as a water pump, used to adjust the forward direction of the butterfly-shaped body 1 and drive the butterfly-shaped body 1 forward; the control unit includes a control and power supply module 4 and a walking drive module 5 installed inside the butterfly-shaped body 1. The detection module 3 and the navigation module are electrically connected to the control and power supply module 4, the control and power supply module 4 is electrically connected to the walking drive module 5, and the walking drive module 5 is electrically connected to the walking thruster 2. The control and power supply module 4 is configured to automatically activate the water pump to provide directional thrust for getting out of trouble or adjusting attitude when the underwater robot is detected to be stuck or in an abnormal movement state. The water pump can also be remotely activated by the communication module. This invention features a compact structure, allowing it to freely navigate through complex pipes and caves, reaching areas difficult for humans to access. It possesses the ability to drift naturally with water flow and be propelled by low-interference pulses (self-rescue capability in case of obstruction). Furthermore, it has comprehensive information collection capabilities. By incorporating a detection module 3 and a navigation module, it can accurately acquire information on the direction, scale, and structure of karst pipes and caves. This expands the adaptability of traditional underwater robots in narrow, complex flow, and low-light environments, enabling its widespread application in scenarios such as underground cave exploration, water flow direction analysis, and underwater space modeling. It has significant scientific research and application value.

[0126] Specifically, the butterfly-shaped body 1 includes an upper shell 11, a frame 12, and a lower shell 13. The upper shell 11, frame 12, and lower shell 13 are stacked sequentially from top to bottom and connected together by multiple circumferentially distributed screws, thus forming an installation space between the upper shell 11, frame 12, and lower shell 13. The detection module 3 is installed on the frame 12. The navigation module is installed on the frame 12, upper shell 11, and lower shell 13. The walking thruster 2 is installed on the lower shell 13. The control and power supply module 4 and the walking drive module 5 are both installed on the frame 12. The control and power supply module 4 includes an electrically connected control module and a power supply module. The power supply module is selected as a lithium battery pack as the power source and is centrally arranged near the center of gravity axis of the underwater robot to maintain the overall weight balance. The buoyancy material is coated on the outer surface of the lower shell 13, and the buoyancy material is in the form of a ring or sheet.

[0127] In this embodiment, the skeleton 12 is disc-shaped, with several equally spaced through holes on its periphery, and threaded nests are provided in the through holes; the upper shell 11 and the lower shell 13 are provided with through holes at corresponding positions, which are aligned with the through holes on the skeleton 12; each screw passes through the through holes on the upper shell 11 and the lower shell 13 and is fastened to the threaded nests on the skeleton 12, thereby pressing the upper shell 11, the skeleton 12 and the lower shell 13 together.

[0128] Specifically, the detection module 3 includes a binocular camera 31, and the circumferential wall of the skeleton 12 has a detection hole 121. A waterproof plate is installed at the outer end or inside of the detection hole 121. In this embodiment, the waterproof plate is transparent glass. The binocular camera 31 is installed on the upper platform of the skeleton 12 near the upper shell 11, and the camera direction of the binocular camera 31 corresponds to the position of the detection hole 121. It is used to capture images of the internal structure of the cave, the direction of water flow, or obstacles, and to provide visual information acquisition. The binocular camera 31 is electrically connected to the control and power supply module 4.

[0129] To further optimize the technical solution of the present invention, the detection module 3 also includes a searchlight 32, which is installed on the top of the upper housing 11 and the illumination direction of the searchlight 32 is the same as the imaging direction of the binocular camera 31, thereby improving the image quality and making it particularly suitable for dark, muddy or irregularly structured areas.

[0130] Specifically, the navigation module includes an optical flow sensor 6, an inertial measurement unit (IMU) 7, and a Doppler velocimeter 8 (DVL). The optical flow sensor 6 is vertically mounted in the middle of the upper shell 11 through a waterproof housing 9, and is used to detect the structure of the cave roof or observe changes in water surface characteristics during water drift. The inertial measurement unit 7 is mounted on the frame 12 and is used for real-time attitude angle measurement. The Doppler velocimeter 8 is mounted in the middle of the lower shell 13 and is used to measure the velocity vector relative to the ground, thereby improving underwater autonomous positioning capabilities. The optical flow sensor 6, the inertial measurement unit 7, and the Doppler velocimeter 8 are electrically connected to the control and power supply module 4, respectively.

[0131] To further optimize the technical solution of the present invention, the navigation module also includes an LED fill light 10, which is mounted on the upper housing 11 and located on both sides of the optical flow sensor 6.

[0132] Specifically, sealing rings are provided between the upper housing 11 and the frame 12, and between the lower housing 13 and the frame 12. In this embodiment, annular grooves are provided on the top and bottom surfaces of the outer edge of the frame 12, and O-rings are embedded in the annular grooves, so that the O-rings are located between the frame 12 and the upper housing 11, and between the contact surfaces of the frame 12 and the lower housing 13, and are pressed and sealed by the preload of the screws; sealing covers are installed around the inertial measurement unit 7, the control and power supply module 4, and the walking drive module 5.

[0133] This invention employs technologies such as disc-shaped hydrodynamics + skeleton sealing integration, three-source fusion navigation, non-powered drifting + water pump self-rescue, and post-processing modeling to solve the inherent defects of traditional karst pipeline detection methods in terms of mission flexibility, efficiency, and safety. It is suitable for complex underwater operation scenarios such as karst pipeline detection.

[0134] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0135] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting karst caves, characterized in that, Includes the following steps: 1) Water entry and initial deployment The underwater robot was deployed into the water at the entrance of the karst cave. 2) Non-powered drifting navigation The underwater robot relies on the natural flow of underground water to drift freely, forming a non-powered drifting working mode. The navigation module uses a three-source fusion algorithm to calculate the relative positioning. 3) Observation and data collection inside the cave The binocular camera captures video images of the forward direction, while the searchlight provides active illumination for the forward field of view. All images and raw data from the navigation module are packaged and encoded, and then saved through the storage module in the control and power supply module, or transmitted back in real time or accessed remotely through the communication module interface. 4) Obstacle assessment and self-rescue control The Doppler velocimeter detects abnormal speed and identifies drifting and stopping. The inertial measurement unit detects abnormal rotation or floating and overturning by detecting angular velocity. Combined with optical flow anomalies and image recognition, it is determined whether it is in a blind spot or dead zone. Once it is determined that it has entered a distress state, the control and power supply module controls the start of the walking thruster, so that the underwater robot can get out of trouble and resume its natural drift state. 5) Continuous drifting and path recording After escaping, it continued to drift with the water, repeatedly completing image acquisition, optical flow positioning, and inertial navigation calculation, recording the cave structure, flow characteristics, and path morphology segment by segment; 6) Three-dimensional cave construction module After completing a full cave rafting mission, the recorded binocular images, velocity information from the inertial measurement unit, optical flow sensor, and Doppler velocimeter, as well as magnetic field-aided information, are exported to a host computer for processing. First, a lightweight visual SLAM framework is used to perform feature matching and inter-frame pose estimation on the binocular images. The data detected by the inertial measurement unit and Doppler velocimeter are then combined to perform trajectory correction and time synchronization. When some visual frames fail, displacement estimation by the optical flow sensor is introduced to maintain trajectory continuity. Subsequently, using the synchronized trajectory and image data, a dense 3D point cloud model of the cave is generated through structured light reconstruction, dense matching, or depth map fusion methods.

2. The method for detecting karst caves according to claim 1, characterized in that, The specific process of the three-source fusion algorithm in step 2) is as follows: 1) Optical flow visual perception The optical flow sensor continuously captures visual features of the cave ceiling, water surface light and shadow, and skylight structure, calculating the pixel displacement of feature points between adjacent frames. and sampling time interval The relative velocity vector is obtained. : ； in, This is the camera imaging scaling factor; 2) Inertial Measurement Unit The bottom integrates a high-sampling-rate three-axis gyroscope and accelerometer, outputting angular velocity. With linear acceleration attitude angle Obtained by integrating angular velocity: ； Among them, linear acceleration Describes the linear acceleration of an object in three-dimensional space. For along Acceleration in the axial direction, For along Acceleration in the axial direction, For along Acceleration in the axial direction, Indicates the current moment attitude angle Its three components This is the roll angle. The pitch angle, Yaw angle Indicates the initial time. The attitude angle is the initial condition or reference value for integration. This represents the lower limit of integration. Indicates the upper limit of the integral. This represents the three-axis angular velocity vector measured by the gyroscope in the inertial measurement unit. A function that varies with time τ, where, Around angular velocity of the axis, Around angular velocity of the axis, Around Angular velocity of the axis; 3) Doppler velocimeters Multiple beams of sound waves are emitted downwards, and the relative velocity is measured by the reflected echoes from the bottom of the water or suspended particles, outputting a three-dimensional velocity vector. ; The relative displacement can be obtained by integrating the velocity: ； in, This is relative displacement. The three-dimensional velocity vector measured by DVL; 4) Multi-source data fusion The standard discrete-time Kalman filter algorithm is introduced into the data fusion process to predict and update multi-source sensor data. Its mathematical expression is as follows: = ； ； ； ； =(I ) ； In the state prediction formula: for Predicted state estimation at time; for -1 hour arrives The state transition matrix at time t; For a moment -1 state estimation; In the covariance prediction formula: covariance matrix express The covariance of the estimation error at time -1; express The prediction error covariance at time; obtained after measurement update express Error covariance after update at time step; The process noise input matrix; for -1 time step noise covariance; In the Kalman gain formula: for The Kalman gain at time step 1 is a weight matrix; for Time-based observation matrix; for Observe the noise covariance at all times; In the state update formula, for Posterior state estimation at time 1; for Actual measured value at any given time; To measure the residual, it is the difference between the actual observed value and the predicted observed value. The differences between them; In the covariance update formula This is the updated error covariance.

3. An underwater robot for performing the method as described in claim 1 or 2, characterized in that, include: The butterfly-shaped fuselage has a buoyancy material on its lower surface. The detection unit includes a detection module and a navigation module installed inside the butterfly-shaped fuselage; A walking propulsion device, which is mounted on the butterfly-shaped body; The control unit includes a control and power supply module and a walking drive module installed inside the butterfly-shaped fuselage. The detection module and the navigation module are electrically connected to the control and power supply module, the control and power supply module is electrically connected to the walking drive module, and the walking drive module is electrically connected to the walking propeller.

4. An underwater robot according to claim 3, characterized in that, The butterfly-shaped fuselage includes an upper shell, a frame, and a lower shell. The upper shell, the frame, and the lower shell are stacked sequentially from top to bottom and connected together by multiple circumferentially distributed screws, thereby forming an installation space between the upper shell, the frame, and the lower shell. The detection module is mounted on the frame; the navigation module is mounted on the frame, the upper shell, and the lower shell; the walking propulsion unit is mounted on the lower shell; the control and power supply module and the walking drive module are both mounted on the frame; and the buoyancy material is coated on the outer surface of the lower shell.

5. An underwater robot according to claim 4, characterized in that, The detection module includes a binocular camera. The circumferential wall of the frame has a detection hole, and a waterproof plate is installed at the outer end or inside of the detection hole. The binocular camera is installed on the upper platform of the frame near the upper housing, and the camera direction of the binocular camera corresponds to the position of the detection hole. The binocular camera is electrically connected to the control and power supply module.

6. An underwater robot according to claim 5, characterized in that, The detection module also includes a searchlight, which is mounted on the top of the upper housing, and the direction of illumination of the searchlight is the same as the direction of imaging of the binocular camera.

7. An underwater robot according to claim 4, characterized in that, The navigation module includes an optical flow sensor, an inertial measurement unit, and a Doppler velocimeter. The optical flow sensor is vertically mounted in the middle of the upper housing through a waterproof shell; the inertial measurement unit is mounted on the frame; and the Doppler velocimeter is mounted in the middle of the lower housing. The optical flow sensor, the inertial measurement unit, and the Doppler velocimeter are electrically connected to the control and power supply module, respectively.

8. An underwater robot according to claim 7, characterized in that, The navigation module also includes LED fill lights, which are mounted on the upper housing and located on both sides of the optical flow sensor.

9. An underwater robot according to claim 7, characterized in that, Sealing rings are provided between the upper housing and the frame, and between the lower housing and the frame; sealing covers are installed around the inertial measurement unit, the control and power supply module, and the walking drive module.

Images