Jacket foundation concrete pouring real-time monitoring system and method based on underwater robot

Abstract

The invention provides a jacket foundation concrete pouring real-time monitoring system and method based on an underwater robot. The method comprises the steps that the underwater robot generates an enhanced image through a binocular optical camera in combination with LED light supplement, calculates the ratio of a gap volume to a CAD theoretical volume based on binocular parallax, quantifies plumpness, and meanwhile detects defects through a model; the underwater robot analyzes the uniformity of the slurry through an HSV space, detects the segregation rate of the clustered slurry through edges, and collects temperature, pH and viscosity parameters; the underwater robot locks an overflow port, calculates the pixel flow rate through optical flow, converts the real speed through binocular depth, filters, and sends a pump frequency adjusting instruction to the pump station through a communication protocol according to a threshold value; and after the underwater robot finishes pouring, cruising, performing secondary scanning to verify plumpness and defects, floating, recovering and generating a report. According to the invention, real-time monitoring and comprehensive evaluation of the jacket foundation concrete pouring quality and the surrounding environment state are realized, and the controllability and engineering safety of underwater pouring operation are improved.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a real-time monitoring system and method for concrete pouring of jacket foundation based on an underwater robot. Background Technology

[0002] Underwater engineering plays a crucial role in modern marine development, especially in offshore pile foundation construction, where quality directly impacts the safety and long-term stability of the project. The importance of this field is self-evident, as pile foundations serve as the basis for supporting offshore facilities, and any quality issues can lead to catastrophic consequences. However, current underwater pile foundation construction monitoring methods are insufficient to meet the demands of the complex marine environment, necessitating breakthroughs in existing technologies to ensure the reliability and safety of projects.

[0003] Existing technical solutions generally suffer from insufficient adaptability to the complexity of underwater environments. Many methods often fail to accurately obtain critical information during construction when faced with low visibility, strong ocean currents, and varied seabed topography. This limitation is not only reflected in the real-time assessment of construction quality but also in the insufficient ability to perceive changes in the surrounding environment, making it difficult to detect and address potential risks in a timely manner.

[0004] A deeper technical challenge lies in achieving precise monitoring and management of the grouting process in an underwater environment. As a core component of pile foundation construction, the fullness of grouting directly affects the stability of the pile foundation. However, due to limitations in visualization and perception in the underwater environment, issues such as the uniformity of grouting and the presence of voids or defects are difficult to accurately identify. Furthermore, this monitoring challenge is closely related to dynamic environmental changes during construction. For example, ocean currents can erode the seabed, creating pits that affect the stability around the pile foundation. Existing technologies often cannot simultaneously address both grouting quality and the comprehensive assessment of environmental changes.

[0005] Specifically, in actual business scenarios, construction workers often face the following dilemma: when grouting is being carried out, it is impossible to visually determine whether the grout has completely filled every corner of the pile foundation, or whether there are local voids or unevenness; at the same time, changes in the seabed topography may also occur quietly during construction, such as the formation of scour pits that may weaken the support of the pile foundation, and these changes are often only discovered afterward, increasing potential engineering risks.

[0006] Therefore, how to accurately monitor the grouting quality in complex underwater environments and conduct real-time assessments of changes in the seabed environment has become a key issue in ensuring the safety and stability of offshore pile foundation construction. Summary of the Invention

[0007] This invention provides a real-time monitoring method for concrete pouring of jacket foundation based on an underwater robot, mainly including: The underwater robot is deployed to the offshore pile foundation area via the mother ship's crane and umbilical cable. After self-checking the binocular optical camera, sonar system, thruster, auxiliary sensors and infrared auxiliary imaging module, it is imported into the CAD pile foundation model for path pre-planning to obtain the pre-planned grid obstacle avoidance path. After the underwater robot integrates GPS and INS positioning and locks onto the pile foundation through forward-looking sonar, it dives to the target water depth at a gradient speed according to the pre-planned grid obstacle avoidance path, completes attitude adjustment, and hovers close to the overflow outlet. The underwater robot generates enhanced images using a binocular optical camera combined with LED supplemental lighting, and quantifies fullness by calculating the ratio of gap volume to CAD theoretical volume based on binocular parallax, while simultaneously using a model to detect defects. The underwater robot uses HSV spatial analysis to determine the homogeneity of the slurry, detects edges to cluster the slurry segregation rate, and collects parameters such as temperature, pH, and viscosity. The underwater robot locks onto the overflow outlet and calculates the pixel flow velocity using optical flow. After combining the binocular depth measurement with the actual velocity, it filters the data and sends a pump frequency adjustment command to the pumping station based on the threshold through a communication protocol. The underwater robot constructs a seabed topography model through a multibeam echo sounder and calculates the depth, volume, area and slope factor of scour pits to generate a risk index. The risk index is transmitted to the ground control station in real time, and corresponding response measures are triggered according to the judgment level of the risk index. After the underwater robot finishes casting, it cruises and performs a second scan to verify the fullness and defects before surfacing to retrieve the vessel and generate a report.

[0008] Furthermore, the underwater robot is deployed to the offshore pile foundation area via a mother ship crane and umbilical cable. After self-checking its binocular optical camera, sonar system, thrusters, auxiliary sensors, and infrared auxiliary imaging module, it is imported into the CAD pile foundation model for path pre-planning, resulting in a pre-planned grid obstacle avoidance path, including: After confirming that the sea state parameters meet the standards through the meteorological station and supporting testing equipment, the underwater robot is hoisted into the water by a crane and connected to the umbilical cable. The dual-optical camera performs distortion calibration, the forward-looking sonar performs circumferential scanning verification, the bottom sonar performs accuracy verification, the thruster performs parameter adjustment, the auxiliary sensor performs zero-point verification, and the infrared auxiliary imaging module performs functional self-test. The control station imports the CAD model and uses a unified coordinate system to plan the grid obstacle avoidance path, thus obtaining the pre-planned grid obstacle avoidance path.

[0009] Furthermore, after the underwater robot integrates GPS and INS positioning and locks onto the pile foundation using forward-looking sonar, it descends to the target water depth at a gradient speed according to a pre-planned grid obstacle avoidance path, completes attitude adjustment, and hovers near the overflow outlet, including: After initial fusion of GPS and INS, positioning accuracy is calibrated based on a unified coordinate system, and GPS is turned off if the depth exceeds a threshold. Forward-looking sonar locates the pile foundation through feature extraction, compares real-time positioning data with the pre-planned path, and dynamically adjusts the diving direction; After adjusting attitude according to the pre-planned path, the vehicle descends in a gradient manner, maintaining depth fluctuations. The deflection angle is monitored in real time using IMU data. When the deflection angle exceeds the set threshold or deviates from the pre-planned path, a compensation mechanism is triggered to adjust the propulsion torque. The underwater robot transmits video data to the ground control station and switches to infrared-assisted imaging when the sonar signal weakens. Approaching the pre-planned hovering point, PID control is used to achieve hovering stability, and video and log data are collected and uploaded.

[0010] Furthermore, the step of generating enhanced images using a binocular optical camera combined with LED fill light includes: After the binocular optical camera acquires images, it estimates the underwater global background light and underwater transmittance based on the preset underwater dark channel defogging model, and then reconstructs and generates a fog-free image. After multi-scale fusion weighting and color restoration factor processing, edge preservation is achieved through guided filtering. Adaptive stretching is applied to the image after guided filtering to obtain the final enhanced frame.

[0011] Furthermore, the method of quantifying fullness based on the ratio of gap volume calculated using binocular parallax to the theoretical volume in CAD, while simultaneously using a model to detect defects, includes: Depth data of the annular gap of the catheter holder is obtained by binocular parallax, and the actual total volume V of the annular gap of the catheter holder is obtained by integrating the depth data. actual The actual total volume is the maximum volume that the annular gap can fill; Define the theoretical volume in CAD as the total design volume V of the annular gap of the jacket structure. design The actual filling volume V of the slurry was calculated by integrating the depth data of the slurry-filled area using binocular parallax analysis. filled ; The fullness is quantified by calculating the ratio of unfilled gap volume to fullness, and the fullness directly reflects the degree of filling of the annular gap; By detecting the continuous discharge status of the slurry and the absence of air bubbles and voids, and combining the above fullness calculation results, it is determined whether the fullness error is within the allowable range. An alarm is triggered when the model detects a defect with a confidence level exceeding a threshold. The collected temperature, pH, and viscosity parameters are compared with preset qualified thresholds. Combined with the uniformity test results and segregation rate test results, the quality of the slurry is comprehensively determined to meet the standards. If any parameter exceeds the preset threshold or the comprehensive test results do not meet the requirements, a slurry quality alarm is triggered.

[0012] Furthermore, the process of locking the overflow outlet, calculating pixel flow rate using optical flow, converting the actual velocity using binocular depth, filtering, and then sending a pump frequency adjustment command to the pump station via a communication protocol based on a threshold includes: The model locks the overflow port area, and optical flow calculations eliminate outliers by selecting quantile pixel values. The speed of converting the center depth of the binocular depth map to the actual speed by combining pixel size and focal length; The flow rate value is filtered. If the pressure boosting command is continuously sent at low speed and the pump stops if the problem is not resolved, a warning is issued at medium speed and normal monitoring is performed at high speed.

[0013] Furthermore, the step of constructing a seabed topography model using a multibeam echo sounder and calculating the risk index for the generation of scour pit depth, volume, area, and slope factors includes: Multibeam echo sounding voxel filtering model registration; Based on the theoretical seabed height corresponding to the outer wall of the pile, the actual height of the terrain model is subtracted to calculate the depth of the scour pit. The scour pit volume is obtained by summing the scour pit depths. At the same time, the maximum depth of the scour pit, the area enclosed by contour lines in the region where the depth is greater than the preset depth threshold, and the local slope factor are determined. The risk index is calculated and generated. Based on the preset threshold, the risk index is divided into three judgment levels: safe, concern, and danger. The judgment level and risk index are transmitted to the ground control station simultaneously. When the danger level is reached, an emergency response command is triggered, prompting the implementation of measures such as dumping boulders or backfilling with mud.

[0014] Furthermore, after the pouring is completed, a second cruise scan verifies the fullness and defects, then the vessel is floated back to its original position and a report is generated, including: Cruise along a pre-planned cruise path curve, and check for fullness and defects by using multi-beam and optical scanning to detect leakage thresholds; Maintain a safe distance when surfacing, move to the designated recovery area according to the pre-planned recovery route, and rinse the underwater robot and check the degree of corrosion after recovery; In the post-processing stage, the terrain model is registered, a slurry flow heat map is generated, the operation path is optimized, a model analysis report is generated including slurry quality detection, fullness detection, scour risk assessment, and path execution review, and data backup is completed.

[0015] This invention provides a real-time monitoring system for concrete pouring of jacket foundations based on an underwater robot, mainly comprising: The equipment deployment and path pre-planning module is used to deploy the underwater robot to the offshore pile foundation area via the mother ship crane and umbilical cable, complete the self-check of the equipment including the binocular optical camera and sonar system, import the CAD pile foundation model and perform path pre-planning to obtain the pre-planned grid obstacle avoidance path. The positioning, diving, and hovering module is used to locate the pile foundation through GPS and INS fusion positioning and forward-looking sonar, and control the underwater robot to dive to the target water depth at a gradient speed according to the pre-planned grid obstacle avoidance path. After completing attitude adjustment, it approaches the overflow outlet and achieves stable hovering. The image enhancement, fullness quantification and defect detection module is used to generate enhanced images by combining binocular optical cameras with LED supplementary lighting, calculate the volume of the annular gap of the guide frame based on binocular parallax, quantify the fullness by the ratio of the volume to the CAD theoretical volume, and detect defects using a model. The slurry quality parameter monitoring module is used to analyze the homogeneity of the slurry through HSV spatial analysis, obtain the slurry segregation rate by combining edge detection clustering algorithm, and simultaneously collect the temperature, pH and viscosity parameters of the slurry. The overflow velocity monitoring and pump station control module is used to lock the overflow port area and calculate the pixel flow velocity through optical flow. Combined with binocular depth conversion, the actual overflow velocity is obtained. After filtering, the pump frequency adjustment command is sent to the pump station through the communication protocol according to the preset threshold. The seabed topography modeling and scour risk assessment module is used to construct a seabed topography model through a multibeam echo sounder system, calculate the depth, volume, area and slope factor of scour pits and generate a risk index, transmit the risk index to the ground control station in real time, and trigger corresponding response measures according to the judgment level corresponding to the risk index. The cruise verification, recovery, and report generation module is used to control the underwater robot to cruise and perform a second scan to verify fullness and defects after the pouring is completed, complete the surface recovery operation, and generate an analysis report containing key monitoring data.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a real-time monitoring system and method for concrete pouring of jacket foundations based on an underwater robot. Addressing the challenges of quality monitoring, complex environments, and insufficient seabed stability assessment during the concrete pouring process for jacket foundations, this invention integrates underwater robots, multi-sensor technology, and intelligent analysis methods to construct a complete solution. The invention utilizes an underwater robot equipped with a dual-optical camera, sonar system, and various sensors to achieve fully automated monitoring of the entire process, from equipment deployment, self-inspection, path planning to precise positioning. It demonstrates significant advantages, particularly in the quantification of grout fullness, defect detection, and dynamic adjustment of grout flow velocity. Simultaneously, this invention combines a multibeam echo sounder system to construct a seabed topography model, calculate scour pit parameters, and generate a risk index, effectively assessing the stability of the environment surrounding the pile foundation. Ultimately, this invention achieves precise assessment and dynamic control of grouting quality, ensuring project safety and long-term stability, and providing reliable technical support for offshore engineering. Attached Figure Description

[0017] Figure 1 This is a flowchart of the real-time monitoring method for concrete pouring of jacket foundation based on an underwater robot according to the present invention. Detailed Implementation

[0018] To further understand the content of this invention, a detailed description of the invention is provided in conjunction with the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0019] like Figure 1 The real-time monitoring method for the concrete pouring of the jacket foundation based on an underwater robot in this embodiment may specifically include: Step S1: The underwater robot is deployed to the offshore pile foundation area via the mother ship's crane and umbilical cable. After deployment, it performs a self-test, involving functional verification of the binocular optical camera, sonar system, thrusters, and auxiliary sensors. Then, the digital model of the pile foundation is imported for path pre-planning. This process is fundamental to the entire monitoring mission, ensuring the equipment can operate normally and accurately reach the target area in the complex underwater environment. Specifically, path pre-planning calculates the optimal navigation path by importing the digital model, considering factors such as water flow and obstacles, and uses algorithms to optimize the route to minimize energy consumption and time, ensuring the robot efficiently reaches the pile foundation monitoring point.

[0020] In one embodiment, a comprehensive sea condition assessment is required before deployment to ensure safe operation. This assessment includes measuring parameters such as wave height, current velocity, and underwater visibility using a weather station mounted on the mother ship. The crane will only be launched when the wave height is below a preset threshold, for example, no more than 1.5 meters, and the current velocity is within a safe range. After the crane is launched, operators check the umbilical cable connections to ensure there is no tangling or uneven stress. Subsequently, the underwater robot enters a self-test mode, verifying the functional status of each key component. During the self-test, the binocular optical camera undergoes lens cleaning and focus adjustment to ensure clear imaging; the sonar system undergoes signal strength testing to confirm its detection range and accuracy; the thruster undergoes forward and reverse rotation testing to verify stable thrust output; and auxiliary sensors such as thermometers and pressure gauges undergo initial calibration to ensure accurate data acquisition.

[0021] Step S11, regarding the specific self-check process, can be further subdivided into several sub-steps. First is the self-check of the binocular optical camera. The camera needs to capture a preset test pattern in the underwater environment. The control station analyzes the image for distortion or blurring. If a problem is found, the parameters are automatically adjusted or manual intervention is prompted. Second is the self-check of the sonar system. The forward-looking sonar needs to emit a signal and receive the echo to detect signal loss or interference, ensuring accurate identification of the pile foundation structure. Third is the self-check of the thruster. The control station sends different power commands to test the thruster's response time and stability at low, medium, and high speeds. Finally, the self-check of auxiliary sensors, such as the pressure sensor, needs to record the current water depth and compare it with a standard value. If the deviation exceeds the allowable range, recalibration is required. Through this series of self-check steps, it is ensured that the underwater robot will not affect the monitoring effect due to equipment failure in subsequent tasks.

[0022] Step S12: After completing the self-check, the underwater robot needs to import the digital model of the pile foundation for path pre-planning. The digital model of the pile foundation is typically constructed using computer-aided design technology and includes information such as the pile foundation's geometry, spatial location, and surrounding terrain features. After the control station imports this model into the underwater robot's navigation system, it uses artificial intelligence algorithms to pre-plan the path. The goal of path planning is to ensure that the underwater robot can reach the target area with the shortest distance and lowest energy consumption, while avoiding potential obstacles such as seabed rocks or debris. During the planning process, the system divides the path into multiple grid cells, calculates the probability of safe passage within each cell, and prioritizes the path segment with the highest probability.

[0023] For example, in monitoring offshore wind turbine foundations, the foundations are typically located in water depths of 20 to 50 meters, surrounded by complex seabed topography and man-made structures. By importing a digital model, the system can identify the specific location of the foundation and, combined with the underwater robot's current coordinates, plan an optimized path from the mother ship deployment point to the foundation's overflow outlet. During path planning, the system also considers the direction and intensity of ocean currents, aiming to guide the underwater robot downstream to reduce propulsion energy consumption. Furthermore, if high-risk areas exist along the path, such as underwater cables or fishing nets, the system automatically adjusts the path to ensure the robot's safe passage. This process not only improves task efficiency but also significantly reduces the risk of equipment damage.

[0024] Step S2: After the underwater robot integrates positioning technology and locks onto the pile foundation using sonar, it descends to the target water depth at a gradient velocity, completes attitude adjustment, and hovers near the overflow outlet. This stage is a crucial transitional link from deployment to the execution of monitoring tasks, involving multiple technologies such as positioning, navigation, and attitude control to ensure the robot can accurately reach the grouting area and maintain a stable state. Specifically, the underwater robot first determines its position using the fusion technology of the Global Positioning System (GPS) and Inertial Navigation System (INS). Then, it uses forward-looking sonar to detect the pile foundation structure and adjusts its descent speed and direction based on the detection results. Upon approaching the target water depth, the robot needs to work in conjunction with attitude sensors and thrusters to complete attitude adjustment, ultimately hovering near the overflow outlet.

[0025] In step S21, during positioning and diving, the underwater robot needs to integrate multiple positioning technologies to cope with the complexity of the underwater environment. Initially, the robot relies on the Global Positioning System (GPS) to acquire surface position information, while simultaneously recording its trajectory and direction changes through an inertial navigation system. When the water depth exceeds a certain threshold, such as 10 meters, the GPS signal significantly weakens, at which point the system automatically shuts down this function, relying entirely on the inertial navigation system for position calculation. The inertial navigation system collects data through accelerometers and gyroscopes, calculates the robot's displacement and orientation in real time, and performs error correction by combining historical data. Furthermore, forward-looking sonar plays a crucial role in this stage, emitting sound waves and receiving echoes to detect the contour features of the pile foundation ahead, thereby locking onto the target location.

[0026] For example, in a monitoring mission of offshore pile foundations at a water depth of 30 meters, the underwater robot begins its descent from the mother ship's deployment point, initially relying on the Global Positioning System (GPS) and inertial navigation system to determine its approximate direction. When the water depth reaches 15 meters, the GPS signal is interrupted, and the system switches to inertial navigation mode, simultaneously activating forward-looking sonar for target detection. The sonar system scans the area ahead, identifies the cylindrical structure of the pile foundation, and extracts its center point coordinates as the navigation target. Subsequently, based on the sonar feedback data, the robot dynamically adjusts its descent angle and speed to ensure it approaches the pile foundation via the shortest path. This process effectively avoids path deviations caused by low underwater visibility or positioning errors.

[0027] In step S22, during the descent, the underwater robot employs a gradient velocity control strategy to adapt to environmental changes at different water depths. Gradient velocity control refers to gradually adjusting the descent speed based on water depth and current strength to avoid attitude loss or equipment damage due to excessive speed. In shallow water areas, the robot descends at a slower speed to respond promptly to potential obstacles or emergencies; in medium-deep water areas, the speed can be appropriately increased to shorten the descent time; as it approaches the target depth, the speed is reduced again to ensure the accuracy of attitude adjustment. Furthermore, attitude sensors monitor the robot's tilt angle and deflection in real time. If depth fluctuations or deflections exceed the allowable range, the system automatically triggers the thrusters to adjust torque for compensatory control.

[0028] In one possible implementation, gradient velocity control can be divided into three stages. In the shallow water stage (0-10 meters deep), the diving speed is controlled at 0.5 meters per second to ensure safe passage through any floating objects or shallow obstacles. In the medium water stage (10-25 meters deep), the speed increases to 1.0 meter per second to improve efficiency. In the deep water stage (approaching the target depth of 25-30 meters), the speed decreases to 0.3 meters per second, and an attitude adjustment program is initiated to ensure the robot approaches the overflow outlet at the optimal angle. During attitude adjustment, the system collects roll, pitch, and yaw angle data through attitude sensors. If the roll angle deviation exceeds 5 degrees, differential torque is output through the left and right thrusters for correction. This control strategy effectively improves the stability and safety of the diving process.

[0029] In step S23, during the dive, the underwater robot needs to transmit video and sonar data to the control station in real time so that operators can monitor its status. If the sonar signal is attenuated due to underwater environmental interference, such as encountering high-density suspended objects or complex terrain, the system will automatically switch to auxiliary imaging mode. Auxiliary imaging mode typically relies on binocular optical cameras, using high-brightness supplementary lighting to illuminate the area in front and obtain clear visual information. In addition, the system will also record key log data during the dive, including parameters such as speed, depth, and attitude angles, and upload them to the control station via umbilical cable for subsequent analysis and troubleshooting.

[0030] For example, during an offshore pile foundation grouting monitoring mission, the underwater robot experienced a brief interruption in its sonar signal due to interference from seabed debris when it descended to a depth of 20 meters. Upon detecting that the signal strength was below a preset threshold, the system immediately switched to auxiliary imaging mode, activating a binocular optical camera and supplementary lighting to successfully capture the blurred outline of the pile foundation ahead. Subsequently, the operator adjusted the robot's direction via the control station to ensure it continued its descent along the correct path. Simultaneously, the system logged the signal interruption event, marking it with a timestamp and location information, providing a reference for subsequent optimization of sonar parameters. This flexible switching mechanism significantly improved the mission's robustness.

[0031] In step S24, upon approaching the overflow outlet, the underwater robot needs to achieve hovering stability to ensure the smooth progress of subsequent monitoring tasks. Hovering control relies on precise thruster adjustment and attitude feedback mechanisms. The system monitors the robot's relative position to the overflow outlet in real time using depth and attitude sensors. If a positional shift or depth fluctuation is detected, the thrusters output fine-tuning torque for correction. Simultaneously, dual-lens optical cameras continuously collect video data from the overflow outlet area and upload it to the control station for operator confirmation of the hovering effect. Furthermore, the system records key parameters during hovering, such as thruster power and attitude angle changes, generating a detailed log to provide data support for subsequent task optimization.

[0032] In one embodiment, hovering control can be achieved through a multi-point positioning strategy. The system first identifies the geometric features of the overflow outlet using a binocular optical camera, using its center point as the hovering target. Then, combining depth sensor data, it calculates the three-dimensional distance difference between the robot and the target point. If the distance difference exceeds a preset range, such as 0.2 meters, the position is adjusted via the thrusters. During hovering, if the current causes slight drift in the robot, the system adjusts the thruster output in real time based on the deflection angle feedback from the attitude sensor, ensuring the robot remains at the target position. This control method effectively addresses dynamic interference in the underwater environment, providing a stable platform for subsequent grouting quality monitoring.

[0033] In step S3, the underwater robot generates enhanced images using a binocular optical camera combined with supplementary lighting. It then calculates the ratio of the gap volume to the theoretical volume based on binocular parallax to quantify the fullness, while simultaneously using a model to detect defects. Specifically, in low-light underwater conditions, the binocular optical camera enhances image brightness using supplementary lighting, and a series of image processing steps generate a clear enhanced image. Subsequently, the system calculates the actual volume of the grouting area using the binocular parallax principle and compares it with the theoretically designed volume to derive a fullness index. Furthermore, the system uses a pre-trained defect detection model to identify potential voids, cracks, and other problems in the grouting area.

[0034] In step S31, after the binocular optical camera acquires images, it needs to go through a series of processing steps to generate enhanced images to cope with insufficient light and scattering interference in the underwater environment. First, the camera, with the support of supplementary lighting equipment, acquires the original image of the grouting area. The supplementary lighting equipment usually uses high-brightness LED lamps, with the illumination range covering the camera's field of view to ensure uniform image brightness. Subsequently, the system performs noise reduction processing on the original image to filter out noise points caused by underwater suspended objects. Next, through transmittance estimation and color restoration techniques, a fog-free image is generated to improve the image's contrast and clarity.

[0035] For example, in a pile foundation grouting monitoring task at a water depth of 30 meters, the original images captured by the binocular optical cameras appear blurry due to underwater light scattering, making it difficult to discern details in the grouting area. After the system activates the supplementary lighting equipment, the image brightness is significantly improved, but interference from suspended particles remains. Subsequently, the system filters out noise points using a denoising algorithm and restores the true colors of the image based on transmittance estimation technology, ultimately generating a clear, enhanced image. Through this processing, operators can clearly observe the surface condition of the grouting area, laying the foundation for subsequent volume calculations and defect detection.

[0036] In step S32, after generating the enhanced image, the system further optimizes the image quality using multi-scale fusion technology. Multi-scale fusion refers to decomposing the image into feature layers of different scales, processing brightness, color, and edge information separately, and then generating the final image through weighted fusion. During this process, the system pays special attention to edge preservation, using guided filtering technology to ensure that the boundary details of the grouting area are not smoothed. In addition, the system performs adaptive stretching on the image, dynamically adjusting brightness and contrast according to ambient light intensity to ensure good visualization under different underwater conditions.

[0037] In one possible implementation, multi-scale fusion processing can be divided into three levels. First, the system extracts the low-frequency layer of the image, adjusting the overall brightness and color distribution to restore the true tone of the grouting area. Second, it extracts the mid-frequency layer to enhance texture details, making subtle changes on the grouting surface more apparent. Finally, it extracts the high-frequency layer to preserve edge information and ensure clear boundaries. Subsequently, the system recombines the features from the three levels using a weighted approach and further optimizes the edge effect using guided filtering. This processing method effectively improves image quality, making subsequent volumetric calculations and defect detection more accurate.

[0038] Step S33: Based on the enhanced image, the system calculates the gap volume of the grouting area using the principle of binocular parallax. Binocular parallax refers to inferring the depth information of the target object by using the pixel offset between images captured by two cameras. The system first performs stereo matching on the binocular images to determine the depth value of each pixel within the grouting area, and then calculates the actual volume through depth integration, specifically V=∑(A i *d i ), where V is the actual volume, A i Let d be the projected area of ​​pixel i. i This corresponds to the depth value. Defines the volume of unfilled gaps. , where V actual V represents the actual total volume of the annular gap. filled The actual filling volume of the slurry is determined by the formula. Calculate the fullness ratio, where V design The total design volume of the annular gap of the jacket structure is pre-loaded from the CAD model and serves as a quantitative indicator of the grouting filling effect. If the fullness ratio is lower than the preset threshold, such as 0.9, it indicates that the grouting may be insufficient and further inspection is required.

[0039] For example, in a pile foundation grouting quality monitoring task, the system calculated the actual volume of the grouting area to be 1.8 cubic meters using binocular parallax, while the theoretical design volume was 2.0 cubic meters, resulting in a fullness ratio of 0.9, close to the critical value. Further analysis of the image data by the operators revealed a small amount of unfilled areas at the top of the grouting zone, possibly due to uneven grout flow. The system then recorded this ratio and marked the relevant locations, providing a reference for subsequent adjustments to grouting parameters. This quantitative assessment method directly reflects the grouting effect and provides an important basis for project quality control.

[0040] In step S34, the system simultaneously analyzes the enhanced image using a pre-trained YOLOv5 defect detection model to identify potential issues such as voids, cracks, or air bubbles in the grouting area. The defect detection model is typically trained on a large amount of labeled data and can identify defect features of different types and scales. During the detection process, the system divides the image into blocks, extracts features from each block, and calculates the defect confidence score. If the confidence score of a certain area exceeds a preset threshold, such as 0.8, a defect is identified in that area, and an alarm is triggered to alert the operator. Furthermore, the system records the location and type of the defect, generating a detailed detection log.

[0041] In one embodiment, defect detection can be optimized for different grouting stages. In the early stages of grouting, the system focuses on detecting air bubbles or voids to prevent insufficient structural strength due to residual air. In the middle stages, the system monitors the grout surface for cracks, which may be caused by uneven grout solidification. In the later stages, the system checks for unfilled areas to ensure overall fullness. For a specific task, the system detected a crack defect with a treatment confidence level of 0.85 in the middle stages of grouting, located on the right side of the top of the pile. After confirmation by the operator using images, the grouting pressure was adjusted promptly to prevent further crack expansion. This detection mechanism significantly improves the reliability of grouting quality.

[0042] In step S4, the underwater robot analyzes the grout homogeneity and edge detection to cluster the grout segregation rate using color space analysis, and collects parameters such as temperature, pH, and viscosity. This stage is a crucial part of grouting quality monitoring, aiming to comprehensively evaluate the physical properties and distribution of the grout through image analysis and sensor data acquisition, ensuring that the grouting effect meets engineering requirements. Specifically, the underwater robot uses enhanced images acquired by a binocular optical camera to analyze the grout homogeneity using specific color space conversion techniques, while employing edge detection and clustering methods to quantify potential segregation phenomena within the grout. Furthermore, onboard sensors collect relevant grout parameters in real time, providing multi-dimensional data support for subsequent quality analysis.

[0043] In step S41, when analyzing the uniformity of the slurry, the underwater robot first converts the enhanced image from a conventional color space to a color space more suitable for analysis, such as characterizing the color distribution characteristics of the image through three dimensions: hue, saturation, and brightness. This conversion method can effectively highlight the variation characteristics of the slurry's color and brightness, making it easier to identify whether there are color inhomogeneities or layering phenomena. The system divides the converted image into regions and calculates the statistical characteristics of the color distribution in each region, such as the mean and variance of hue. If the variance of a certain region exceeds a preset threshold, it indicates that there may be a uniformity problem in that region, requiring further attention.

[0044] For example, during an offshore pile foundation grouting monitoring mission, after performing color space conversion on enhanced images near the grout overflow outlet, the underwater robot discovered a large variance in the hue distribution near the top of the pile, indicating potential uneven mixing of the grout in this area. The system marked this area as a key area of ​​concern and recorded relevant data, uploading it to the control station. Operators, observing the image, noted that the area's color was lighter, possibly due to excessive water content in the grout. Subsequently, by adjusting the output pressure of the grouting pump, more uniform grout mixing was ensured. This analysis effectively identified potential quality issues, providing a basis for timely adjustments.

[0045] Step S42: Based on the uniformity analysis, the system further identifies segregation phenomena in the slurry using edge detection technology. Edge detection refers to extracting boundary regions with significant changes in brightness or color within the slurry area using image processing methods. These boundaries often correspond to stratification or separation regions of different components in the slurry. The system clusters the detected edges, grouping adjacent edges with similar characteristics into the same group and calculating the area ratio of each group as a quantitative indicator of the segregation rate. If the segregation rate exceeds a preset threshold, such as 5%, it indicates that there may be significant component separation in the slurry, which needs to be recorded and the operator alerted.

[0046] In one embodiment, edge detection can be divided into several steps. First, the system preprocesses the enhanced image to filter out minor noise to improve the accuracy of edge recognition. Then, a gradient calculation method is used to extract regions with significant brightness changes in the image and generate an edge map. Next, a clustering method is used to group the line segments in the edge map, calculate the area enclosed by each group of line segments, and compare it with the total area to obtain the segregation rate. In a specific task, the system detected a segregation rate of 6% in the bottom area of ​​the pile foundation, exceeding the threshold. Operators, combining the image with their observations, confirmed that there was obvious stratification in this area, possibly due to inconsistent grout settlement rates. This discovery helps to take timely measures to prevent further deterioration of the grouting quality.

[0047] In step S43, simultaneously, the underwater robot collects the temperature, pH, and viscosity parameters of the grout using its onboard sensors. These parameters directly reflect the physical and chemical properties of the grout and are crucial for evaluating grouting quality. The temperature sensor measures the real-time temperature of the grout near the overflow outlet to determine if there are any abnormal temperature rises or falls; the pH sensor detects the chemical stability of the grout to ensure that acid-base imbalance does not affect the curing effect; and the viscosity sensor measures the flow resistance of the grout to assess whether its fluidity meets design requirements. The collected data is uploaded to the control station in real time via an umbilical cable, forming a time-series record.

[0048] For example, in a pile foundation grouting task at a water depth of 35 meters, the underwater robot detected a grout temperature of 18.5 degrees Celsius near the overflow outlet using a temperature sensor, slightly lower than expected, possibly due to rapid heat dissipation caused by the deep-water environment. The pH sensor showed a value of 7.2, within the normal range, indicating stable chemical properties of the grout. The viscosity sensor, however, recorded low grout fluidity, potentially affecting its filling effect. Based on this data, operators adjusted the output parameters of the grouting pump, increasing the grout temperature and fluidity to ensure a smooth grouting process. This multi-parameter data acquisition method provided a comprehensive basis for quality assessment.

[0049] In step S5, the underwater robot locates the overflow outlet and calculates the pixel velocity using optical flow. After converting this velocity to the actual velocity using binocular depth information, it performs filtering and sends a pump frequency adjustment command to the pump station via a communication protocol based on a threshold value. This stage is crucial for the dynamic control of the grouting process. It aims to dynamically adjust the output frequency of the grouting pump by monitoring the grout flow velocity in real time, ensuring that the grouting speed and pressure are within an appropriate range. Specifically, the underwater robot uses binocular optical cameras to locate the overflow outlet area, calculates the pixel velocity of the grout flow through image sequence analysis, converts it to the actual velocity using depth information, removes noise interference through filtering, and finally sends an adjustment command to the pump station based on the comparison between the flow velocity value and a preset threshold.

[0050] Step S51: First, the underwater robot uses a dual-lens optical camera to locate the overflow area and performs optical flow analysis on consecutive frame images. Optical flow analysis calculates the pixel velocity of the slurry flow by comparing the displacement of pixels in adjacent frames. During the calculation, the system removes outliers, such as those caused by underwater suspended matter, and selects representative pixel values ​​as the final result. This process effectively reflects the flow state of the slurry near the overflow outlet, providing reliable data for subsequent velocity conversion.

[0051] For example, in a grouting monitoring task, an underwater robot performed optical flow analysis on an image sequence of the overflow area. It found that most pixels had the same displacement direction, indicating relatively stable grout flow. However, a few abnormal displacement points existed at the image edges, possibly due to water flow interference or obstruction by suspended debris. The system automatically removed these abnormal points and selected the pixel displacement values ​​from the central region as representative values ​​to calculate the average pixel velocity. This processing method effectively reduced the impact of interference factors and ensured the accuracy of the flow velocity calculation.

[0052] Step S52: Subsequently, the system combines the binocular depth information to convert the pixel velocity into the actual velocity. The binocular depth information is calculated using the parallax of images acquired by the binocular optical cameras, reflecting the actual spatial distance in the overflow port area. The system calculates the actual velocity based on the camera's focal length and pixel size using the formula... To estimate the actual velocity of the slurry flow, where v pixel_85 Z represents the pixel flow rate at the 85th percentile within the ROI. center The pixel represents the depth value of the real-time depth map of a stereo SGBM at the center point of the ROI. size =2.2μm, f=8mm. This conversion process takes into account the refractive effect of the underwater environment, and the accuracy of the results is ensured by correcting the parameters.

[0053] In one embodiment, the conversion process can be divided into two steps. First, the system extracts the depth value of the central region of the overflow port from the binocular image, for example, measuring a depth of 0.5 meters. Second, by combining the camera's internal parameters and pixel displacement values, the true velocity is calculated. In one task, the system measured a pixel velocity of 10 pixels per second. Combining the depth information and correction parameters, the true velocity is calculated to be 0.2 meters per second. This velocity value directly reflects the flow state of the slurry, providing an important reference for subsequent adjustments.

[0054] In step S53, after obtaining the actual velocity, the system removes noise interference through filtering and determines the flow velocity status based on a preset threshold, then sends an adjustment command to the pump station. The filtering process typically employs a time-series smoothing method to reduce velocity value jumps caused by water flow fluctuations or equipment vibrations. If the filtered flow velocity is below the low-speed threshold, for example, 0.1 meters per second, it indicates that the slurry flow is too slow, and the system will continuously send a pressurization command; if the flow velocity remains low without improvement, a pump stop command is sent to avoid equipment overload; if the flow velocity is in the medium-speed range, for example, between 0.1 and 0.3 meters per second, a warning command is sent to alert the operator; if the flow velocity reaches the high-speed range, for example, exceeding 0.3 meters per second, it is considered a normal state, and only routine monitoring is performed.

[0055] For example, in a pile foundation grouting task at a water depth of 40 meters, the system measured the grout flow velocity to be 0.08 meters per second after filtering, which was below the low-velocity threshold. The system automatically sent a pressurization command to the pump station, increasing the pump's output frequency. After 5 minutes, if the flow velocity still hadn't returned to the normal range, the system sent another command, requesting a temporary pump stop and a check for pipe blockage. Operators confirmed through the control station that a small amount of sediment was present in the pipe. The pipe was then cleared, and the pump station was restarted, ultimately restoring the flow velocity to 0.25 meters per second. This dynamic adjustment mechanism effectively prevented grouting quality problems caused by abnormal flow velocity.

[0056] In step S6, the underwater robot constructs a seabed topography model using a multibeam echo sounder and calculates the depth, volume, area, and slope factor of the scour pit to generate a risk index. This stage is an important supplementary step in grouting quality monitoring, aiming to assess the stability of the seabed around the pile foundation, identify potential risks caused by water scouring, and ensure the long-term safe operation of the pile foundation. Specifically, the underwater robot uses a multibeam echo sounder to collect seabed depth data, constructs a three-dimensional topography model, and then calculates relevant parameters of the scour pit by analyzing the model characteristics, and comprehensively generates a risk index to provide a basis for decision-making in engineering maintenance.

[0057] Step S61: First, the underwater robot transmits acoustic signals via a multibeam echo sounder and receives the reflected echoes from the seabed, acquiring high-precision depth data. The system performs voxel filtering on the acquired data to remove outliers caused by underwater suspended objects or noise. Then, the filtered data is spatially registered with the pile foundation location to construct a three-dimensional topographic model of the seabed. This model visually reflects the undulations and scouring of the seabed around the pile foundation, laying the foundation for subsequent analysis.

[0058] For example, in a monitoring mission for offshore wind turbine foundations, an underwater robot used a multibeam echo sounder to scan the seabed area within a 10-meter radius around the foundations, collecting thousands of depth data points. After voxel filtering the data to remove outliers caused by fine particles carried by the water flow, the system registered the data with the coordinates of the foundation's center point to construct a three-dimensional topographic model of the seabed. The model showed a distinct depression on the southeast side of the foundation, likely formed by long-term water erosion. This model provided a reliable basis for subsequent calculations of erosion pit parameters.

[0059] In step S62, based on the terrain model, the system further calculates relevant parameters of the scour pit, including depth, volume, area, and slope factor. Depth calculation involves subtracting the lowest point height of the scour pit from the surrounding seabed reference height to obtain the scour depth; volume calculation estimates the total volume of the scour pit area by integrating the area; area calculation involves extracting the boundary of the scour pit and calculating its projected area; and the slope factor assesses its stability by analyzing the inclination angle of the scour pit's edges. The system records these parameters as input data for risk assessment.

[0060] In one embodiment, parameter calculations can be optimized for different scour patterns. For shallow but large scour pits, the system focuses on calculating their area and volume to assess their impact on the overall stability of the pile foundation. For deep but small scour pits, the system focuses on analyzing their slope factor to determine if there is a risk of further expansion. In one task, the system measured the depth of the scour pit on the north side of the pile foundation to be 1.2 meters, its volume to be 3.5 cubic meters, and its area to be 4.0 square meters. The high slope factor indicates that the edge of the scour pit is relatively steep and may continue to deteriorate. This multi-parameter analysis method provides a comprehensive perspective for risk assessment.

[0061] Step S63: Subsequently, the system calculates the risk index of the scour pit using a weighted method. Weighted calculation refers to assigning different weights to depth, volume, area, and slope factors based on their relative importance, using a formula... Generate a comprehensive risk index, in which Rush ProbThe values ​​range from 0 to 100%, with w1=0.45, w2=0.30, w3=0.15, and w4=0.10 representing weights optimized based on measurements from 28 wind farms across the country; D max To determine the maximum depth of the scour pit, V pit To scour the pit volume, A 50 Slope is the area enclosed by contour lines in a region with a depth ≥ 0.5m. factor =Local maximum slope / 45°. If the risk index exceeds the preset threshold, such as 0.7, it indicates that the scour pit poses a potential threat to the stability of the pile foundation, and the system will mark the area and prompt the operator to pay attention; if the index is below the threshold, it is considered a normal state, and only the data is recorded for subsequent analysis.

[0062] For example, in a pile foundation monitoring task at a water depth of 25 meters, the system calculated a risk index of scour pit to be 0.75, exceeding the threshold. Analysis showed that the scour pit had a high depth and slope factor, possibly due to recent strengthening of ocean currents. Based on this result, the operators planned to reinforce the area after grouting to prevent further scour. This risk assessment mechanism provides an important guarantee for the long-term safe operation of the pile foundation.

[0063] Step S7: After the underwater robot completes the grouting process, it cruises and performs a second scan to verify the fullness and defects. It then surfaces and is recovered, generating a report. This stage is the final step in grouting quality monitoring, aiming to confirm the grouting effect through a second inspection, record final data, and provide a complete report for project acceptance. Specifically, the underwater robot cruises along a predetermined path, using a multibeam echo sounder and optical camera to comprehensively scan the grouting area, verifying fullness and defects. It then safely surfaces and is recovered, finally generating a detailed report through post-processing.

[0064] Step S71: First, the underwater robot cruises along a pre-planned curved path, performing a secondary scan of the pile foundation grouting area. During the scan, a multibeam echo sounder is used to detect the overall morphology of the grouting area and determine whether there are leaks or unfilled areas; a binocular optical camera is used to capture surface details and check for defects such as cracks or bubbles. The system analyzes the scan data in real time. If the fullness is found to be below a preset threshold or the number of defects exceeds the allowable range, the relevant locations are marked and detailed data is recorded.

[0065] For example, after a grouting task, an underwater robot cruised along a spiral path around the pile foundation, scanning and covering the entire area from top to bottom. The multibeam echo sounder detected a fullness of 0.92 in the top area, meeting the requirements, but a small unfilled area was found in the bottom area. A binocular optical camera further confirmed that there were no obvious cracks in this area, only localized insufficiency. The system recorded this information for subsequent processing by the operators. This secondary scanning process effectively ensured comprehensive verification of the grouting quality.

[0066] In step S72, after completing the cruise scan, the underwater robot begins its ascent for recovery. During the ascent, the system must maintain a safe distance from the mother ship to avoid collisions caused by currents or waves. Simultaneously, the robot uses its onboard water spray system to rinse external equipment, removing attached sea mud or salt to prevent long-term corrosion. Furthermore, the system performs status checks on critical components, such as the thruster's operational status and sensor calibration, to ensure the equipment remains usable after recovery.

[0067] In one embodiment, the ascent process can be divided into several stages. In the initial stage, the robot ascends at a low speed, maintaining depth fluctuations within a safe range while using its sonar system to detect any obstacles above. As it approaches the surface, the speed is appropriately increased, but a lateral distance of at least 5 meters must be maintained from the mother ship to avoid entanglement with the umbilical cable. After recovery, operators conduct a thorough inspection of the robot's outer shell, discovering a small amount of salt deposits, which are subsequently removed by rinsing with fresh water. This meticulous recovery process effectively extends the equipment's lifespan.

[0068] In step S73, the system finally performs post-processing on all collected data to generate a detailed report. Post-processing includes integrating and analyzing image data, terrain models, and sensor parameters to generate a heat map of the grouting area, visually displaying the fullness and defect distribution. Simultaneously, the system optimizes and records the cruise path, providing a reference for subsequent tasks. Furthermore, all data is backed up and stored to ensure information integrity. The final report includes grouting quality assessment, scour risk index, and equipment status, providing a basis for project acceptance and maintenance.

[0069] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring of concrete pouring for jacket foundations based on underwater robots, characterized in that, include: The underwater robot is deployed to the offshore pile foundation area via the mother ship's crane and umbilical cable. After self-checking the binocular optical camera, sonar system, thruster, auxiliary sensors and infrared auxiliary imaging module, it is imported into the CAD pile foundation model for path pre-planning to obtain the pre-planned grid obstacle avoidance path. After the underwater robot integrates GPS and INS positioning and locks onto the pile foundation through forward-looking sonar, it dives to the target water depth at a gradient speed according to the pre-planned grid obstacle avoidance path, completes attitude adjustment, and hovers close to the overflow outlet. The underwater robot generates enhanced images using a binocular optical camera combined with LED supplemental lighting, and quantifies fullness by calculating the ratio of gap volume to CAD theoretical volume based on binocular parallax, while simultaneously using a model to detect defects. The underwater robot uses HSV spatial analysis to determine the homogeneity of the slurry, detects edges to cluster the slurry segregation rate, and collects parameters such as temperature, pH, and viscosity. The underwater robot locks onto the overflow outlet and calculates the pixel flow velocity using optical flow. After combining the binocular depth measurement with the actual velocity, it filters the data and sends a pump frequency adjustment command to the pumping station based on the threshold through a communication protocol. The underwater robot constructs a seabed topography model through a multibeam echo sounder and calculates the depth, volume, area and slope factor of scour pits to generate a risk index. The risk index is transmitted to the ground control station in real time, and corresponding response measures are triggered according to the judgment level of the risk index. After the underwater robot finishes casting, it cruises and performs a second scan to verify the fullness and defects before surfacing to retrieve the vessel and generate a report.

2. The method as claimed in claim 1, characterized in that, The underwater robot is deployed to the offshore pile foundation area via a mother ship crane and umbilical cable. After self-checking its binocular optical camera, sonar system, thrusters, auxiliary sensors, and infrared auxiliary imaging module, it is imported into the CAD pile foundation model for path pre-planning, resulting in a pre-planned grid obstacle avoidance path, including: After confirming that the sea state parameters meet the standards through the meteorological station and supporting testing equipment, the underwater robot is hoisted into the water by a crane and connected to the umbilical cable. The dual-optical camera performs distortion calibration, the forward-looking sonar performs circumferential scanning verification, the bottom sonar performs accuracy verification, the thruster performs parameter adjustment, the auxiliary sensor performs zero-point verification, and the infrared auxiliary imaging module performs functional self-test. The control station imports the CAD model and uses a unified coordinate system to plan the grid obstacle avoidance path, thus obtaining the pre-planned grid obstacle avoidance path.

3. The method as described in claim 1, characterized in that, The underwater robot integrates GPS and INS positioning and locks onto the pile foundation using forward-looking sonar. Following a pre-planned grid obstacle avoidance path, it descends at a gradient speed to the target water depth, completes attitude adjustment, and hovers near the overflow outlet, including: After initial fusion of GPS and INS, positioning accuracy is calibrated based on a unified coordinate system, and GPS is turned off if the depth exceeds a threshold. Forward-looking sonar locates the pile foundation through feature extraction, compares real-time positioning data with the pre-planned path, and dynamically adjusts the diving direction; After adjusting attitude according to the pre-planned path, the vehicle descends in a gradient manner, maintaining depth fluctuations. The deflection angle is monitored in real time using IMU data. When the deflection angle exceeds the set threshold or deviates from the pre-planned path, a compensation mechanism is triggered to adjust the propulsion torque. The underwater robot transmits video data to the ground control station and switches to infrared-assisted imaging when the sonar signal weakens. Approaching the pre-planned hovering point, PID control is used to achieve hovering stability, and video and log data are collected and uploaded.

4. The method as described in claim 1, characterized in that, The method of generating enhanced images using a dual-lens optical camera combined with LED fill light includes: After the binocular optical camera acquires images, it estimates the underwater global background light and underwater transmittance based on the preset underwater dark channel defogging model, and then reconstructs and generates a fog-free image. After multi-scale fusion weighting and color restoration factor processing, edge preservation is achieved through guided filtering. Adaptive stretching is applied to the image after guided filtering to obtain the final enhanced frame.

5. The method as described in claim 1, characterized in that, The method of quantifying fullness based on the ratio of gap volume calculated by binocular parallax to CAD theoretical volume, while simultaneously using a model to detect defects, includes: Depth data of the annular gap of the catheter holder is obtained by binocular parallax, and the actual total volume V of the annular gap of the catheter holder is obtained by integrating the depth data. actual The actual total volume is the maximum volume that the annular gap can fill; Define the theoretical volume in CAD as the total design volume V of the annular gap of the jacket structure. design The actual filling volume V of the slurry was calculated by integrating the depth data of the slurry-filled area using binocular parallax analysis. filled ; The fullness is quantified by calculating the ratio of unfilled gap volume to fullness, and the fullness directly reflects the degree of filling of the annular gap; By detecting the continuous discharge status of the slurry and the absence of air bubbles and voids, and combining the above fullness calculation results, it is determined whether the fullness error is within the allowable range. An alarm is triggered when the model detects a defect with a confidence level exceeding a threshold. The collected temperature, pH, and viscosity parameters are compared with preset qualified thresholds. Combined with the uniformity test results and segregation rate test results, the quality of the slurry is comprehensively determined to meet the standards. If any parameter exceeds the preset threshold or the comprehensive test results do not meet the requirements, a slurry quality alarm is triggered.

6. The method as described in claim 1, characterized in that, The process of locking the overflow outlet, calculating pixel flow rate using optical flow, converting actual velocity using binocular depth, filtering, and then sending pump frequency adjustment commands to the pump station via a communication protocol based on a threshold includes: The model locks the overflow port area, and optical flow calculations eliminate outliers by selecting quantile pixel values. The speed of converting the center depth of the binocular depth map to the actual speed by combining pixel size and focal length; The flow rate value is filtered. If the pressure boosting command is continuously sent at low speed and the pump stops if the problem is not resolved, a warning is issued at medium speed and normal monitoring is performed at high speed.

7. The method as described in claim 1, characterized in that, The process of constructing a seabed topography model using a multibeam echo sounder and calculating the depth, volume, area, and slope factor of scour pits to generate a risk index includes: Multibeam echo sounding voxel filtering model registration; Based on the theoretical seabed height corresponding to the outer wall of the pile, the actual height of the terrain model is subtracted to calculate the depth of the scour pit. The scour pit volume is obtained by summing the scour pit depths. At the same time, the maximum depth of the scour pit, the area enclosed by contour lines in the region where the depth is greater than the preset depth threshold, and the local slope factor are determined. The risk index is calculated and generated. Based on the preset threshold, the risk index is divided into three judgment levels: safe, concern, and danger. The judgment level and risk index are transmitted to the ground control station simultaneously. When the danger level is reached, an emergency response command is triggered, prompting the implementation of measures such as dumping boulders or backfilling with mud.

8. The method as described in claim 1, characterized in that, After the pouring is completed, a second cruise scan is performed to verify the fullness and defects. The vessel then floats back to shore and generates a report, including: Cruise along a pre-planned cruise path curve, and check for fullness and defects by using multi-beam and optical scanning to detect leakage thresholds; Maintain a safe distance when surfacing, move to the designated recovery area according to the pre-planned recovery route, and rinse the underwater robot and check the degree of corrosion after recovery; In the post-processing stage, the terrain model is registered, a slurry flow heat map is generated, the operation path is optimized, a model analysis report is generated including slurry quality detection, fullness detection, scour risk assessment, and path execution review, and data backup is completed.

9. A real-time monitoring system for concrete pouring of jacket foundation based on an underwater robot, characterized in that, The system includes: The equipment deployment and path pre-planning module is used to deploy the underwater robot to the offshore pile foundation area via the mother ship crane and umbilical cable, complete the self-check of the equipment including the binocular optical camera and sonar system, import the CAD pile foundation model and perform path pre-planning to obtain the pre-planned grid obstacle avoidance path. The positioning, diving, and hovering module is used to locate the pile foundation through GPS and INS fusion positioning and forward-looking sonar, and control the underwater robot to dive to the target water depth at a gradient speed according to the pre-planned grid obstacle avoidance path. After completing attitude adjustment, it approaches the overflow outlet and achieves stable hovering. The image enhancement, fullness quantification and defect detection module is used to generate enhanced images by combining binocular optical cameras with LED supplementary lighting, calculate the volume of the annular gap of the guide frame based on binocular parallax, quantify the fullness by the ratio of the volume to the CAD theoretical volume, and detect defects using a model. The slurry quality parameter monitoring module is used to analyze the homogeneity of the slurry through HSV spatial analysis, obtain the slurry segregation rate by combining edge detection clustering algorithm, and simultaneously collect the temperature, pH and viscosity parameters of the slurry. The overflow velocity monitoring and pump station control module is used to lock the overflow port area and calculate the pixel flow velocity through optical flow. Combined with binocular depth conversion, the actual overflow velocity is obtained. After filtering, the pump frequency adjustment command is sent to the pump station through the communication protocol according to the preset threshold. The seabed topography modeling and scour risk assessment module is used to construct a seabed topography model through a multibeam echo sounder system, calculate the depth, volume, area and slope factor of scour pits and generate a risk index, transmit the risk index to the ground control station in real time, and trigger corresponding response measures according to the judgment level corresponding to the risk index. The cruise verification, recovery, and report generation module is used to control the underwater robot to cruise and perform a second scan to verify fullness and defects after the pouring is completed, complete the surface recovery operation, and generate an analysis report containing key monitoring data.

Images