Underwater target observation method and device and underwater robot

Abstract

The invention relates to the technical field of underwater detection, and discloses an underwater target observation method and device, and an underwater robot, and the method controls the underwater robot to perform fixed-distance observation on an object to be measured along a main navigation path, and obtains a sonar image at the outer side of the main navigation path in real time. When it is judged that a target area meeting a preset condition exists according to the sonar image, the underwater robot is controlled to interrupt the main task, and the current interruption position and posture are recorded; an observation point is determined based on the interruption position and the target area, and the underwater robot is controlled to actively observe the target area; and after observation is completed, the system is controlled to return to the interruption position and recover the interruption posture, and fixed-distance observation continues to be executed along the original main navigation path. According to the application, by expanding lateral perception and constructing an interruption recovery closed loop, automatic response to key targets outside a path is realized while the scanning continuity and integrity are ensured, and the coverage integrity and operation efficiency of underwater observation are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of underwater detection technology, and in particular to a method, device, and underwater robot for observing underwater targets. Background Technology

[0002] In underwater structure surface inspection operations, underwater robots equipped with forward-facing sonar typically scan along a pre-planned path. This method relies on the pre-set path and the robot's tracking accuracy, and has significant limitations in real-world complex underwater environments. Since forward-facing sonar can only acquire information within a limited area directly in front of the robot, if the pre-planned path does not perfectly match the actual shape of the structure, or if the robot deviates from its trajectory due to interference from water currents, some surface areas may remain uncovered by sound waves, creating scanning blind spots and affecting the completeness and reliability of the inspection. Summary of the Invention

[0003] Based on this, it is necessary to address the technical problem that existing underwater observation technologies have scanning blind spots, which affect the integrity and reliability of detection. Therefore, a method, device, and underwater robot for observing underwater targets are proposed.

[0004] Firstly, a method for observing underwater targets is provided, the method comprising: The underwater robot performs fixed-distance observations of the object under test according to a preset main navigation path; the main navigation path is a continuous observation path planned for the object under test. During the observation of the main navigation path, sonar images of the spatial region outside the main navigation path are acquired in real time; Based on the sonar image, determine whether there is a target area that meets the preset conditions; When the target area is present, the underwater robot interrupts its observation of the object under test and records the current interruption position and interruption posture. Based on the interruption location and the target area, an observation point is determined, and the underwater robot is controlled to actively observe the target area. After completing the observation of the target area, the underwater robot is controlled to return to the interrupted position and restore the interrupted attitude, and continue to perform fixed-distance observation of the object under test along the main navigation path.

[0005] Secondly, an underwater target observation device is provided, the device comprising: The navigation module is used to perform fixed-distance observation of the object under test according to a preset main navigation path; the main navigation path is a continuous observation path planned for the object under test. The acquisition module is used to acquire sonar images of the spatial region outside the main navigation path in real time during the observation of the main navigation path; The judgment module is used to determine whether there is a target area that meets preset conditions based on the sonar image; An interrupt module is used to interrupt the observation of the object under test when the judgment result of the judgment module is yes, and to record the current interruption position and interruption posture. The observation module is used to determine observation points based on the interruption location and the target area, and control the underwater robot to actively observe the target area; The recovery module is used to control the underwater robot to return to the interrupted position and restore the interrupted posture after completing the observation of the target area, and continue to perform fixed-distance observation of the object under test along the main navigation path.

[0006] Thirdly, an underwater robot is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described underwater target observation method.

[0007] Beneficial effects: I. Compared to existing methods that rely solely on forward-facing sonar scanning along a fixed path, this application acquires real-time sonar images of the outer side of the path while simultaneously performing fixed-range observations along the main navigation path. This enables the underwater robot to simultaneously perceive environmental information outside the main scanning corridor, effectively extending the robot's effective detection range from a single, narrow forward-facing area to a wider lateral space. Through real-time analysis and judgment of the lateral sonar images, it can proactively discover target areas that meet the characteristics of defects but are located outside the preset observation path, thereby overcoming the problem of missed detections caused by path planning deviations or limited forward field of view, and significantly improving the probability of detecting surface defects and the completeness of coverage in the entire observation mission.

[0008] II. Upon discovering the target area, this application defines a control closed loop that includes recording the interruption location and posture, returning to the interruption location, and restoring the interruption posture. By accurately recording the spatiotemporal state at the moment of interruption, and driving the robot to accurately return to the recorded state after completing the observation of the target area, the robot can temporarily leave the preset path to perform maneuvering observations, and then accurately resume the original path tracking and distance observation tasks. This ensures the continuity and integrity of the main navigation path observation, avoiding the path coverage gaps or chaos that may be caused by temporary turning in existing technologies. Thus, while flexibly responding to sudden discoveries, it maintains the efficiency and orderliness of systematic scanning.

[0009] Third, this application integrates multiple stages, including sonar image acquisition, target condition judgment, interruption decision recording, active observation execution, and state recovery, into an automated process. The entire system no longer passively executes preset commands but possesses the ability to autonomously judge and plan actions based on real-time environmental perception. When a target meeting preset conditions appears, it can autonomously trigger a complete set of standardized response actions without human intervention, achieving full automation from perception to anomaly detection to detailed investigation and regression. This significantly improves the intelligence level and response speed of operations, enabling underwater robots to efficiently meet the dual needs of systematic general surveys and detailed investigations of key targets in unattended or remotely monitored modes. Attached Figure Description

[0010] 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] in: Figure 1 This is a flowchart of an underwater target observation method in one embodiment; Figure 2 This is a schematic diagram of the main navigation path in one embodiment; Figure 3 This is a schematic diagram illustrating the principle of observing a target area in one embodiment; Figure 4 This is a schematic diagram of the process for controlling the observation perspective of an underwater robot in one embodiment; Figure 5 This is a flowchart illustrating hover control in one embodiment; Figure 6 This is a schematic diagram of the process for controlling the spatial position of an underwater robot in one embodiment; Figure 7 This is a flowchart illustrating the interruption control of an underwater robot in one embodiment; Figure 8 This is a structural block diagram of an underwater target observation device in one embodiment; Figure 9 This is a structural block diagram of an underwater robot in one embodiment. Detailed Implementation

[0012] 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, not all, of the embodiments of the present invention. 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.

[0013] The underwater observation system of this application includes: a management device, a surface base station, and at least one underwater robot. The management device can communicate with the underwater robot via cable or the surface base station. In some embodiments, the management device can also be located on the surface base station. The underwater robot mentioned above includes, but is not limited to, types of underwater robots such as remotely operated vehicles (ROVs) and autonomous remotely controlled vehicles (ARVs), and can also be underwater detection equipment, underwater submarine equipment, or other underwater operation equipment; this application does not impose any limitations on this.

[0014] The management equipment is installed in the aquatic environment to provide underwater robots with a main navigation path for observing underwater objects (such as bridge piers, ship hulls, etc.) and to configure the main navigation path to the designated underwater robot. The underwater robot observes the object based on the main navigation path to monitor for defects such as cracks and corrosion. The management equipment can be a mobile device, tablet computer, or fixed computer, etc., and this application does not impose any restrictions on this.

[0015] Floating base stations can also take the form of ship hulls or other waterborne equipment. They are typically equipped with GNSS (Global Navigation Satellite System) and a USBL (Ultra-Short Baseline) transducer array positioned below the water surface. Floating base stations also serve as communication hubs between management equipment and underwater robots, handling task scheduling and ensuring coordinated responses from management equipment commands and underwater robots. In some embodiments, the floating base station also has the capability to supply power to the underwater robot.

[0016] In this embodiment, the underwater robot is equipped with a USBL transponder, which works in conjunction with the USBL transducer array set up by the surface base station for cooperative positioning. The positioning principle is as follows: the surface base station obtains the position coordinates of the surface robot based on the GNSS module, the underwater robot uses the USBL transponder to measure the relative position offset between itself and the surface base station, and then calculates its own current position coordinates based on the relative position offset and the position coordinates of the surface base station.

[0017] The present invention will now be described in detail through specific embodiments.

[0018] Please see Figure 1 As shown, Figure 1 A flowchart illustrating an underwater target observation method provided in an embodiment of the present invention includes the following steps: S1. The underwater robot performs fixed-distance observation of the object under test according to the preset main navigation path.

[0019] The main navigation path is a continuous observation path planned for the object under test. It is a pre-planned, continuous spatial trajectory designed to complete the systematic observation task of a specific object. The underwater robot acquires the preset main navigation path from the management device via underwater cable or wirelessly. This path can be pre-generated by the management device's host computer software based on the object's 3D model, preset observation coverage requirements, and the optimal working distance range of the forward-facing sonar ranging device, using a path planning algorithm / model. Alternatively, it can be pre-generated by the management device's host computer software using a path planning algorithm / model based on spatial observation intervals. The main navigation path is represented as a series of sequentially connected spatial waypoints, each containing 3D position coordinates and optional desired attitude angle information. The underwater robot reads this path data through its navigation computer. In actual operation, the underwater robot may not traverse all points on the main navigation path during execution, depending on the actual observation conditions.

[0020] During the path execution phase, the underwater robot's integrated navigation system continuously provides its real-time position and attitude data. This system typically integrates data from inertial measurement units, Doppler logs, and ultra-short baseline positioning sensors. The motion controller compares the current real-time position with the position of the current target waypoint on the path, calculates the position deviation, and generates thrust commands for each thruster based on a pre-defined tracking control law, such as proportional-integral-derivative control, driving the robot along the preset path towards the target waypoint. Upon reaching a waypoint, it automatically switches to the next waypoint, thus achieving continuous path tracking.

[0021] Meanwhile, to ensure the observation effect, in the preferred embodiment of this application, the underwater robot will perform a fixed-distance observation mode on the object being measured. That is, the forward sonar ranging device carried by the underwater robot will continuously emit sound waves and receive the echo from the surface of the object being measured, calculate the straight-line distance between the front of the robot and the surface of the object being measured in real time, and dynamically adjust the observation position of the underwater robot to ensure that the underwater robot can always maintain a good observation effect.

[0022] For example, see Figure 2 The diagram shows the main navigation path.

[0023] Figure 2 The main navigation path shown is bow-shaped. When the underwater robot conducts crack inspection on the surface of an underwater bridge pier, the object to be measured is a certain vertical side of the bridge pier. The main navigation path is planned as a bow-shaped path covering a specific rectangular area on this side. The planning of this path is based on the length of the top edge, the length of the bottom edge, and the vertical height of the area to be monitored. The underwater robot starts execution from the starting point at the upper left corner of the top of this area.

[0024] First, the robot keeps the depth unchanged and moves horizontally from left to right (for example: from point L to point M). At the same time, its forward sonar continuously measures the distance from the surface of the bridge pier, and through distance control, it maintains an expected observation distance of about 2 meters. After reaching the right boundary, the first horizontal survey line is completed. Then, according to the path planning, while keeping the forward sonar pointing at the surface of the bridge pier, the robot vertically moves downward a predetermined distance (i.e., Figure 2 △h in ), such as 0.5 meters. After reaching the next depth, the robot then moves horizontally from right to left, executes the second horizontal survey line, and also maintains distance-based observation. Repeating this process, it first moves horizontally to cover a certain width, then vertically dives a certain depth, and then moves horizontally in the reverse direction, forming a spatial broken-line trajectory similar to a bow shape.

[0025] During the entire execution process of the bow-shaped path, the motion controller of the underwater robot is responsible for coordinating the timing and accuracy of horizontal movement and vertical diving, while the distance-based observation controller continuously works throughout the process to ensure that the observation distance between the robot and the surface of the bridge pier is stably controlled at the set distance whether the robot is in the horizontal movement stage or the depth transformation stage, so as to achieve continuous, uniform, and constant-distance scanning coverage of the bridge pier surface.

[0026] S2. During the observation process of the main navigation path, sonar images of the spatial area outside the main navigation path are obtained in real time.

[0027] Among them, the generation of sonar images depends on the sonar imaging equipment carried by the underwater robot. For example: the sonar imaging equipment can be a side-scan sonar, a multi-beam bathymetric sonar, etc., which are installed on both sides or specific orientations of the robot body. During the whole process of the robot moving along the main navigation path, this sonar imaging equipment continuously works.

[0028] The working process is as follows: the sonar transmitter emits a fan-shaped sound wave beam in a direction perpendicular to the robot's heading or at a specific side-inclination angle with the heading according to the set pulse repetition interval. When the sound wave propagates and encounters the seabed, water body structure, or the extended part of the object to be measured outside the main navigation path, it will scatter, and part of the echo signal is captured by the sonar receiver. The received original analog signal is amplified, filtered, and digitized, and then converted into a digital echo sequence.

[0029] To generate a two-dimensional sonar image, the distance information corresponding to each echo signal is calculated based on its arrival time, and pixel grayscale values ​​are assigned according to the echo intensity. Simultaneously, combined with the precise position and attitude data provided in real time by the underwater robot's navigation system, particularly the heading and roll angles, motion compensation and geographic coordinate correction are performed on each echo data beam. Finally, the continuously acquired and corrected scanline data are sequentially stitched together to generate and update a two-dimensional acoustic image reflecting the acoustic characteristics of the terrain surface outside the path in real time.

[0030] The principle for generating 3D sonar images or point clouds is typically based on multibeam echo sounding or interferometry. By simultaneously acquiring echoes from multiple adjacent beams and processing the phase difference or angle of arrival information of each beam echo, not only can the target's distance information be obtained, but its azimuth and depth information relative to the sonar can also be calculated, thus directly generating 3D spatial point cloud data. Whether generating a 2D image or a 3D point cloud, this data acquisition process is synchronized and parallel in time with the forward ranging observation process in step S1, coordinated and scheduled by the robot's data processing unit to ensure synchronous perception of the core observation path and its surrounding environment.

[0031] For example, an underwater robot is performing a fixed-distance observation along a bow-shaped path towards the front of a bridge pier. As the robot moves from left to right along one of the horizontal survey lines, its side-scan sonar, mounted on its right side, operates synchronously. The sonar's sound wave emission surface faces outwards on the robot's starboard side, i.e., towards the right side of the pier. The sound waves propagate outwards in a fan-shaped beam, covering an area of ​​a certain width and distance to the right of the robot.

[0032] As the robot moves, the sonar continuously emits pulses and receives echoes. For each pulse echo, the system calculates the slant distance between the reflection point and the robot based on the speed of sound in water and the echo return time. Combined with the robot's real-time high-precision integrated navigation data, knowing its precise latitude, longitude, depth, heading, and attitude at every moment, the system can locate each reflection point on each scan line in the global coordinate system and assign it a sound intensity value. Continuous scan lines are stitched together to form a two-dimensional acoustic image showing the surface condition of the right side of the bridge pier in real time. If a three-dimensional imaging sonar is used, the output is a three-dimensional point cloud model of the side surface, which can directly reflect the geometric undulations of the surface. This imaging process is completely autonomous, independent of the main navigation task running directly in front of the robot, realizing real-time monitoring of the space outside the predetermined observation area.

[0033] S3. Based on the sonar image, determine whether there is a target area that meets the preset conditions.

[0034] The raw sonar data is optimized to improve the reliability of the analysis. For two-dimensional images, median filtering or Gaussian filtering algorithms can be used to suppress random noise, and histogram equalization can be used to enhance image contrast, making potential target features easier to identify. For three-dimensional point cloud data, statistical filtering methods can be used to remove isolated noise points, and voxel downsampling may be performed to improve processing efficiency while preserving geometric features.

[0035] The system automatically detects potential target regions of interest from preprocessed data. For 2D images, edge detection algorithms may be used to identify obvious linear contours, or region growing and connected component analysis may be used to find anomalous patches with uniform acoustic reflectivity. For 3D point clouds, surface geometric anomalies, such as protrusions or depressions, can be located through surface normal vector analysis or elevation change detection.

[0036] For each identified potential target region, its key feature parameters are calculated. These parameters typically include geometric and spatial parameters. Geometric parameters, depending on the target's shape, may include calculated actual length, width, area, perimeter, aspect ratio, or depth difference. Spatial parameters mainly refer to the positional relationship of the target region relative to the underwater robot body. These are usually calculated using a sonar geometric measurement model and the robot's real-time pose data to determine the horizontal distance or slant distance between the geometric center of the target region and the robot.

[0037] The quantized parameters of each potential target region are compared one by one with preset discrimination conditions. These preset conditions are combinations of logical rules predefined according to the specific observation task, typically including spatial distance constraints and feature threshold constraints. For example, a complete discrimination rule is: the distance between the target region and the robot is less than 5 meters, and the feature length identified as a crack is greater than 0.2 meters. Only when all relevant parameters of a potential target region simultaneously meet the requirements of all clauses in the preset conditions is it confirmed as a valid target region, and the type, precise location, and feature parameters of the target region are output.

[0038] In one possible embodiment of this application, the above-mentioned preset conditions are: the distance between the target area and the underwater robot is less than or equal to a distance threshold, and the quantification index of the disease corresponding to the image features of the target area in the sonar image exceeds a preset feature threshold.

[0039] The first step is to calculate the first metric: the real-time distance between the target area and the underwater robot. Once a potential target area is identified in the sonar image, the linear slant distance or horizontal projection distance between the target's pixel position in the sonar image, the geometric projection model of the sonar beam, and the underwater robot's real-time position and attitude data provided by its navigation system are calculated. This calculation result is the first quantified value for subsequent judgments.

[0040] Next, the second indicator, namely the disease quantification index, is calculated. Image features related to specific disease types are extracted from sonar images. For example, for crack disease, key features are its linear shape and size. Continuous or discontinuous edges in the sonar image are identified and tracked, fitted into a linear structure, and their total length, maximum width, or average width at the actual scale are calculated. For corrosion or porosity disease, key features might be the area and contrast of the region. Connected regions with abnormal acoustic reflection characteristics (too strong or too weak) are segmented, their actual projected area is calculated, and the difference between the average gray level of the region and the average gray level of the background may be calculated. These calculated parameters, such as length, area, and contrast, constitute the disease quantification index for the target area.

[0041] The final conditional decision is a simultaneous verification of the two quantified values ​​mentioned above. The calculated real-time distance value is compared with a preset distance threshold, and the calculated disease quantification index is compared with a preset corresponding feature threshold. Only when both comparison results are true—that is, the real-time distance is less than or equal to the distance threshold, and the disease quantification index exceeds the feature threshold—will the system ultimately determine that the currently identified area is a target area that meets the preset conditions and trigger subsequent processes. If either condition is not met, the area is considered an invalid or secondary target and will not be responded to.

[0042] For example, when an underwater robot inspects a bridge pier, the preset conditions are: the target distance must be less than or equal to 5 meters, and the crack length must exceed 0.2 meters.

[0043] When a linear feature resembling a crack appears in the sonar image, the system first performs distance measurement. Based on the feature's location in the image and the sonar parameters, the distance from the center of the crack to the robot's current position is calculated to be 3.5 meters. This value satisfies the condition of being less than 5 meters.

[0044] Subsequently, a quantitative analysis was performed on the linear feature. After image processing and scale conversion, the actual length of the linear feature was calculated to be 0.3 meters. This value satisfies the condition of exceeding 0.2 meters.

[0045] Since the distance value of 3.5 meters and the length value of 0.3 meters both satisfy the two preset threshold conditions, the area is determined to be a valid target area and marked as a crack that needs further observation. Conversely, if the feature length is 0.15 meters, even if the distance is satisfied, the system will ignore it because it does not meet the feature threshold.

[0046] It should be noted that the distance threshold and feature threshold are not arbitrarily set, but are calculated or calibrated in advance using a systematic method based on the physical performance constraints of the underwater robot, the working characteristics of the sensors, the structural safety standards of the object under test, and the specific requirements of the detection task. This determination process is a standard technical preparation step in planning similar operations in this field.

[0047] The determination of the distance threshold is mainly based on a comprehensive consideration of the following three aspects: First, the maneuverability of the underwater robot, including its minimum turning radius under load, braking distance from cruising speed to a standstill, and hovering positioning accuracy. This determines whether the robot can safely and efficiently interrupt the main path and move to the operational space boundary of the target to the side. Second, the effective range and measurement accuracy of the side-scan sonar. Sonar images at long distances will experience a decrease in resolution due to sound wave attenuation and beam spread. Setting a distance threshold must ensure that within this distance, the image quality is sufficient to support reliable feature recognition and quantization. Third, the balance between the overall efficiency and safety of the task. Setting the threshold too high will cause the robot to frequently respond to secondary targets at a distance, affecting the efficiency of the main task. Setting it too low may cause important defects in the vicinity to be missed. Considering these factors, the distance threshold is usually set to a value less than or equal to the robot's maximum effective response radius, which can be measured through simulation or water trials.

[0048] The determination of feature thresholds is primarily based on the testing standards of the industry to which the object under test belongs, structural health diagnostic standards, and statistical analysis of historical defect data. First, the type of target defect (e.g., cracks, corrosion, attachments) needs to be clearly defined. For cracks, the feature threshold usually refers to the minimum reported length or width, derived from engineering structural safety assessment standards. For example, the underwater testing code for concrete structures may stipulate that cracks exceeding 0.2 meters in length must be recorded and reported. For corrosion, the threshold may refer to the minimum area or maximum depth, based on material corrosion allowance and structural strength calculations. When determining specific values, relevant national and industry standards or the owner's specific technical specifications will be consulted. Furthermore, in the early stages of the project, sonar scanning of sample areas in known states (health and defects) to establish a correlation model between sonar image feature parameters (e.g., linear length, area, grayscale contrast) and actual defect measurements also helps to calibrate and set more accurate feature thresholds, ensuring that the quantitative indicators of the sonar correspond to the actual engineering significance of the defect.

[0049] Taking the detection of corrosion in underwater steel pile foundations as an example, the distance threshold is set at 4 meters. This value is determined based on the following: under operational load, the maximum reliable lateral maneuver radius for stable turning and positioning of this type of underwater robot is approximately 5 meters, as tested; the side-scan sonar used can guarantee the identification of corrosion patches with a diameter greater than 0.1 meters within a distance of 4 meters; and considering the overall task efficiency evaluation, limiting the response range to within 4 meters is reasonable.

[0050] The characteristic threshold was set at 0.05 square meters for a single corrosion area. This value was determined based on the following: according to the technical specifications for the maintenance of port facilities in this sea area, for this type of load-bearing pile foundation, corrosion patches exceeding 0.05 square meters in area are required to be recorded and evaluated. During the project preparation phase, technicians used a sonar system to image a test plate with a corrosion area of ​​known size (calibrated to 0.08 square meters). The system was confirmed to be able to accurately segment and calculate the area from the image, thus verifying the feasibility and reliability of using 0.05 square meters as the characteristic threshold.

[0051] S4. When a target area exists, the underwater robot interrupts the observation of the object under test and records the current interruption position and interruption posture.

[0052] When the judgment result of step S3 is yes, step S4 is executed. Interruption refers to the underwater robot pausing its currently executing main navigation path tracking and ranging observation tasks. The underwater robot's main control system sends high-level instructions to the motion controller, which then stops generating tracking control instructions based on preset waypoints and switches to a position-holding mode. In this mode, using the underwater robot's current position as the setpoint, closed-loop control drives the thrusters to counteract disturbances such as water flow, allowing the robot to hover stably near the current point, providing a relatively static state basis for subsequent recording operations.

[0053] While the robot remains stable, it simultaneously records critical states related to the interruption position and posture. This process essentially involves reading and freezing the measurements taken by the robot's core navigation sensors at the moment of interruption or immediately after stabilization.

[0054] Recording the interruption position relies on the robot's integrated navigation system. From the data fusion output of this navigation system, the three-dimensional coordinates of the robot's body coordinate system origin or a specified reference point in the global or task coordinate system are acquired and stored. These coordinates typically include longitude, latitude, and depth information, or X, Y, and Z coordinate values ​​in a local coordinate system with a fixed origin. Data acquisition needs to be synchronized with the navigation system's data update cycle to ensure timestamp consistency.

[0055] Recording the interrupted attitude primarily relies on the inertial measurement unit (IMU). It reads and stores the robot's three Euler angles in space: yaw, pitch, and roll. The yaw angle defines the robot's horizontal orientation, the pitch angle reflects its heel-to-tail tilt, and the roll angle reflects its left-to-right tilt. These three angles uniquely determine the robot's instantaneous spatial orientation relative to the horizontal reference plane and north. The recorded attitude data is bound to the position data and stored together in a specific area of ​​non-volatile memory, often labeled "task interruption scene" or similar.

[0056] Suppose an underwater robot is performing a bow-shaped path observation of the front of a bridge pier. When it reaches point A on the path, its side-scan sonar identifies a corrosion area on its right that meets certain criteria. The robot's motion controller stops executing commands to proceed to the next preset waypoint. Instead, the motion controller activates a position-keeping algorithm, fine-tuning the thruster output to allow the robot to smoothly transition from cruising mode and eventually hover stably in the water near point A.

[0057] After the robot achieves stable hovering, it immediately performs state recording. Precise positioning information is obtained from the integrated navigation system, recorded in the local task coordinate system as "Position: X=102.5 m, Y=56.8 m, Depth Z=15.2 m". Simultaneously, attitude information is obtained from the inertial measurement unit, recorded as "Attitude: Heading angle = 90.0 degrees (due east), Pitch angle = 0.5 degrees (slight bow pitch), Roll angle = -0.2 degrees (slight starboard roll)". This complete set of position and attitude data is packaged and saved as the critical state of "Interruption Scene A" to the robot's solid-state memory.

[0058] It should be noted that if the judgment result of step S3 is negative, that is, if there is no target area, the underwater robot continues to perform fixed-distance observation according to the main navigation route.

[0059] S5. Determine the observation point based on the interruption location and target area, and control the underwater robot to actively observe the target area.

[0060] The observation point is a spatial location near the target area, and its selection must meet several conditions. The point should be located in a position that the robot can safely reach and easily maintain stability in; the robot's primary observation sensor (such as a forward-facing high-resolution sonar or optical camera) should cover the target area at a near-vertical or optimal angle of incidence; and the distance between the point and the target area should be within the optimal operating range of the primary observation sensor. The specific method for determining the observation point typically involves geometric calculations. For example, based on the spatial coordinates of the target area and its geometric extension direction, and considering the robot's dimensions and safety margins, one or more candidate point coordinates are calculated at the optimal observation distance along the normal direction or the perpendicular direction of the principal feature direction of the target area, and one of these is selected as the final observation point.

[0061] Subsequently, the robot is controlled to move towards the observation point and conduct active observation. Based on the location of the interruption point, the robot's current attitude, and the coordinates of the final observation point, a local path is planned. This path typically consists of a straight line or a gentle curve to ensure safe maneuverability. The robot autonomously navigates along this local path to the observation point. Upon arrival, the robot first performs precise attitude adjustments. Its control system calculates the required heading and pitch angle corrections based on the target area's position relative to the observation point, and uses the thrusters to precisely align the robot's main observation sensor axis with the center of the target area.

[0062] After stabilizing in the correct position and orientation, the robot initiates an active observation program. This includes controlling the main observation sensor to operate in specific modes, such as forward-facing sonar performing high-density fan-shaped scanning or multi-beam measurement to acquire detailed distance information and two-dimensional images of the target area; or controlling the optical camera to take pictures from multiple angles. To further construct a three-dimensional model of the target area, the robot is controlled to perform small-range translations or slow rotations around a specific axis near the observation point, thereby collecting data from multiple perspectives. Throughout the active observation process, the acquired raw data, along with the target area's unique identifier, timestamp, and pose information at the time of acquisition, are synchronously packaged and stored.

[0063] For example, see Figure 3 The diagram shown illustrates the principle of an underwater robot observing a target area.

[0064] The underwater robot interrupted its inspection at point A on the front of the bridge pier and discovered a corrosion target on its right side. Based on the geometric center and approximate outline of the corrosion patch, the direction of its surface normal was calculated. Along this normal direction, an observation point C was set approximately 1.5 meters from the center of the patch (this is the preferred working distance for forward-facing high-definition sonar). Point C is located in the water outside the side of the bridge pier.

[0065] A straight path from point A to point C was then planned. The robot turned and navigated along this path to point C. Upon arrival, the robot adjusted its heading and pitch angles based on the coordinates of the corrosion patch center relative to point C, ensuring its forward sonar beam was perpendicularly aligned with the central region of the patch. After adjustment and stable hovering, the robot activated its forward sonar to perform a fine scan, combining this with the main observation sensor to obtain millimeter-resolution images and distance data of the corrosion area. To obtain 3D information, the robot also slowly changed its pitch angle while maintaining its position at point C, performing several scans from different perspectives. All scan data was labeled "Corrosion Area_001" and stored, completing this active observation task.

[0066] S6. After completing the observation of the target area, control the underwater robot to return to the interrupted position and restore the interrupted attitude, and continue to perform fixed-distance observation of the object under test along the main navigation path.

[0067] In this process, the underwater robot retrieves the interruption location coordinates and interruption attitude angle data recorded in step S4 from non-volatile memory. Based on the robot's current observation point position and the interruption location, the system plans a return path. This path planning takes into account both efficiency and safety, and is typically a straight line connecting the two points or a smooth curve that takes obstacle avoidance into account.

[0068] The underwater robot is controlled to navigate along a planned return path until it reaches the vicinity of the interruption point. Upon arrival, the robot does not immediately resume navigation but performs a precise pose return operation. The robot's current actual position is compared with the recorded interruption point, and the thruster output is fine-tuned to ensure that the spatial error between the robot's positioning reference point and the interruption point coordinates is less than a preset tolerance range, such as within a few centimeters. Next, attitude recovery is performed. The robot's current real-time measurements of heading, pitch, and roll angles are compared with the recorded interruption attitude angle data. Through the attitude control loop, the thrusters are driven to adjust the robot's orientation and level until the deviation of each attitude angle from the recorded values ​​is less than the set angular tolerance, such as within 1 degree. This process ensures that the robot not only returns to its original position but also restores the exact same spatial orientation as at the moment of interruption.

[0069] After the position and attitude are restored, the main control system resends the task continuation command to the motion controller and the range observation controller. The waypoint that was interrupted in the main navigation path is reset as the current target point. The motion controller recalculates the path tracking command based on the robot's current position and attitude. At the same time, the forward sonar ranging device is reactivated, and the range observation controller starts working again, controlling the robot to maintain a preset observation distance with the object under test based on real-time distance feedback. Thus, the underwater robot has achieved a seamless switch from state to task, continuing to perform range observation operations after the interruption point along the original main navigation path as if the interruption event had never occurred.

[0070] For example, after the underwater robot completes the observation of corrosion points on the side of the bridge pier, it is located at observation point C. The previously recorded interruption site data is retrieved: the coordinates of the interruption location A are (X=102.5 meters, Y=56.8 meters, depth Z=15.2 meters), and the interruption attitude is (heading angle 90.0 degrees, pitch angle 0.5 degrees, roll angle -0.2 degrees).

[0071] Plan a straight path from point C back to point A. The robot travels along this path, and upon reaching the vicinity of point A, it activates the precise position-keeping mode, making minor adjustments until the coordinates displayed by its positioning system match the recorded values. Subsequently, the robot adjusts its attitude, gradually correcting its heading angle from the current orientation back to 90.0 degrees, adjusting the pitch angle to 0.5 degrees, and the roll angle to -0.2 degrees.

[0072] Once the position and attitude have been restored, the main navigation task is reactivated. The robot faces the pier again, uses its forward-facing sonar to determine the current distance, and begins fixed-distance control. Starting from the next waypoint after point A, the robot continues to execute the previously interrupted bow-shaped path, moving horizontally and maintaining a fixed-distance observation of the pier surface at a distance of 2 meters, allowing subsequent inspection tasks to proceed continuously.

[0073] See Figure 4 This is a schematic diagram of the process for controlling the observation perspective of an underwater robot provided in an embodiment of this application.

[0074] S5. Determine observation points based on the interruption location and target area, and control the underwater robot to actively observe the target area, including: S51. Obtain the first distance information between the interruption location and the target area; S52. Estimate the travel position based on the first distance information and the set observation distance; S53. Control the underwater robot to move to the travel position and keep it hovering; S54. Adjust the underwater robot's observation attitude and position according to the set observation distance so that the target area is within the underwater robot's observation field of view.

[0075] In step S51, the interruption location has already been recorded and stored as coordinate data in step S4. The location information of the target area is obtained in step S3 by analyzing the sonar image, specifically including the coordinates of the geometric center or feature points of the area in the global or robot-relative coordinate system. Obtaining the first distance information involves calculating the spatial straight-line distance between these two points. Based on the three-dimensional coordinates of the interruption location and the three-dimensional coordinates of the target area, the Euclidean distance formula is directly applied for calculation. The coordinate data required for the calculation all come from the fusion output of the robot navigation system and the sonar imaging system, ensuring that the data has consistency in a unified coordinate system.

[0076] For example, the coordinates of the interruption location A in the local task coordinate system are (10 meters, 5 meters, depth 20 meters). The target area, analyzed by sonar imagery, has its geometric center determined to be at (13 meters, 9 meters, depth 20.5 meters) in the same coordinate system. The system calculates the straight-line distance between the two points in three-dimensional space, i.e., the first distance information, which is approximately 5 meters.

[0077] In step S52, the set observation distance is a fixed value or adjustable parameter pre-determined based on the optimal operating range of the main observation sensor, such as 2 meters. The goal of estimating the travel position is to find a spatial point from which observation of the target area meets the set observation distance requirement while also ensuring the robot's safe access. The estimation process is based on geometric principles. Typically, using the target area location as a reference point, a candidate point is calculated based on the difference between the first distance information and the set observation distance, along the vector direction from the interrupted position to the target area or its opposite direction. A more common strategy is to directly use the target area location as the center and, while ensuring the set observation distance, calculate a travel position coordinate in an orientation that the robot can safely approach (such as the direction of the surface normal of the target area). This calculation also needs to consider the robot's dimensions and safe operating margin.

[0078] Continuing the previous example, the first distance information is 5 meters, and the observation distance is set to 2 meters. Using the target area location as a reference, along a direction roughly from the interruption point A towards the target area, a travel position point B is estimated at a distance of 2 meters from the target area. Through vector calculation, the coordinates of point B are likely to be approximately (12.2 meters, 8.2 meters, depth 20.2 meters). This point lies on the line connecting point A and the target area, approximately 2 meters from the target area.

[0079] In step S53, based on the robot's current position (i.e., near the interruption position) and its current travel position, an efficient and safe motion trajectory is planned, typically a straight line or a gentle curve. The underwater robot's propulsion control system receives this trajectory command and drives each thruster to work in coordination, enabling the robot to move along the planned trajectory towards the current travel position. Throughout the movement, the integrated navigation system provides real-time pose feedback to achieve closed-loop control. When the robot reaches the vicinity of the current travel position, the control system switches to a high-precision position-holding mode, continuously fine-tuning the thruster output to counteract water flow disturbances, allowing the robot to hover stably at the current travel position with the position error controlled within a preset range (e.g., centimeter level), preparing for subsequent fine adjustments.

[0080] The robot starts from point A (the point of interruption), and the motion controller plans a straight path to point B (the point of travel). The robot activates its thrusters and travels along this straight line. Real-time navigation data shows that the robot is gradually approaching the coordinates of point B. When the distance and position errors are within a threshold, the controller activates hovering mode, and the robot stabilizes in the water near point B, preparing for the next adjustment.

[0081] In step S54, based on the hovering position, the underwater robot performs fine adjustments to ensure optimal observation. The adjustments are divided into observation position fine-tuning and observation attitude adjustment. First, based on the precise contour of the target area and the set observation distance, the robot's hovering position is fine-tuned so that the actual distance between the robot and the feature points of the target area precisely matches the set observation distance. Then, the observation attitude is adjusted. The azimuth and pitch angles of the target area relative to the robot's current position are calculated. Based on these calculated angle values, the attitude controller drives the robot to change its heading and pitch angles, ensuring that the axis of the onboard forward-facing sonar or optical camera, or other main observation sensor, is precisely aligned with the center or key parts of the target area. During the adjustment process, the sensor may provide real-time image feedback to assist in precise alignment. The ultimate goal is to ensure that the target area falls completely within the center of the main observation sensor's field of view, while the robot maintains the set observation distance, thereby creating optimal conditions for acquiring high-quality observation data.

[0082] For example, after hovering at point B, the actual distance to the target area was detected to be 2.1 meters, slightly greater than the set observation distance of 2 meters. Therefore, the robot was controlled to slowly move forward 0.1 meters. After fine-tuning the position, calculations showed that the target area was located to the lower right of the robot's current orientation. The attitude controller then adjusted the robot's heading angle to the right by 5 degrees, and simultaneously adjusted the pitch angle to the downwards by 3 degrees. After the adjustments were completed, the real-time image from the forward-facing sonar showed that the target corrosion area was now centered in the image, and the sonar ranging showed a distance of 2.0 meters, meeting all observation conditions. The robot maintained this state and initiated a high-definition scan.

[0083] This technical solution enables underwater robots to autonomously, efficiently, and accurately deploy to optimal observation points based on the interrupted scene and identified targets, and establish a stable observation posture. This not only ensures detailed observation quality of unexpectedly discovered target areas but also improves the reliability and consistency of the entire system's response to abnormal events through standardized operating procedures.

[0084] See Figure 5 This is a schematic diagram of the hovering control process provided in the embodiments of this application.

[0085] S53. Control the underwater robot to move to the travel position and keep it hovering, including: S531. Plan an unobstructed travel path based on the coordinates of the interruption location, the travel location, and the underwater environment; S532. Control the underwater robot to move along the travel path in an anti-current posture; S533. Once the travel position is reached, maintain the position in an anti-current posture.

[0086] In step S531, the inputs include the three-dimensional coordinates of the interruption position recorded in step S4, the three-dimensional coordinates of the travel position estimated in step S52, and underwater environmental information. Environmental information can be derived from prior electronic nautical charts, real-time environmental maps constructed during the initial scanning phase of this mission, or obstacle data obtained in real-time by the obstacle avoidance sonar onboard the robot before maneuvering. Based on this information, a safe and efficient navigation path is searched in three-dimensional space, starting from the interruption position and ending at the travel position. Safety is paramount, as the path must maintain a distance of at least a safe radius from all known or real-time perceived obstacles (such as structures, reefs, and aquatic vegetation). Efficiency typically means minimizing path length or energy consumption while meeting safety constraints. Common planning methods include applying the A* search algorithm to a three-dimensional grid map or generating smooth paths based on spline curves. The final result is a smooth, unobstructed trajectory composed of a series of ordered spatial waypoints.

[0087] For example, the underwater robot stops at point A on the front of the bridge pier, and its travel position at point B is located on the side of the pier. The planning module loads an existing 3D model of the bridge pier as a static obstacle. Simultaneously, the forward-looking obstacle avoidance sonar detects a suspended fishing net between points A and B in real time. Taking into account both the static bridge pier model and the dynamic fishing net obstacle, a 3D curved path is calculated, starting from point A, first moving slightly outward away from the bridge pier to avoid the fishing net, and then heading towards point B. All points on this path maintain a safe distance of at least 1 meter from the bridge pier surface and the fishing net.

[0088] In step S532, after receiving the planned travel path, the underwater robot proceeds along the path. During operation, it adjusts its attitude to an anti-current attitude based on the real-time underwater environment. An anti-current attitude typically refers to adjusting the robot's heading and pitch angles so that its longitudinal axis forms a certain angle of attack with the predicted or measured direction of the bottom current, or directly maintaining the robot in a low-drag attitude that is most conducive to resisting the current and reducing lateral movement, such as facing the head or tail of the underwater robot towards the main direction of the current.

[0089] The robot's attitude setting is integrated with path tracking. Based on the current real-time position and the positions of the next target waypoint on the path, the tracking error is calculated, and integrated thrust commands for each thruster are generated by considering the overall attitude requirements. The thrusters execute these commands, driving the robot to move along the predetermined path. Simultaneously, continuous closed-loop control dynamically adjusts the attitude and position to counteract the drift and rotational torque caused by the water flow, ensuring that the robot can accurately move towards the predetermined three-dimensional trajectory.

[0090] Continuing the previous example, the robot moves from point A to point B along the planned curve. Doppler log data detects a lateral water flow from left to right. It is then calculated that to counteract this flow, the robot's heading angle needs to be skewed 10 degrees to the left to maintain tracking of the predetermined route. Throughout the journey, this yaw angle and the thrust of each thruster are dynamically adjusted based on changes in the water flow and path deviations, ensuring the robot consistently moves stably along the curve that avoids the fishing net.

[0091] In step S533, when the underwater robot's navigation system determines that the spatial error between its current position and its travel position is less than a preset arrival threshold (e.g., 0.2 meters), it determines that it has reached the travel position and then switches from path tracking mode to high-precision position holding mode.

[0092] In this mode, the control objective is no longer tracking moving waypoints, but rather simultaneously locking the robot's current position and the set anti-drift attitude. The position-keeping controller uses the traveling position coordinates as the setpoint and the robot's real-time position as feedback, calculating the anti-drift thrust required to maintain the position using a proportional-integral-derivative (PID) control algorithm. The attitude-keeping controller, on the other hand, uses the current attitude as the setpoint and the real-time attitude angle as feedback, calculating the anti-rotational torque required to maintain the attitude.

[0093] The output commands from the two controllers are fused and sent to the propulsion system. Based on these fused commands, continuous and minute thrust adjustments are made to actively counteract any displacement and attitude changes in the robot's six degrees of freedom caused by environmental disturbances such as water flow and waves. This enables the robot to hover stably in its traveling position for a long period of time and maintain a fixed orientation that is conducive to subsequent observation.

[0094] For example, after the robot reaches point B, the position-holding mode is activated. The coordinates of point B and the previously used 10-degree leftward heading angle are simultaneously set as control targets. If a new current causes the robot to drift 5 centimeters to the right and its heading angle to decrease by 2 degrees, these errors will be immediately detected. The robot will quickly calculate and instruct the left thruster to briefly increase thrust to push the robot back to its original coordinates at point B, while simultaneously instructing the tail vertical rudder to fine-tune to restore the 10-degree leftward heading angle. Through this continuous fine-tuning, the robot can stably hover at point B in complex water currents.

[0095] See Figure 6 This is a schematic diagram of the process for controlling the spatial position of an underwater robot according to an embodiment of this application.

[0096] S54. Adjust the underwater robot's observation attitude and position according to the set observation distance to ensure that the target area is within the underwater robot's observation field of view, including: S541. Adjust the underwater robot's observation angle according to the coordinate information of the target area to ensure that the target area is within the observation range of the underwater robot; S542. Adjust the underwater robot to a horizontal attitude and obtain the second distance information between the underwater robot and the target area; S543. Adjust the spatial position of the underwater robot based on the second distance information and the preset observation distance.

[0097] In step S541, the observation viewpoint is adjusted, i.e., the robot's heading angle is adjusted, so that the pointing center of its onboard main observation sensor (such as a forward-facing sonar) is aligned with the target area. First, the relative coordinates of the center point or main feature points of the target area in the robot's current coordinate system are obtained. This is typically accomplished through coordinate transformation, converting the global coordinates or sonar image coordinates of the target area, combined with the robot's current precise position and attitude, into a local coordinate system with the robot as the origin. Next, the horizontal azimuth angle of this relative coordinate is calculated, i.e., the left and right deviation angle of the target area relative to the robot's bow. This angle is the heading angle increment or target heading angle that needs adjustment. The attitude controller receives this angle command and drives the robot to rotate around its vertical axis by controlling the robot's bow thrusters or control surfaces until its actual heading angle matches the calculated target heading angle. During the adjustment process, the forward-facing observation sensor provides real-time image feedback to assist in achieving precise alignment, ensuring that the target area appears in the center of the sensor's field of view.

[0098] For example, when the underwater robot hovers at position B, its forward-facing sonar's initial field of view is not fully aligned with the corroded area to its side. Based on the known coordinates of the corroded area and the robot's own coordinates, the area is calculated to be approximately 15 degrees to the right of the robot's current heading. The attitude controller then instructs the port thruster to advance slightly and the starboard thruster to brake slightly, causing the robot's bow to slowly rotate 15 degrees to the right. During the rotation, the forward-facing sonar image shows the corroded area gradually moving towards the center of the screen. The rotation stops when the heading angle is adjusted and the target is centered in the sonar image.

[0099] In step S542, after aligning the heading angle with the target, the robot needs to be adjusted to a horizontal attitude to ensure that the reference axis for distance measurement is consistent with the sensor axis, avoiding measurement errors caused by attitude tilt. A horizontal attitude mainly refers to the robot's pitch and roll angles being close to zero degrees, meaning the robot's main coordinate system horizontal plane is parallel to the geographic horizontal plane. The attitude controller reads the real-time pitch and roll angles fed back by the inertial measurement unit. If their absolute values ​​exceed the allowable threshold (e.g., 0.5 degrees), it generates control commands to drive the corresponding thrusters (usually vertical or lateral thrusters) to adjust the robot's pitch and roll until both pitch and roll angles stabilize within the threshold range.

[0100] After the robot reaches and maintains a stable horizontal posture, it acquires a second distance information using a forward-facing sonar ranging device. At this point, the sonar beam is stably aligned with the target area due to the horizontal posture and the adjusted heading angle. The sonar emits detection pulses and receives echoes from the surface of the target area. The signal processing unit calculates the straight-line distance between the surface of the robot's sonar transducer and the reflection point of the target area based on the round-trip time of the sound wave and the speed of sound in water. This distance is the second distance information. This distance value is more accurate and reliable than the distance roughly estimated at the previous travel position.

[0101] Following the previous example, after the robot adjusted its course, the inertial measurement unit showed a 1-degree bow roll and a 0.8-degree port roll. The attitude controller instructed the stern thruster to make minor adjustments to raise the bow, and simultaneously instructed the starboard thruster to make minor adjustments to correct the port roll. After several seconds of adjustment, both the pitch and roll angles stabilized within 0.2 degrees. Subsequently, the forward-facing sonar emitted ranging sound waves towards the already aligned corroded area, and the precise straight-line distance was calculated to be 2.3 meters based on the echo time; this is the second distance information.

[0102] In step S543, the preset observation distance is the optimal working distance required for the main observation sensor to acquire high-quality data; it is a known fixed value or range. The second distance information obtained in step S542 is compared with this preset observation distance to calculate the distance error. If the second distance is greater than the preset observation distance, the robot needs to be controlled to move closer to the target area; if it is less than the preset distance, the robot needs to be controlled to move away from the target area.

[0103] The adjustment process is executed by the position controller. Based on the calculated distance error value and its trend, combined with current environmental disturbances, the controller generates micro-motion control commands for the robot's forward or backward thrusters. The thrusters output precise thrust according to the commands, driving the robot to perform a small-range linear translation on the horizontal plane along the currently aligned line of sight (i.e., the robot's heading). Throughout the translation, the real-time distance is continuously monitored via forward sonar, forming a closed-loop control system. When the monitored real-time distance reaches a point where the error with the preset observation distance falls within an allowable range (e.g., ±0.1 meters), the position controller stops the translation command and may switch to a precise position-holding mode, stabilizing the robot in a new spatial position that satisfies the optimal observation distance.

[0104] For example, the preset observation distance requirement is 2.0 meters, while the second distance measurement is 2.3 meters, resulting in an error of 0.3 meters. The position controller instructs the main thrusters to propel the robot slightly forward, causing it to slowly move forward along the direction currently aligned with the corrosion area. During this movement, the forward-facing sonar continuously provides real-time distance feedback. When the real-time distance decreases to 2.05 meters, the controller determines that it has entered the allowable range, stops propulsion, and initiates position holding. Ultimately, the robot stabilizes at a position approximately 2.05 meters from the corrosion area.

[0105] In one possible embodiment of this application, it further includes: At the observation position, the underwater robot's heading angle and / or pitch angle are adjusted to acquire multi-view image information, and the target area is 3D modeled by combining the distance data corresponding to each image information, and the 3D modeling data is uploaded to the management device.

[0106] The underwater robot hovers stably at a position that meets the preset observation distance, aligns itself with the target area, collects multi-view data through controllable attitude adjustment, constructs a three-dimensional model of the target area based on this data, and finally completes the data transmission.

[0107] First, multi-view image information and corresponding distance data are acquired. The underwater robot maintains its current position essentially unchanged, and through its attitude control system, it sequentially adjusts its heading angle, pitch angle, or a combination of both, thereby changing the pointing angle of its onboard high-definition camera or forward-scanning sonar imaging sensors. After each attitude adjustment, the system controls the imaging sensor to acquire a frame of image information of the target area; this image can be an optical photograph or an acoustic image. Simultaneously, the forward-scanning sonar ranging device works synchronously, accurately measuring and recording the real-time straight-line distance between the center point of the imaging sensor lens or sonar transducer and the observed point on the target area surface. Each imaging is strictly bound to the precise distance data and the robot's precise pose data at that moment, forming a complete multi-view observation data package.

[0108] Subsequently, 3D modeling is performed. These data packets carrying angle, distance, and pose information are sent to the robot's onboard data processing unit or uploaded to the surface control station for calculation via a high-speed underwater acoustic communication module. The 3D modeling algorithm uses this data to perform 3D reconstruction based on multi-view geometry principles. Specifically, the algorithm identifies the same feature points in images from different viewpoints and uses the two-dimensional pixel coordinates of these feature points in the image, the corresponding imaging sensor attitude angle, and known sensor intrinsic parameters, combined with the precise distance data corresponding to each viewpoint as scale constraints, to calculate the true coordinates of these feature points in 3D space. By fitting the spatial coordinates of a large number of feature points, a 3D point cloud model describing the surface geometry of the target area can be generated, which can be further converted into a 3D mesh model or a surface model.

[0109] Finally, data is uploaded. After completing the 3D modeling, the generated 3D model data, the associated original multi-view data packets, the unique identifier of the target area, the modeling timestamp, etc., are packaged according to a predetermined data encapsulation format. The underwater robot transmits the data packets to surface management equipment or shore-based control centers via its onboard underwater acoustic communication unit or fiber optic microcable. The transmission process may employ reliable data transmission protocols to ensure data integrity.

[0110] For example, the underwater robot has stabilized at a distance of 2.0 meters from the corroded area of ​​the bridge pier. First, the robot maintains its current position and, through attitude control, slowly rotates on the horizontal plane, taking a high-resolution optical photograph and recording the corresponding distance every 15 degrees of heading angle, completing one full horizontal rotation around the target. Then, the robot returns to its initial heading, adjusts its pitch angle, and takes another set of photographs at different pitch angles (e.g., -10 degrees, 0 degrees, +10 degrees), recording the distances. This process yields more than ten sets of images from different perspectives and precise distance data.

[0111] These data are processed in real time. The modeling algorithm identifies common feature points at the edges of the corrosion patches in all photos. Using the robot's heading and pitch angles at the time of taking the photos, as well as the 2.0-meter distance benchmark corresponding to each photo, the three-dimensional spatial position of these edge feature points is calculated, thereby constructing a three-dimensional point cloud model of the depth and extent of the corrosion area, clearly showing its three-dimensional morphology.

[0112] After the model is built, the robot sends a data packet containing the 3D point cloud data and the identifier "corrosion area_001" to the control computer on the surface workboat via the underwater acoustic communication link to complete the information reporting.

[0113] This technical solution effectively overcomes the limitations of single-view underwater observation by actively adjusting the attitude to acquire multi-view binding distance data, providing a sufficient and scale-constrained data source for 3D reconstruction. The generated 3D model can intuitively and quantitatively reflect the three-dimensional morphology and size of the target area (such as crack depth and corrosion pit volume), greatly improving the accuracy and assessment value of disease detection.

[0114] See Figure 7 This is a flowchart illustrating the interrupt control process provided in an embodiment of this application.

[0115] S4. When a target area exists, the underwater robot interrupts the observation of the object under test and records the current interruption position and interruption posture, including: S41. When a target region exists, calculate the actual distance between the currently identified target region and any labeled target region in the database. S42. Calculate the similarity between the contour features of the target region and the contour features of the labeled target region; S43. If the actual distance is greater than or equal to the preset distance threshold, or the similarity is lower than the preset similarity threshold, then the target area is determined as a new target area, and the observation of the object under test is interrupted, and the current interruption position and interruption posture are recorded.

[0116] In step S41, the underwater robot extracts the geometric center coordinates or other feature point coordinates of the target area from the current identification results. Simultaneously, it accesses a database storing information about the observed target areas, either locally or via a communication link. This database records a unique identifier for each labeled target area and its corresponding spatial coordinates, typically the center position recorded during detailed observation and modeling.

[0117] For each marked target region record in the database, the spatial straight-line distance between its coordinates and the coordinates of the current target region is calculated. This is a Euclidean distance calculation in three-dimensional space, which needs to consider the depth dimension. During the calculation, it must be ensured that all coordinates are in the same global or local coordinate system, which is uniformly provided and maintained by the underwater robot's navigation system. The system will traverse all relevant marked target records in the database, calculate a series of actual distance values, and usually record the minimum value as a key judgment criterion.

[0118] For example, during an underwater robot's inspection, a suspected corrosion area is newly identified at its current location. The calculated geometric center coordinates are (X=150 m, Y=80 m, depth Z=22 m). A database query reveals three recorded corrosion targets with coordinates Target_A (149 m, 85 m, 22.1 m), Target_B (155 m, 70 m, 21.5 m), and Target_C (160 m, 90 m, 23 m). The system calculates the distances between the new target and these three targets sequentially: approximately 5.1 meters to Target_A, approximately 11.2 meters to Target_B, and approximately 14.1 meters to Target_C. The minimum actual distance is 5.1 meters (to Target_A).

[0119] In step S42, contour feature similarity calculation can be performed in parallel, either simultaneously with or after the spatial distance calculation. Contour features are a set of quantized parameters extracted from sonar images or 3D point clouds that describe the shape and structure of the target region. For a newly identified target region, the system extracts its contour features, such as the vertex sequence after polygon approximation, area, perimeter, aspect ratio, direction of the minimum bounding rectangle, or more advanced feature descriptors such as Hu moments.

[0120] The system retrieves the contour feature data of the labeled target region that is spatially closest to the current target region (i.e., the region or regions corresponding to the minimum actual distance calculated in S41) from the database. Next, a specific similarity metric algorithm is used to calculate the similarity between the two sets of features. One common method is to calculate the cosine similarity or the inverse of the Euclidean distance between the feature vectors; another method is to compare key shape parameters (such as area ratio, aspect ratio, etc.). The result is a similarity score between 0 and 1, with a higher score indicating greater contour similarity.

[0121] Continuing the previous example, the new target region is spatially closest to Target_A. Extracting the contour features of the new target, its sonar image shows it as an approximately elliptical patch with a calculated area of ​​0.12 square meters and a major axis of 30 degrees. Retrieving the contour features of Target_A from the database, it is an approximately elliptical patch with an area of ​​0.11 square meters and a major axis of 28 degrees. The area ratio of the two is calculated to be approximately 1.09, and the directional difference is 2 degrees. Based on a preset similarity calculation formula (e.g., considering both the closeness of the area ratio to 1 and the directional difference), the contour feature similarity between the two is calculated to be 0.88 (out of 1).

[0122] In step S43, the minimum actual distance calculated in S41 is compared with a preset distance threshold. This distance threshold defines the spatial tolerance range for determining whether two targets are the same entity; for example, it can be set to 3 meters. Simultaneously, the contour feature similarity calculated in S42 is compared with a preset similarity threshold. This threshold defines the minimum requirement for shape similarity; for example, it can be set to 0.7.

[0123] If the minimum actual distance is greater than or equal to the distance threshold, it indicates that the new target is spatially far from all known targets and may be an independent new target. Alternatively, if the contour feature similarity is lower than the similarity threshold, it indicates that even if they are spatially close, significant shape differences suggest they are likely not the same target, but a neighboring new target. As long as either of these two conditions is met, the currently identified target region is ultimately determined to be a new target region.

[0124] When a new target area is identified, the system triggers, interrupts, and records the following steps: The main navigation task is paused, the robot is stabilized, and the current interruption position coordinates and interruption attitude angle are accurately recorded for subsequent active observation. Conversely, if the minimum actual distance is less than the distance threshold and the contour feature similarity is higher than or equal to the similarity threshold, the system determines it to be the same marked target area, does not trigger an interruption, and the underwater robot continues to perform its tasks along the main navigation path.

[0125] For example, the preset distance threshold is 3 meters, and the preset similarity threshold is 0.7. Continuing with the previous example, the minimum actual distance between the new target and the nearest known target Target_A is 5.1 meters (greater than 3 meters), and the contour similarity is 0.88 (greater than 0.7). Since the condition of "actual distance greater than or equal to the distance threshold" is met, the system determines that this target is a new target area. Subsequently, the robot immediately interrupts its inspection of the bridge pier at its current position, hovers stably, records its precise position and attitude at this moment, and initiates a detailed observation process for this new corrosion point.

[0126] This technical solution can improve the intelligence level and operational efficiency of underwater robot autonomous inspection, ensure that system resources are concentrated on the detection and recording of newly discovered targets, avoid ineffective repeated observation of the same diseased area, and thus achieve more efficient and accurate automated monitoring in complex underwater environments.

[0127] It should be noted that after completing the calculations and comparisons in steps S41 and S42, if the judgment conditions simultaneously satisfy "the actual distance is less than the preset distance threshold" and "the similarity is not lower than the preset similarity threshold", then the underwater robot determines that the currently identified target area and a certain marked target area in the database are the same entity. At this time, no task interruption command will be triggered, and the current identification result will be regarded as a repeated identification or re-passing of a known target.

[0128] Specifically, the robot records the identification event and related comparison data (such as the matched target ID, calculated actual distance, and similarity) as a log entry, which can be optionally stored in the operation log for later task analysis and data traceability. Subsequently, the computing resources and cached data temporarily allocated for this identification are cleared, and control is completely returned to the main navigation task module.

[0129] The underwater robot will continue its original operational state as defined in step S1. It will continue to receive instructions from the main navigation path sequence, controlling the robot to navigate along the preset path; the ranging observation controller will continue to maintain the set observation distance between the robot and the object under test based on the ranging feedback from the forward sonar. The side-scan sonar imaging and the real-time judgment in step S3 will continue to operate normally, unaffected by this event, preparing to process the next frame of new sonar data.

[0130] In one possible embodiment of this application, the main navigation path consists of at least two alternating segments of first navigation lines and at least two segments of second navigation lines; the first navigation lines are used to provide a depth direction of travel and a reference travel distance in the depth direction of travel; the second navigation lines are used to provide a reference travel direction in a fixed depth plane and a travel distance in the reference travel direction; the reference travel directions provided by two adjacent segments of second navigation lines are different or the depth directions of travel provided by two adjacent segments of first navigation lines are different.

[0131] The main navigation path in this technical solution is a three-dimensional spatial trajectory planned to enable the underwater robot to systematically scan objects with extended surfaces (such as bridge piers, ship hulls, and pipelines). This path is not a simple straight line or planar curve, but a composite path composed of two types of functionally defined basic navigation line segments connected in an orderly and alternating manner in space. Its core design idea is to achieve coverage scanning of the depth and horizontal dimensions through segmented control, thereby efficiently completing the observation of the two-dimensional unfolded surface of the object.

[0132] The first navigation path is specifically responsible for controlling the underwater robot's movement in the vertical direction (i.e., the depth direction). It defines the robot's "depth movement direction" (diving or surfacing) and the "reference movement distance" in that direction during its movement along this path. When the robot executes the first navigation path, its main task is to change its depth position along a direction perpendicular to the horizontal plane, switching to a new "floor" or profile for subsequent horizontal scanning.

[0133] The second navigation path is specifically responsible for controlling the underwater robot's horizontal movement within a fixed depth plane. It defines the "reference direction of travel" (such as due east, due north, or a specific horizontal heading angle) that the robot should follow during its movement along this path, as well as the "distance traveled" in that direction. When the robot executes the second navigation path, its primary task is to travel in a straight line horizontally at a constant depth, performing continuous, fixed-distance observations of the surface of the object under test at that depth profile.

[0134] The alternating connection of paths indicates that the first navigation path and the second navigation path are connected end to end in a cycle. A typical work cycle is as follows: the robot first performs a horizontal scan at depth A along a segment of the second navigation path; after that, it immediately descends (or ascends) to depth B along a segment of the first navigation path; then, it performs a horizontal scan at depth B along the next segment of the second navigation path, and so on.

[0135] The different reference directions provided by adjacent segments of the second navigation path ensure that the horizontal scanning directions are staggered at adjacent depth layers. For example, the first segment's horizontal scanning direction is 0 degrees (due north), and after completion and descent, the second segment's horizontal scanning direction changes to 180 degrees (due south). This design makes the robot's reciprocating horizontal scanning lines parallel but not overlapping on the depth projection plane, forming a coverage pattern like a "bow" or "zigzag," avoiding scanning blind spots and improving coverage efficiency.

[0136] Similarly, if two adjacent first navigation routes provide different depth directions, the path can be adjusted back and forth in the vertical direction. For example, a descending first navigation route can be executed first, followed by an ascending first navigation route, which increases the flexibility of the path when climbing or circling complex three-dimensional structural surfaces.

[0137] Please see Figure 8 As shown, in one embodiment, an underwater target observation device is provided, the device comprising: The navigation module 901 is used to perform fixed-distance observation of the object under test according to a preset main navigation path; the main navigation path is a continuous observation path planned for the object under test. The acquisition module 902 is used to acquire sonar images of the space region outside the main navigation path in real time during the observation of the main navigation path; The judgment module 903 is used to determine whether there is a target area that meets preset conditions based on the sonar image; The interrupt module 904 is used to interrupt the observation of the object under test when the judgment result of the judgment module is yes, and to record the current interrupt position and interrupt posture. The observation module 905 is used to determine the observation point based on the interruption location and the target area, and control the underwater robot to actively observe the target area; The recovery module 906 is used to control the underwater robot to return to the interrupted position and restore the interrupted posture after completing the observation of the target area, and continue to perform fixed-distance observation of the object under test along the main navigation path.

[0138] In one possible embodiment, the preset conditions include: The distance between the target area and the underwater robot is less than or equal to a distance threshold; and The disease quantification index corresponding to the image features of the target region in the sonar image exceeds a preset feature threshold.

[0139] In one possible embodiment, determining the observation point based on the interruption location and the target area, and controlling the underwater robot to actively observe the target area, includes: Obtain the first distance information between the interruption location and the target area; Based on the first distance information and the set observation distance, the estimated travel position is determined; Control the underwater robot to move to the travel position and keep it hovering; The underwater robot's observation posture and position are adjusted according to the set observation distance so that the target area is within the underwater robot's observation field of view.

[0140] In one possible embodiment, controlling the underwater robot to move to the travel position and maintain hovering includes: Based on the coordinates of the interruption location, the travel location, and the underwater environment, an unobstructed travel path is planned. Control the underwater robot to move along the travel path in an anti-current posture; Once the travel position is reached, the position is maintained in the anti-current posture.

[0141] In one possible embodiment, adjusting the observation attitude and position of the underwater robot according to the set observation distance to ensure that the target area is within the observation field of view of the underwater robot includes: The underwater robot's observation angle is adjusted according to the coordinate information of the target area to ensure that the target area is within the observation range of the underwater robot; the observation angle is the heading angle. Adjust the underwater robot to a horizontal position and obtain the second distance information between the underwater robot and the target area; The spatial position of the underwater robot is adjusted based on the second distance information and the preset observation distance.

[0142] In one possible embodiment, after the step of adjusting the observation attitude and position of the underwater robot according to the set observation distance so that the target area is within the observation field of view of the underwater robot, the method further includes: At the observation position, the underwater robot's heading angle and / or pitch angle are adjusted to acquire multi-view image information, and the target area is 3D modeled by combining the distance data corresponding to each image information, and the 3D modeling data is uploaded to the management device.

[0143] In one possible embodiment, when the target area is present, the underwater robot interrupts its observation of the object under test and records the current interruption position and interruption posture, including: When the target region exists, calculate the actual distance between the currently identified target region and any marked target region in the database; Calculate the similarity between the contour features of the target region and the contour features of the labeled target region; If the actual distance is greater than or equal to a preset distance threshold, or the similarity is lower than a preset similarity threshold, then the target region is determined as a new target region, the observation of the object under test is interrupted, and the current interruption position and interruption posture are recorded.

[0144] In one possible embodiment, the main navigation path consists of at least two alternating segments of first navigation lines and at least two segments of second navigation lines; the first navigation lines are used to provide a depth direction of travel and a reference travel distance in the depth direction of travel; the second navigation lines are used to provide a reference travel direction in a fixed depth plane and a travel distance in the reference travel direction of travel; the reference travel directions provided by two adjacent segments of second navigation lines are different or the depth directions of travel provided by two adjacent segments of first navigation lines are different.

[0145] In one embodiment, an underwater robot is provided, which can be a server-side component, and its internal structure diagram can be as follows: Figure 9As shown, the underwater robot includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface allows communication with external clients via a network connection. When executed by the processor, the computer program implements the functions or steps of an underwater target observation method.

[0146] In one embodiment, an underwater robot is proposed, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement... Figure 1 The method shown can be referred to for details. Figure 1 As shown, it will not be elaborated further here.

[0147] In one embodiment, a computer-readable storage medium is provided that stores a computer program, which is loaded and executed by a processor as described above. Figure 1 The method steps of the illustrated embodiment can be found in the following documentation for detailed execution. Figure 1 The specific details of the illustrated embodiments will not be elaborated here. It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or the underwater robot described above can be referred to the relevant descriptions on the server side and the client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0148] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0149] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0150] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for observing underwater targets, characterized in that, The method includes: The underwater robot performs fixed-distance observations of the object under test according to a preset main navigation path; the main navigation path is a continuous observation path planned for the object under test. During the observation of the main navigation path, sonar images of the spatial region outside the main navigation path are acquired in real time; Based on the sonar image, determine whether there is a target area that meets the preset conditions; When the target area is present, the underwater robot interrupts its observation of the object under test and records the current interruption position and interruption posture. Based on the interruption location and the target area, an observation point is determined, and the underwater robot is controlled to actively observe the target area. After completing the observation of the target area, the underwater robot is controlled to return to the interrupted position and restore the interrupted posture, and continue to perform fixed-distance observation of the object under test along the main navigation path.

2. The underwater target observation method according to claim 1, characterized in that, The preset conditions include: The distance between the target area and the underwater robot is less than or equal to a distance threshold; and The disease quantification index corresponding to the image features of the target region in the sonar image exceeds a preset feature threshold.

3. The underwater target observation method according to claim 1, characterized in that, The step of determining observation points based on the interruption location and the target area, and controlling the underwater robot to actively observe the target area, includes: Obtain the first distance information between the interruption location and the target area; Based on the first distance information and the set observation distance, the estimated travel position is determined; Control the underwater robot to move to the travel position and keep it hovering; The underwater robot's observation posture and position are adjusted according to the set observation distance so that the target area is within the underwater robot's observation field of view.

4. The underwater target observation method according to claim 3, characterized in that, The control of the underwater robot to move to the travel position and maintain hovering includes: Based on the coordinates of the interruption location, the travel location, and the underwater environment, an unobstructed travel path is planned. Control the underwater robot to move along the travel path in an anti-current posture; Once the travel position is reached, the position is maintained in the anti-current posture.

5. The underwater target observation method according to claim 3, characterized in that, The step of adjusting the observation attitude and position of the underwater robot according to the set observation distance, so that the target area is within the observation field of view of the underwater robot, includes: The underwater robot's observation angle is adjusted according to the coordinate information of the target area to ensure that the target area is within the observation range of the underwater robot; the observation angle is the heading angle. Adjust the underwater robot to a horizontal position and obtain the second distance information between the underwater robot and the target area; The spatial position of the underwater robot is adjusted based on the second distance information and the preset observation distance.

6. The underwater target observation method according to claim 3, characterized in that, After the step of adjusting the observation attitude and position of the underwater robot according to the set observation distance so that the target area is within the observation field of view of the underwater robot, the method further includes: At the observation position, the underwater robot's heading angle and / or pitch angle are adjusted to acquire multi-view image information, and the target area is 3D modeled by combining the distance data corresponding to each image information, and the 3D modeling data is uploaded to the management device.

7. The underwater target observation method according to claim 1, characterized in that, When the target area exists, the underwater robot interrupts its observation of the object under test and records the current interruption position and attitude, including: When the target region exists, calculate the actual distance between the currently identified target region and any marked target region in the database; Calculate the similarity between the contour features of the target region and the contour features of the labeled target region; If the actual distance is greater than or equal to a preset distance threshold, or the similarity is lower than a preset similarity threshold, then the target region is determined as a new target region, the observation of the object under test is interrupted, and the current interruption position and interruption posture are recorded.

8. The underwater target observation method according to claim 1, characterized in that, The main navigation path consists of at least two alternating segments of first navigation lines and at least two segments of second navigation lines; the first navigation lines are used to provide a depth direction of travel and a reference travel distance in the depth direction of travel. The second navigation line is used to provide a reference direction of travel within a fixed depth plane and a travel distance along the reference direction of travel; The reference running directions provided by two adjacent segments of the second navigation line are different, or the depth running directions provided by two adjacent segments of the first navigation line are different.

9. An underwater target observation device, characterized in that, The device includes: The navigation module is used to perform fixed-distance observation of the object under test according to a preset main navigation path; the main navigation path is a continuous observation path planned for the object under test. The acquisition module is used to acquire sonar images of the spatial region outside the main navigation path in real time during the observation of the main navigation path; The judgment module is used to determine whether there is a target area that meets preset conditions based on the sonar image; An interrupt module is used to interrupt the observation of the object under test when the judgment result of the judgment module is yes, and to record the current interruption position and interruption posture. The observation module is used to determine observation points based on the interruption location and the target area, and control the underwater robot to actively observe the target area; The recovery module is used to control the underwater robot to return to the interrupted position and restore the interrupted posture after completing the observation of the target area, and continue to perform fixed-distance observation of the object under test along the main navigation path.

10. An underwater robot, characterized in that, The underwater robot includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the underwater target observation method as described in any one of claims 1 to 7.

Images