A multi-beam sonar dynamic three-dimensional visualization system for underwater construction

Abstract

This invention relates to the field of underwater engineering detection technology and discloses a multi-beam sonar dynamic three-dimensional visualization system for underwater construction, including sonar equipment, a main controller, an inertial navigation system, a data processing host, a graphics display device, and an underwater current meter. The invention utilizes a timing delay module to calculate delay parameters and generate a delay rate, and a mesh construction module to generate dynamic mesh edge length values ​​to create a hash mesh, eliminating time misalignment errors caused by the diffusion of suspended sediment clumps. An envelope calculation module combines occupancy integral values ​​and echo intensity variance values ​​to generate a steady-state factor value. An error correction module calculates the feature anchor point spatial normal residual and inversely compensates the theoretical envelope boundary to generate a corrected envelope, correcting spatial position errors caused by stress deformation. A multi-level engine module generates classified point clouds and calculates memory weight values ​​to allocate independent memory resources, preventing global memory allocation congestion in the graphics processor.

Description

Technical Field

[0001] This invention relates to the field of underwater engineering detection technology, specifically a multibeam sonar dynamic three-dimensional visualization system for underwater construction. Background Technology

[0002] Multibeam sonar detection systems suffer from several technical drawbacks when performing 3D visualization of engineering machinery movements in underwater construction environments. The excavation actions of the machinery disrupt the underwater geological layer, generating suspended sediment clumps. These clumps diffuse with the water flow into the sonar detection area, resulting in spatial motion lag. This causes a time misalignment error between the received signals of the engineering machinery's movements and the acoustic detection signals. Furthermore, the suspended sediment clumps generate numerous interfering point clouds, increasing the system's data processing and computational load.

[0003] When construction machinery performs underwater operations, the reaction force from excavation causes deformation of connecting rods and displacement of joint clearances, resulting in spatial position errors between the theoretical envelope boundary and the actual spatial position of the construction machinery. As a result, the system has difficulty extracting the actual surface feature data of the machinery around the boundary where the error exists.

[0004] The underwater construction environment contains mechanical bodies, static underwater geological layers, and dynamic sediment spatial targets. The system has difficulty performing spatial target decoupling and separation actions based on the sonar point cloud attribute characteristics. Massive point cloud data floods into the graphics rendering pipeline concurrently. The system does not allocate independent video memory resources for the bandwidth throughput requirements of point clouds with different attributes, causing congestion in the global video memory resource allocation within the graphics processor.

[0005] Therefore, this invention proposes a multibeam sonar dynamic three-dimensional visualization system for underwater construction to address the shortcomings of existing technologies. Summary of the Invention

[0006] The technical problem solved by this invention is that multibeam sonar detection of underwater construction machinery suffers from data time errors caused by differences in sensor sampling frequencies and image misalignment caused by inconsistent spatial coordinate references. Excavation operations generate suspended silt masses that interfere with acoustic detection. Mechanical deformation under stress causes spatial deviations in the theoretical envelope boundary. Concurrent rendering of massive point cloud data causes congestion in the allocation of graphics processor memory resources.

[0007] To address the existing technical problems, the present invention provides the following technical solution: This invention provides a multi-beam sonar dynamic three-dimensional visualization system for underwater construction, including sonar equipment, main controller, inertial navigation equipment, data processing host, graphics display equipment, and underwater current meter.

[0008] The sonar equipment is installed in the underwater environment and outputs raw point clouds. The main controller is connected to the engineering machinery and outputs joint pose and pressure variation. The inertial navigation equipment includes a first inertial navigation unit and a second inertial navigation unit. The first inertial navigation unit is rigidly connected to the sonar equipment base and outputs the sonar's 6-degree attitude. The second inertial navigation unit is rigidly connected to the engineering machinery base and outputs the machinery's 6-degree attitude. The sonar equipment, main controller, inertial navigation equipment, and underwater current meter are all electrically connected to the data processing host input port through a communication network. The graphics display equipment is connected to the data processing host output port and receives rendering commands to execute the final image output action.

[0009] The data processing host integrates a synchronization alignment module, a timing delay module, a mesh construction module, an envelope calculation module, an error correction module, and a multi-level engine module. The synchronization alignment module receives the original point cloud, joint pose, pressure variability, sonar 6-degree attitude, and mechanical 6-degree attitude, and performs timestamp alignment and coordinate transformation to generate a unified coordinate point cloud. The timing delay module receives the flow velocity and straight-line distance to calculate the delay parameters and combines them with the pressure variability to generate the delay variability. The mesh construction module receives the delay variability and performs a scale expansion to generate a hash mesh. The envelope calculation module calculates the theoretical envelope boundary within the hash mesh by combining the joint pose and the unified coordinate point cloud to generate a steady-state factor value. The error correction module selects feature anchor points based on the steady-state factor value and the theoretical envelope boundary and inversely calculates the spatial normal residual to generate a corrected envelope. The multi-level engine module receives the corrected envelope and the steady-state factor value, performs a classification operation on the point cloud to be rendered, and distributes the classified point cloud to the asynchronous rendering queue to generate output frames.

[0010] The synchronization alignment module parses the timestamps written to the data frames at the hardware control layer of the receiving port, compares the hardware sampling frequencies of the sonar device and the main controller, and calculates the instantaneous joint pose, instantaneous pressure rate, instantaneous sonar 6-degree attitude, and instantaneous mechanical 6-degree attitude using cubic spline interpolation and spherical linear interpolation algorithms to generate a synchronization data stream. The synchronization alignment module uses the instantaneous sonar 6-degree attitude and instantaneous mechanical 6-degree attitude to construct a homogeneous transformation matrix corresponding to the multi-source local reference coordinate system. It then multiplies the original point cloud carrying local 3D coordinates and the joint pose by the homogeneous transformation matrix to generate a unified coordinate point cloud in absolute geodetic coordinate space.

[0011] The time-series delay module extracts the linear distance corresponding to the spatial locking point of the suspended sediment cluster and performs a division operation with the flow velocity output by the underwater current meter to generate a delay parameter. This module applies the delay parameter to the pressure variability time axis, shifting the pressure variability backward by the time length corresponding to the delay parameter to generate a delayed variability. The mesh construction module performs a dot product of the delayed variability and the mapping coefficient value to obtain the dynamic compensation increment, and linearly superimposes the dynamic compensation increment with the static mesh edge length value to generate the dynamic mesh edge length value. This dynamic mesh is then generated using a spatial hash mapping function.

[0012] The envelope solving module extracts the 3D geometric model of the engineering machinery, link length values, and joint poses, performs mathematical operations to generate local transformation matrices, performs matrix multiplication to generate positive kinematic matrices, and uses triangular patches to connect surface vertices to generate a 3D closed surface shape, which is then established as the theoretical envelope boundary. Within the hash grid, the module establishes a time window to extract a unified coordinate point cloud, statistically analyzes the frequency of occurrence of independent voxel units to generate occupancy integral values, extracts echo intensity values, performs dispersion measurement to generate echo intensity variance values, and uses the occupancy integral values ​​and echo intensity variance values ​​to generate a steady-state factor value.

[0013] The error correction module performs a spatial expansion operation along the outer normal direction of the theoretical envelope boundary according to a set distance parameter to generate an external spatial neighborhood. The error correction module utilizes the instantaneous mechanical motion rate to perform negative adaptive compensation on the hard object threshold. For unified coordinate point clouds with steady-state factor values ​​greater than the hard object threshold, the error correction module performs an independent extraction operation to generate feature anchor points. It calculates the coordinates of the feature anchor points and the projected points, performs vector subtraction to generate a deviation vector, and combines the deviation vector with the surface normal vector to perform a vector dot product operation to generate the spatial normal residual.

[0014] The error correction module uses the spatial normal residual combined with the distance weight parameter to perform a multiplication operation to generate a scalar compensation distance. It then performs a vector scalar multiplication operation with the surface normal vector to generate a compensation vector. Finally, it performs a spatial vector addition operation with the theoretical envelope boundary surface vertex coordinates to generate corrected vertex coordinates. The corrected vertex coordinates are then used to reconstruct the three-dimensional surface shape to generate a corrected envelope.

[0015] The multi-level engine module extracts a unified coordinate point cloud carrying steady-state factor values ​​and defines it as the point cloud to be rendered. It then calculates the spatial relationship between the 3D coordinates of the point cloud to be rendered and the boundary of the modified envelope surface. When the multi-level engine module determines that the 3D coordinates of the point cloud to be rendered are outside the modified envelope space, it performs an interval comparison operation with the steady-state factor values ​​and the geological and sediment thresholds, and merges the mechanical classification point cloud, the geological classification point cloud, and the sediment classification point cloud to generate a classified point cloud.

[0016] The multi-level engine module establishes an asynchronous rendering queue within the graphics processor. It performs weighted independent calculations using the total amount of local point cloud data and the baseline refresh rate parameter to generate video memory weight values. It then uses these video memory weight values ​​to allocate isolated independent video memory resources across the global video memory capacity. Finally, it issues refresh commands with time-difference characteristics to the asynchronous rendering queue and extracts spatial pixel depth values. Combined with depth buffer test rules, it performs layer compositing to generate output frames.

[0017] This invention provides a multi-beam sonar dynamic three-dimensional visualization system for underwater construction, which has the following advantages: 1. This invention extracts the straight-line distance and water flow velocity through the time delay module to calculate the delay parameters. The delay parameters are applied to the pressure variability time axis to generate the delay variability. The grid construction module combines the delay variability to generate dynamic grid side length values ​​to generate a hash grid, thereby eliminating the time misalignment error caused by the diffusion of suspended sediment masses due to the excavation action. The dynamic grid side length values ​​are used to reduce the data processing and computational load of the suspended sediment mass area.

[0018] 2. This invention generates a steady-state factor value by combining the occupancy integral value and the echo intensity variance value through the envelope calculation module. The error correction module searches for feature anchor points in the spatial neighborhood outside the theoretical envelope boundary whose steady-state factor value is greater than the hard object threshold. The spatial normal residual generated by calculating the feature anchor points is used to inversely compensate the theoretical envelope boundary to generate a corrected envelope, thereby repairing the spatial position error caused by the stress deformation phenomenon and joint clearance offset phenomenon of engineering machinery.

[0019] 3. This invention uses a multi-level engine module to perform interval comparison between steady-state factor values ​​and geological and sediment thresholds to generate classified point clouds. It calculates the video memory weight value by combining the total amount of local point cloud data with the benchmark refresh frequency parameter, and allocates independent video memory resources by cutting the global video memory capacity inside the graphics processor. This completes the decoupling and separation of three-dimensional spatial targets and matches the bandwidth throughput requirements of specific classified point clouds in the graphics rendering process, preventing the concurrency of global video memory resource allocation caused by the concurrent transmission of massive point clouds. Attached Figure Description

[0020] Figure 1 This is a structural block diagram of a multibeam sonar dynamic three-dimensional visualization system for underwater construction according to the present invention.

[0021] Figure 2 This is a flowchart illustrating the overall steps of a multibeam sonar dynamic three-dimensional visualization method for underwater construction according to the present invention.

[0022] Figure 3 This is a detailed flowchart of the sub-steps of step S100 in this invention.

[0023] Figure 4 This is a detailed flowchart of the sub-steps of step S200 in this invention.

[0024] Figure 5 This is a detailed flowchart of the sub-steps of step S300 in this invention.

[0025] Figure 6 This is a detailed flowchart of the sub-steps of step S400 in this invention.

[0026] Figure 7 This is a detailed flowchart of the sub-steps of step S500 in this invention.

[0027] Figure 8This is a dynamic comparison chart of the screen refresh rate of the present invention. Detailed Implementation

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] See Figure 1 This invention provides a multi-beam sonar dynamic three-dimensional visualization system for underwater construction, including sonar equipment, main controller, inertial navigation equipment, data processing host, graphics display equipment and underwater current meter.

[0030] The sonar equipment is installed in the underwater environment and outputs raw point clouds. The main controller is connected to the engineering machinery and outputs joint pose and pressure variation. The inertial navigation equipment includes a first inertial navigation unit and a second inertial navigation unit. The first inertial navigation unit is rigidly connected to the sonar equipment base and outputs the sonar's 6-degree attitude. The second inertial navigation unit is rigidly connected to the engineering machinery base and outputs the machinery's 6-degree attitude. The sonar equipment, main controller, inertial navigation equipment, and underwater current meter are all electrically connected to the data processing host input port through a communication network. The graphics display equipment is connected to the data processing host output port and receives rendering commands to execute the final image output action.

[0031] The data processing host integrates a synchronization alignment module, a timing delay module, a mesh construction module, an envelope calculation module, an error correction module, and a multi-level engine module. The synchronization alignment module receives the original point cloud, joint pose, pressure variability, sonar 6-degree attitude, and mechanical 6-degree attitude, and performs timestamp alignment and coordinate transformation to generate a unified coordinate point cloud. The timing delay module receives the flow velocity and straight-line distance to calculate the delay parameters and combines them with the pressure variability to generate the delay variability. The mesh construction module receives the delay variability and performs a scale expansion to generate a hash mesh. The envelope calculation module calculates the theoretical envelope boundary within the hash mesh by combining the joint pose and the unified coordinate point cloud to generate a steady-state factor value. The error correction module selects feature anchor points based on the steady-state factor value and the theoretical envelope boundary and inversely calculates the spatial normal residual to generate a corrected envelope. The multi-level engine module receives the corrected envelope and the steady-state factor value, performs a classification operation on the point cloud to be rendered, and distributes the classified point cloud to the asynchronous rendering queue to generate output frames.

[0032] The system execution process follows a fixed sequence. The system's underlying communication bus collects and synchronizes external sensor signals and engineering machinery status signals. The timing delay module and mesh construction module reconstruct the hash mesh side length values ​​based on the flow rate and straight-line distance. The envelope calculation module calculates the theoretical envelope boundary based on the joint pose and extracts the statistical features of the point cloud inside the spatial data structure unit to generate steady-state factor values. The error correction module uses feature anchor points to define the hard object point cloud cluster and calculates the reverse compensation vector to repair the deformation error of the theoretical envelope boundary and generate a corrected envelope. The multi-level engine module determines the point cloud attributes based on the corrected envelope spatial range and steady-state factor values. The graphics processor allocates video memory resources for different point cloud attributes and executes differentiated refresh instructions to generate the final 3D frame.

[0033] See Figure 2 This invention provides a method for dynamic 3D visualization of multibeam sonar for underwater construction, based on a multibeam sonar dynamic 3D visualization system for underwater construction. The method includes the following steps: S100, the synchronization alignment module receives the original point cloud output by the sonar device, the joint pose and pressure rate output by the main controller, and the sonar 6-degree attitude and mechanical 6-degree attitude output by the inertial navigation device. The synchronization alignment module aligns the timestamps of the original point cloud, joint pose, pressure rate, sonar 6-degree attitude and mechanical 6-degree attitude according to the clock protocol to generate a synchronization data stream. The synchronization alignment module calculates the corresponding transformation matrix using the sonar 6-degree attitude and mechanical 6-degree attitude respectively, and performs coordinate transformation by multiplying the original point cloud and joint pose by the transformation matrix to generate a unified coordinate point cloud. S200, the timing delay module extracts the straight-line distance and flow velocity to calculate the delay parameters. The timing delay module combines the delay parameters with the pressure variability to perform a delay superposition action to generate the delay variability. The grid construction module receives the delay variability and performs a scale expansion action to adjust the side length values ​​of the spatial data structure unit to generate a hash grid. S300, the envelope solution module extracts joint poses and performs forward kinematics calculation to generate theoretical envelope boundaries. The envelope solution module extracts unified coordinate point clouds within the time window for voxel units inside the hash grid. The envelope solution module calculates the unified coordinate point cloud occupancy integral value and echo intensity variance value and performs steady-state evaluation to generate steady-state factor values. S400, the error correction module finds unified coordinate point cloud with steady-state factor values ​​greater than the hard object threshold in the spatial neighborhood outside the theoretical envelope boundary to extract feature anchor points. The error correction module calculates the spatial normal residual between the feature anchor points and the surface coordinates of the theoretical envelope boundary. The error correction module uses the spatial normal residual to inversely compensate the theoretical envelope boundary to repair deformation errors and generate a corrected envelope. The S500 multi-level engine module extracts and corrects the envelope space coverage and steady-state factor values, performs spatial classification on the point cloud to be rendered, and generates classified point clouds. The multi-level engine module pushes the classified point clouds into the asynchronous rendering queue according to the attributes of the classified point clouds. The multi-level engine module allocates video memory resources, executes refresh instructions to generate output frames, and sends the output frames to the graphics display device.

[0034] The technical solution will be explained in detail below.

[0035] See Figure 3 In one embodiment, step S100 may further include the following sub-steps: S101, the internal synchronization and alignment module of the data processing host establishes a data link connecting the sonar equipment, the main controller, and the inertial navigation equipment via a communication network. The sonar equipment uses a multibeam echo sounder, which scans the seabed surface and sends a raw point cloud to the synchronization and alignment module. This raw point cloud contains three-dimensional spatial coordinates and echo intensity values. The main controller uses a programmable logic controller (PLC), which reads the angle values ​​from the absolute encoder of the engineering machinery joints and the pressure transmitter values ​​from the hydraulic circuit, and sends the joint attitude and pressure rate to the synchronization and alignment module. Both the first and second inertial navigation units use integrated navigation units, respectively measuring the three-dimensional spatial translation coordinates and roll, pitch, and yaw angles, and sending the corresponding sonar 6-degree attitude and mechanical 6-degree attitude to the synchronization and alignment module. The communication network is a hybrid of gigabit Ethernet and fieldbus, supporting concurrent transmission of heterogeneous data.

[0036] S102, the synchronization alignment module writes timestamps to the arriving data frames at the hardware control layer of the receiving port. For the underlying hardware timestamp writing and parsing process, those skilled in the art can use the internal media access control layer logic of a programmable gate array to complete the parsing action. The underlying time extraction and parsing mechanism is a well-known technology in the field and will not be elaborated here. The synchronization alignment module parses the data frame timestamps to generate original point clouds with timestamps, joint poses with timestamps, pressure rates with timestamps, sonar 6-degree attitudes with timestamps, and mechanical 6-degree attitudes with timestamps.

[0037] S103, the synchronization alignment module performs timestamp alignment to generate a synchronization data stream. The sonar device, main controller, and inertial navigation device have different hardware sampling frequencies. The synchronization alignment module selects the period of the acoustic signal transmitted by the sonar device as the reference time axis. The synchronization alignment module extracts the original point cloud time tag from the reference time axis and searches for the joint pose, pressure rate of change, and 6-degree attitude with the closest time distance at the original point cloud time tag position. The synchronization alignment module compares the hardware sampling frequencies of the sonar device and the main controller. If the sonar device's sampling frequency is higher than that of the main controller, it uses a cubic spline interpolation algorithm to calculate the instantaneous joint pose and instantaneous pressure rate at the corresponding time of the original point cloud time label. If the sonar device's sampling frequency is lower than that of the main controller, it performs nearest-neighbor time-shaping in the joint sequence to extract the instantaneous joint pose and instantaneous pressure rate. For the extraction of instantaneous sonar 6-degree pose and instantaneous mechanical 6-degree pose, the synchronization alignment module converts the sonar 6-degree pose and mechanical 6-degree pose into quaternion representations and uses the spherical linear interpolation algorithm (Slerp) to calculate the instantaneous quaternion at the corresponding time of the original point cloud time label, and then reversely converts it into instantaneous sonar 6-degree pose and instantaneous mechanical 6-degree pose, eliminating the gimbal lock error caused by spatial angle interpolation. The synchronization alignment module packages and encapsulates the original point cloud with the instantaneous joint pose, instantaneous pressure rate, and instantaneous 6-degree pose, generates a synchronization data stream, and stores it in the memory buffer area of ​​the data processing host.

[0038] S104, the synchronization alignment module extracts the instantaneous sonar 6-degree attitude and instantaneous mechanical 6-degree attitude from the synchronization data stream within the memory buffer area. The underwater construction scenario contains multiple local reference coordinate systems; the sonar equipment has a detector coordinate system, and the engineering machinery has a mechanical base coordinate system. The equipment mounting carrier undergoes spatial displacement due to water flow disturbance, causing spatial deflection and translation of the detector coordinate system and the mechanical base coordinate system relative to the stationary geodetic coordinate system. The synchronization alignment module extracts the corresponding roll angle, pitch angle, and yaw angle values ​​using the instantaneous sonar 6-degree attitude and the instantaneous mechanical 6-degree attitude, respectively. The module then constructs a three-dimensional rotation matrix using these values ​​and combines them with the three-dimensional translation vector of the carrier relative to the geodetic reference origin to construct a homogeneous transformation matrix. The trigonometric function mapping relationship between the roll angle, pitch angle, and yaw angle values ​​and the three-dimensional rotation matrix can be derived using Euler angle transformation formulas by those skilled in the art. The process of constructing the spatial homogeneous transformation matrix is ​​well-known in the field and will not be elaborated further here.

[0039] S105, the synchronization alignment module performs coordinate transformation using a homogeneous transformation matrix. The module extracts local 3D coordinates from the original point cloud within the synchronization data stream and multiplies these local 3D coordinates by the homogeneous transformation matrix corresponding to the sonar device to generate absolute geodetic coordinates. The module also extracts local 3D coordinates from the joint poses within the synchronization data stream and multiplies these local 3D coordinates by the homogeneous transformation matrix corresponding to the engineering machinery to generate absolute geodetic coordinates. Finally, the module maps these absolute geodetic coordinates into a unified coordinate point cloud within the same spatial dimension boundary. This unified coordinate point cloud eliminates the spatial misalignment error caused by the independent movement of the sonar device mounting platform and the engineering machinery mounting platform, providing a static reference coordinate system for subsequent mesh construction steps.

[0040] See Figure 4 In one embodiment, step S200 may further include the following sub-steps: S201, the timing delay module extracts the three-dimensional coordinates of the sonar device in absolute geodetic coordinate space and the absolute geodetic coordinates of the engineering machinery base. The engineering machinery performs underwater excavation, damaging the underwater geological layer. The underwater sediment is disturbed by the excavation power and detaches from the geological layer, producing suspended sediment clumps. The timing delay module extracts the absolute geodetic coordinates of the engineering machinery base corresponding to the peak of the pressure variability in the synchronous data stream as the spatial locking point. The suspended sediment clumps diffuse with the water flow to the detection area of ​​the sonar device, resulting in a spatial motion lag time. The timing delay module calculates the three-dimensional Euclidean distance between the three-dimensional coordinates of the sonar device and the spatial locking point to generate a straight-line distance. The timing delay module receives the water flow rate obtained by the underwater current meter to generate the water flow velocity.

[0041] S202, the timing delay module extracts the straight-line distance and the water flow velocity to calculate the delay parameters. The timing delay module divides the straight-line distance by the water flow velocity to generate the delay parameters. The delay parameters match the time span of the suspended sediment clumps crossing the straight-line distance to reach the detection area of ​​the sonar equipment.

[0042] S203, the timing delay module extracts the internal pressure variation rate of the synchronous data stream and performs a delay superposition action. The hydraulic system of the construction machinery outputs high pressure to drive the hydraulic cylinder to perform digging action. The main controller records the pressure transient signal of the hydraulic circuit to generate pressure variation rate. The increase in the pressure variation rate value indicates the increase in the digging action force and the expansion of the scale of the suspended sediment mass. The timing delay module applies the delay parameter to the pressure variation rate time axis. The timing delay module shifts the pressure variation rate backward by the time length corresponding to the delay parameter in the time dimension to generate a delay variation rate. The delay variation rate eliminates the time misalignment error between the time when the mechanical action occurs and the time when the suspended sediment mass enters the sonar detection area. The delay variation rate is kept synchronized with the actual time when the suspended sediment mass actually interferes with the acoustic detection action.

[0043] S204, the mesh building module receives the delay variability output by the timing delay module. The unified coordinate point cloud is scattered in 3D space. The mesh building module performs voxel segmentation on the 3D space to be processed to generate spatial data structure units. These spatial data structure units are cube-shaped, and the cube's side length determines the granularity of the spatial division. An increase in the delay variability value indicates an increase in the disturbance intensity of suspended sediment masses. The mesh building module extracts the delay variability and performs a scale expansion operation to adjust the cube's side length.

[0044] S205, the mesh construction module extracts the stored static mesh edge length values ​​and mapping coefficient values. The module calculates the dynamic compensation increment by multiplying the delay rate and mapping coefficient values, and then linearly superimposes the dynamic compensation increment with the static mesh edge length values ​​to generate the dynamic mesh edge length values. The dynamic mesh edge length values ​​change synchronously with the delay rate.

[0045] ; In the formula, Represents the side length value of the dynamic grid; Represents the side length of the static grid; Represents the numerical value of the mapping coefficient; This represents the rate of delay.

[0046] S206, the mesh construction module assigns dynamic mesh edge length values ​​to spatial data structure units to generate voxel units. The mesh construction module uses the dynamic mesh edge length values ​​to perform integer division operations on the 3D coordinates of the unified coordinate point cloud to generate a 3D discrete index. The mesh construction module combines this with a spatial hash mapping function to convert the 3D discrete index into hash keys, and establishes a correspondence between hash keys and voxel unit memory addresses to generate a hash mesh. For the underlying code writing process of constructing the mapping relationship between the 3D discrete index and memory addresses using the spatial hash mapping function, those skilled in the art can use a key-value pair mapping algorithm to complete the data storage structure construction. The underlying hash table construction mechanism is a well-known technology in this field and will not be elaborated here. The hash mesh has the characteristic of allocating memory units only in the spatial region containing the unified coordinate point cloud. The hash mesh, combined with the dynamic mesh edge length values, reduces the computational load for data processing in suspended sediment clusters.

[0047] See Figure 5 In one embodiment, step S300 may further include the following sub-steps: S301, the envelope calculation module extracts the joint pose carried in the synchronous data stream. The joint pose records the absolute rotation angle value corresponding to each moving joint of the construction machinery. The construction machinery has a mechanical structure composed of multiple rigid links connected in series. Adjacent rigid links are connected by hinges through moving joints. The envelope calculation module reads the three-dimensional geometric model of the construction machinery and the link length value that are preset and stored in the system.

[0048] S302, the envelope calculation module combines the link length value with the absolute rotation angle value to perform mathematical operations to generate a local transformation matrix. The local transformation matrix defines the spatial relative positional relationship between adjacent rigid links. The envelope calculation module extracts multiple local transformation matrices from the bottom base of the construction machinery to the end-operating equipment and performs matrix multiplication to generate a positive kinematics matrix. The positive kinematics matrix describes the absolute spatial pose of each rigid link of the construction machinery relative to the reference coordinate system.

[0049] S303. For the underlying calculation process of constructing a positive kinematic matrix using link parameters, those skilled in the art can use standard kinematic modeling algorithms to complete the coordinate transformation link construction action. Kinematic parameter modeling technology is a well-known technology in this field and will not be described in detail here.

[0050] S304, the envelope solution module extracts the local coordinates of the surface vertices of the three-dimensional geometric model of the engineering machinery. The envelope solution module uses the positive kinematics matrix to perform a spatial multiplication mapping action on the local coordinates of the surface vertices to generate absolute geodetic coordinates. The envelope solution module uses triangular patches to connect the surface vertices with absolute geodetic coordinates to generate the shape of a three-dimensional closed surface. The envelope solution module establishes the shape of the three-dimensional closed surface as the theoretical envelope boundary.

[0051] S305, the theoretical envelope boundary defines the three-dimensional spatial coverage of the engineering machinery in an ideal, unforced rigid body state. The envelope calculation module maps the theoretical envelope boundary into the absolute geodetic coordinate space inside the hash grid. The theoretical envelope boundary and the unified coordinate point cloud are located in the same absolute geodetic coordinate space dimension. The theoretical envelope boundary provides a basic comparative boundary reference for subsequent acoustic point cloud spatial position determination and interference point cloud removal.

[0052] S306, the envelope solving module performs traversal addressing for independent voxel cells within the hash grid. The envelope solving module establishes a fixed-length time window in the time dimension. This time window adopts a sliding window mechanism and slides backward synchronously with the system clock according to a preset time step. The envelope solving module extracts a unified coordinate point cloud that falls within the spatial coverage of the independent voxel cells and is within the time range of the time window.

[0053] S307, the envelope solution module calculates the frequency of occurrence of independent voxel units in the unified coordinate point cloud and generates the occupancy integral value. Suspended sediment masses move with water flow and have the characteristic of random spatial position changes, while the static geological layer at the bottom of the water has the characteristic of relatively fixed spatial position. The occupancy integral value quantifies the continuous existence of spatial position.

[0054] S308, the envelope calculation module extracts the echo intensity values ​​carried by the unified coordinate point cloud, performs a discreteness calculation, and generates an echo intensity variance value. The uneven distribution of particles within suspended sediment masses causes drastic fluctuations in the reflected acoustic signal intensity. The surface of the static geological layer at the bottom of the water exhibits homogeneous reflectivity, resulting in a stable reflected acoustic signal intensity. The echo intensity variance value quantifies and assesses the stability of the acoustic reflective material. For the underlying mathematical calculations determining the degree to which the echo intensity values ​​deviate from the average value during the discreteness calculation, those skilled in the art can use the standard variance statistical algorithm to complete the data processing. The standard variance statistical algorithm is a well-known technology in this field and will not be elaborated upon here.

[0055] S309, the envelope calculation module adds the echo intensity variance value to the anti-zero constant value to generate the corrected variance value, and the envelope calculation module divides the occupancy integral value by the corrected variance value to generate the steady-state factor value.

[0056] ; In the formula, Represents the steady-state factor value; Represents the percentage of market share as a percentage of total sales. This represents the variance of the echo intensity. This represents the value of the zero constant.

[0057] The zero constant value prevents the program from crashing due to the denominator value being equal to zero. The increase in the steady-state factor value indicates an increased probability of the presence of underwater static geological layer point clouds within independent voxel units. The steady-state assessment action utilizes the steady-state factor value to fuse spatial occupancy characteristics and acoustic reflection characteristics to complete the multi-dimensional attribute quantification conversion action.

[0058] See Figure 6 In one embodiment, step S400 may further include the following sub-steps: S401, the error correction module extracts the three-dimensional coordinates of the vertices of the theoretical envelope boundary surface. When the construction machinery performs construction operations, the excavation reaction force causes the connecting rod deformation and joint clearance offset. The deformation and clearance offset cause a spatial deviation between the spatial position of the theoretical envelope boundary and the actual spatial position of the construction machinery. The error correction module performs a spatial expansion action along the outer normal direction of the theoretical envelope boundary according to the set distance parameters to generate an external spatial neighborhood.

[0059] S402, the error correction module extracts a unified coordinate point cloud within the external spatial neighborhood. The external spatial neighborhood limits the maximum distribution range of spatial deviation distance. The external spatial neighborhood reduces the number of data processing operations. The error correction module extracts the steady-state factor value corresponding to the unified coordinate point cloud. The error correction module reads the hard object threshold pre-written into the storage unit.

[0060] S403, the hard object threshold defines the characteristics of the reflected echo from the metal surface of the construction machinery and the scattered echo from the suspended silt. The error correction module combines the steady-state factor value with the hard object threshold to perform a numerical comparison. The unified coordinate point cloud with a steady-state factor value greater than the hard object threshold represents the data points generated by the sound wave reflection from the actual surface of the construction machinery.

[0061] To prevent excessive decay of the occupancy integral value due to high-speed movement of construction machinery, which could lead to misjudgment of point cloud removal, the error correction module extracts the time first derivative of the joint pose to calculate the instantaneous movement rate of the machinery, and uses the instantaneous movement rate of the machinery to perform negative adaptive compensation on the hard object threshold, thereby reducing the hard object judgment standard when the construction machinery is in a high-speed movement state.

[0062] When construction machinery is in motion, the frequency of its three-dimensional coordinate point cloud occupying independent voxel units decreases, which in turn causes the occupancy integral value to decay. However, the metal surface of construction machinery has extremely uniform acoustic reflection characteristics, which makes the echo intensity variance value within the time window always close to zero. This extremely low denominator value ensures the basic order of magnitude of the steady-state factor value. Combined with the aforementioned negative adaptive compensation action performed on the hard object threshold, it ensures that the steady-state factor value can still exceed the hard object threshold when the machinery is in motion.

[0063] S404, the error correction module performs an independent extraction action to generate feature anchor points for unified coordinate point clouds where the steady-state factor value is greater than the hard object threshold. For the underlying code operation of the spatial retrieval engine extracting point cloud data within a specific spatial range, those skilled in the art can use an octree data structure to complete the three-dimensional spatial coordinate retrieval action. The establishment of the octree data structure and the spatial coordinate retrieval logic are well-known technologies in the field and will not be elaborated here. Feature anchor points remove background noise data and suspended sediment data from the water body while retaining the actual three-dimensional coordinate data of the engineering machinery's surface.

[0064] S405, the error correction module extracts feature anchor points outside the theoretical envelope boundary, and calculates the perpendicular projection points of the feature anchor points on the outer surface of the theoretical envelope boundary to generate projection point coordinates. The error correction module also extracts the surface normal vector of the theoretical envelope boundary at the projection point coordinate position. For the underlying geometric calculation process of calculating the projection point coordinates of a point in three-dimensional space to a complex curved surface, those skilled in the art can use the nearest point iterative retrieval algorithm to complete the mapping action. The nearest point iterative retrieval algorithm is a well-known technology in this field and will not be described in detail here.

[0065] S406, the error correction module extracts the coordinates of the feature anchor points. It then subtracts the projection point coordinates from the feature anchor point coordinates using vector subtraction to generate a deviation vector. Finally, it combines this deviation vector with the surface normal vector using vector dot product to generate the spatial normal residual. This spatial normal residual quantifies the spatial offset caused by the deformation of the engineering machinery surface under stress.

[0066] ; In the formula, Represents the spatial normal residual; Represents the coordinates of the feature anchor point; Represents the coordinates of the projection point; Represents the surface normal vector.

[0067] S407, the error correction module extracts the vertex coordinates of the theoretical envelope boundary surface. The error correction module then calculates the 3D Euclidean distance between the projection point coordinates and the vertex coordinates of the theoretical envelope boundary surface to generate a distance weight parameter. This distance weight parameter decreases in value as the 3D Euclidean distance increases, matching the attenuation characteristics of local deformation propagating to the surrounding area in rigid connecting rods of engineering machinery.

[0068] S408, the error correction module multiplies the spatial normal residual by the distance weight parameter to generate a scalar compensation distance. It then extends this scalar compensation distance by performing a scalar multiplication with the corresponding surface normal vector to generate a compensation vector. Finally, it combines this compensation vector with the coordinates of the theoretical envelope boundary surface vertices to generate corrected vertex coordinates. The error correction module then uses these corrected vertex coordinates to reconstruct the 3D surface shape, generating a corrected envelope. This corrected envelope compensates for spatial position errors caused by mechanical deformation and joint clearance misalignment, ensuring the envelope accurately reflects the actual absolute spatial position of the engineering machinery within the underwater construction environment.

[0069] See Figure 7 In one embodiment, step S500 may further include the following sub-steps: S501: The multi-level engine module extracts a unified coordinate point cloud carrying steady-state factor values ​​and defines this unified coordinate point cloud as the point cloud to be rendered. The multi-level engine module extracts a corrected envelope, which limits the three-dimensional closed coverage area occupied by the engineering machinery within the absolute geodetic coordinate space. The multi-level engine module calculates the spatial positional relationship between the three-dimensional coordinates of the point cloud to be rendered and the boundary of the corrected envelope surface.

[0070] In step S502, when the multi-level engine module determines that the 3D coordinates of the point cloud to be rendered fall within the corrected envelope space, the multi-level engine module adds mechanical body attribute tags to the point cloud to generate a mechanical classification point cloud. For the underlying geometric calculation process of the relationship between the 3D coordinates of spatial points and the interior / exterior space of a 3D closed polygon, those skilled in the art can use a ray penetration intersection test algorithm to complete the spatial delimitation. The ray penetration intersection test algorithm is a well-known technology in this field and will not be elaborated upon here.

[0071] In S503, when the multi-level engine module determines that the 3D coordinates of the point cloud to be rendered are outside the corrected envelope space, it extracts the steady-state factor values ​​carried by the point cloud. The multi-level engine module reads the geological threshold and sediment threshold set within the storage unit. The hard object threshold is greater than the geological threshold, and the geological threshold is greater than the sediment threshold.

[0072] In S504, the multi-level engine module combines the steady-state factor values ​​with geological and sediment thresholds to perform interval comparisons. When the multi-level engine module determines that the steady-state factor value is greater than or equal to the geological threshold, it adds static geological attribute labels to the point cloud to be rendered, generating a geologically classified point cloud.

[0073] In S505, when the multi-level engine module determines that the steady-state factor value is less than the geological threshold but greater than or equal to the sediment threshold, it will attach dynamic sediment attribute labels to the point cloud to be rendered, generating a sediment-classified point cloud. When the multi-level engine module determines that the steady-state factor value is less than the sediment threshold, it will classify the point cloud to be rendered as water noise data and perform spatial culling.

[0074] This interval comparison action causes the original underwater geological layer area damaged by engineering machinery to automatically decrease the steady-state factor value due to the increased randomness of the point cloud spatial distribution. As the system operates, it dynamically degenerates into water noise or sediment classification point cloud, achieving real-time adaptive updates of three-dimensional visualization of underwater topography evolution without human intervention.

[0075] The S506 multi-level engine module merges mechanical, geological, and sediment classification point clouds to generate a classification point cloud. This classification point cloud utilizes spatial coverage and steady-state factor values ​​to perform the decoupling and separation of targets in three-dimensional space. The classification point cloud provides the data foundation for subsequent differentiated graphics memory resource allocation.

[0076] The S507 multi-level engine module establishes multiple parallel asynchronous rendering queues within the graphics processor. These asynchronous rendering queues include a high-frequency rendering queue, a mid-frequency rendering queue, and a low-frequency rendering queue. The multi-level engine module extracts the attribute tags carried by the classification point clouds. It then pushes the mechanical classification point cloud into the high-frequency rendering queue, the sediment classification point cloud into the mid-frequency rendering queue, and the geological classification point cloud into the low-frequency rendering queue. The classification point clouds are distributed to the corresponding parallel rendering channels according to their spatial dynamic evolution frequency.

[0077] S508, the multi-level engine module performs memory allocation calculations. The multi-level engine module calculates the total amount of local point cloud data contained in each asynchronous rendering queue, reads the system's preset baseline refresh rate parameter, and combines it with the normalized constant value. Then, it performs weighted independent calculations using the total amount of local point cloud data for each queue and the baseline refresh rate parameter to generate dedicated memory weight values ​​for high-frequency, mid-frequency, and low-frequency rendering queues.

[0078] ; In the formula, This represents the memory weight value; Represents the total amount of local point cloud data; Represents the baseline refresh rate parameter; This represents the value of the normalization constant.

[0079] The S509's multi-level engine module utilizes memory weight values ​​to allocate global memory capacity within the graphics processor. This module distributes isolated, independent memory resources to high-frequency, mid-frequency, and low-frequency rendering queues. These independent memory resources are matched to the bandwidth throughput requirements of specific point cloud classifications during the graphics rendering process.

[0080] The S510 multi-level engine module issues refresh commands with time-difference characteristics to the asynchronous rendering queue. It issues high-frequency refresh commands to the high-frequency rendering queue to ensure the continuous and smooth operation of the construction machinery. It issues medium-frequency refresh commands to the medium-frequency rendering queue to represent the gradual diffusion of suspended sediment clumps. Finally, it issues low-frequency refresh commands to the low-frequency rendering queue to reduce redundant computing power consumption caused by continuous redrawing of static underwater terrain.

[0081] S511, for the underlying interface control logic of multi-threaded graphics rendering instruction calls and concurrent execution, those skilled in the art can use the standard graphics application programming interface to complete the shader program scheduling action. The underlying graphics rendering pipeline concurrency control mechanism is a well-known technology in this field and will not be described in detail here.

[0082] The S512 multi-level engine module acquires the local rendering sub-layers generated by the parallel processing of various asynchronous rendering queues. The multi-level engine module extracts the spatial pixel depth values ​​from these sub-layers. It then uses depth buffer testing rules to compare the spatial pixel depth values ​​and performs layer compositing to generate output frames. Finally, the multi-level engine module sends these output frames to the graphics display device using a video transmission protocol. The graphics display device then displays a dynamic 3D visualization of the underwater construction operations of engineering machinery within the multi-beam sonar detection area on the screen.

[0083] Specific application examples are as follows: This embodiment was applied to an underwater dredging project in a port channel. The water depth was approximately 20 meters, and the engineering machinery was an underwater excavator equipped with a hydraulic excavating arm. The sonar equipment was fixed on an underwater base 15 meters away from the work area. The underwater current meter measured the current velocity of the surrounding water to be 1.5 meters per second.

[0084] When the excavator bucket of the construction machinery touches the bottom and begins to dig, the following system operation process is triggered: The hydraulic circuit pressure transmitter of the main controller detected a sudden increase in pressure. Since the straight-line distance between the excavation point and the sonar equipment is 15 meters, and the water flow velocity is 1.5 m / s, the timing delay module calculates a delay parameter of 15 / 1.5 = 10 seconds. This means it takes 10 seconds for the suspended sediment to drift to the detection area. After 10 seconds, the timing delay module translates this pressure surge into a delay rate, assuming a delay rate of 8.0 (normalized pressure units).

[0085] The mesh building module extracts the set static mesh side length value as 0.1 meters and the mapping coefficient value as 0.05. According to the formula... The dynamic mesh side length values ​​are calculated. .

[0086] The system predicted that sediment shading was about to occur and coarsened the hash grid resolution of the area from 10 cm to 50 cm in advance, reducing the CPU load caused by useless data processing.

[0087] In the underwater static geological layer region, due to the hardness and stability of the medium, the envelope calculation module calculates its occupancy integral value as 0.95 (frequent occurrence) and its echo intensity variance value as 0.02 (uniform reflection without fluctuation). The zero constant value is set to 0.01.

[0088] According to the formula The stability factor of the geological layer .

[0089] In the suspended sediment region, the disordered particles result in an integral value of 0.4 for the occupancy rate and a variance of 0.8 for the echo intensity. Its steady-state factor value... This difference provides a quantitative basis for subsequent space elimination.

[0090] Due to encountering hard reefs during excavation, the rigid connecting rods of the engineering machinery bent under stress. At this point, the error correction module detected a batch of uniform coordinate point clouds in the external spatial neighborhood with steady-state factor values ​​greater than the hard object threshold, and extracted them as feature anchor points.

[0091] Set the coordinates of a certain feature anchor point The coordinates of its projection point on the theoretical envelope boundary Corresponding surface normal vector .

[0092] According to the formula Calculate the spatial normal residual The error correction module uses this residual to inversely derive and generate a correction envelope, bending the excavator arm in the 3D model by 0.22 meters to make it fit the real point cloud.

[0093] After classifying the point cloud, the multi-level engine module calculates the data volume of the high-frequency rendering queue (mechanical classification point cloud). The base refresh rate is 50,000. 60Hz; Mid-frequency rendering queue (sediment classification point cloud) It is 150,000. The frequency is 15Hz. Let the normalization constant be... It is 10000.

[0094] According to the formula The memory weight value of the high-frequency queue ; intermediate frequency queue .

[0095] This led to a departure from the traditional practice of evenly allocating video memory, with the main bandwidth being allocated to high-speed moving mechanical components.

[0096] The experimental verification and effect comparison are as follows: To verify the effectiveness of this invention, we built a comparative test platform based on the same hardware environment on-site (an industrial data processing host equipped with an NVIDIA RTX A4000 graphics card). The test captured a 60-second complete construction process of the engineering machinery from its launch into the water to its encounter with a large-scale suspended sediment cloud.

[0097] Control group (traditional approach): The entire point cloud is synchronously fed into a single-threaded rendering pipeline, without steady-state evaluation of mud and sand, and without distinguishing between asynchronous rendering queues.

[0098] Experimental group (the present invention): The multi-beam sonar dynamic three-dimensional visualization system is started, and the timing delay module, mesh construction module and multi-level engine module are activated.

[0099] The test results recorded the frame rate (FPS) throughout the process. Since underwater engineering requires operators to control the excavator, a frame rate below 30 FPS is considered severe lag, directly impacting construction safety and efficiency.

[0100] See Figure 8 , Figure 8 The solid line represents the frame rate change trajectory of the output image using the system of this invention, while the dashed line represents the frame rate change trajectory using the traditional 3D visualization scheme. During the period from 0 to 15 seconds, the underwater environment is clear, and both methods maintain a frame rate of approximately 60 FPS. At the 15-second mark, the engineering machinery performs a powerful excavation operation, disrupting the underwater geological layer and generating a large amount of suspended sediment. As the delay parameters were exhausted, the sediment cloud entered the sonar detection area (starting from approximately the 25th second). Traditional solutions, attempting to render massive amounts of disordered sediment point clouds at the highest frequency, caused graphics memory resources to overflow, resulting in a precipitous drop in frame rate to around 15 FPS, with severe lag and tearing in the image. In contrast, the system of this invention reduced the amount of local data by performing scale expansion actions performed by the mesh construction module, and the multi-level engine module pushed the sediment classification point cloud into the mid-frequency rendering queue and reduced the weight allocation of graphics memory. This allowed the system's core computing power to be locked on the correction envelope rendering of the engineering machinery. After a brief fluctuation, the frame rate quickly and adaptively stabilized at over 55 FPS, ensuring the smoothness and accuracy of the dynamic 3D visualization of multibeam sonar for underwater construction throughout the process.

Claims

1. A multibeam sonar dynamic three-dimensional visualization system for underwater construction, characterized in that, It includes sonar equipment, main controller, inertial navigation equipment, underwater current meter, and data processing host; The sonar device outputs the raw point cloud, the main controller outputs the joint pose and pressure rate of the engineering machinery, the inertial navigation device outputs the attitude data of the sonar device and the engineering machinery, and the underwater current meter outputs the water flow velocity of the underwater construction environment. The data processing host integrates a synchronization alignment module, a timing delay module, a mesh construction module, an envelope calculation module, an error correction module, and a multi-level engine module. The synchronization alignment module receives the original point cloud, the joint pose, and the attitude data to generate a unified coordinate point cloud. The timing delay module extracts the absolute geodetic coordinates of the engineering machinery base corresponding to the peak moment of the pressure variability surge to establish a spatial locking point. The timing delay module obtains the straight-line distance between the three-dimensional coordinates of the sonar device and the spatial locking point; The timing delay module generates delay parameters based on the straight-line distance and the water flow speed; The timing delay module applies the delay parameter to the time axis of the pressure rate to generate a delayed rate; The mesh construction module receives the delay rate to generate a hash mesh, and the envelope calculation module combines the joint pose and the unified coordinate point cloud within the hash mesh to generate a theoretical envelope boundary and a steady-state factor value. The error correction module receives the theoretical envelope boundary, the unified coordinate point cloud, and the steady-state factor value to generate a corrected envelope; The multi-level engine module receives the corrected envelope and the steady-state factor value, extracts the unified coordinate point cloud as the point cloud to be rendered, and performs a spatial classification operation on the point cloud to be rendered to generate a classified point cloud.

2. The multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The inertial navigation device includes a first inertial navigation unit and a second inertial navigation unit; The first inertial navigation unit outputs a sonar 6-degree attitude, and the second inertial navigation unit outputs a mechanical 6-degree attitude. The sonar 6-degree attitude and the mechanical 6-degree attitude constitute the attitude data.

3. The multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 2, characterized in that, The synchronization alignment module extracts the time tags of the original point cloud; The synchronization alignment module extracts the instantaneous joint pose, instantaneous pressure rate, instantaneous sonar 6-degree attitude, and instantaneous mechanical 6-degree attitude at the corresponding time tag position. The synchronization alignment module constructs a homogeneous transformation matrix based on the instantaneous sonar 6-degree attitude and the instantaneous mechanical 6-degree attitude; The synchronization alignment module uses the homogeneous transformation matrix to process the original point cloud and the instantaneous joint pose to generate the unified coordinate point cloud.

4. The multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The mesh construction module extracts the static mesh side length values ​​and mapping coefficient values; The mesh construction module generates dynamic mesh side length values ​​based on the delay rate, the mapping coefficient values, and the static mesh side length values. The mesh construction module uses the dynamic mesh side length values ​​to generate a three-dimensional discrete index for the unified coordinate point cloud processing; The grid construction module uses a spatial hash mapping function to convert the three-dimensional discrete index into hash key values; The grid construction module establishes the correspondence between the hash key value and the memory address of the voxel unit to generate the hash grid.

5. A multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The envelope calculation module extracts the three-dimensional geometric model of the engineering machinery and the link length values ​​stored in the system. The envelope calculation module combines the link length value and the joint pose to generate a positive kinematic matrix; The envelope calculation module uses the positive kinematics matrix to map the local coordinates of the surface vertices of the three-dimensional geometric model of the engineering machinery to generate absolute geodetic coordinates. The envelope calculation module uses triangular facets to connect surface vertices with the absolute geodetic coordinates to generate the theoretical envelope boundary.

6. The multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The envelope solving module establishes a time window for the independent voxel units within the hash grid; The envelope calculation module extracts the unified coordinate point cloud that falls within the coverage area of ​​the independent voxel unit and is within the time window. The envelope calculation module calculates the frequency of occurrence of the unified coordinate point cloud occupying the independent voxel unit and generates an occupancy integral value. The envelope calculation module extracts the echo intensity values ​​from the unified coordinate point cloud to generate echo intensity variance values. The envelope calculation module generates the steady-state factor value based on the echo intensity variance value, the zero-prevention constant value, and the occupancy integral value.

7. A multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The error correction module generates an external spatial neighborhood along the outside of the theoretical envelope boundary according to a set distance parameter. The error correction module compares the steady-state factor value with the hard object threshold within the external spatial neighborhood to generate a feature anchor point. The error correction module calculates the spatial normal residual of the feature anchor point and the theoretical envelope boundary surface; The error correction module uses the spatial normal residual to compensate for the deformation error at the theoretical envelope boundary and generate the corrected envelope.

8. A multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 1, characterized in that, The multi-level engine module generates a mechanical classification point cloud when it determines that the point cloud to be rendered falls inside the correction envelope. When the multi-level engine module determines that the point cloud to be rendered is outside the correction envelope, it reads the geological threshold and the sediment threshold. The multi-level engine module combines the steady-state factor value with the geological threshold and the sediment threshold to perform an interval comparison. The multi-level engine module generates geological classification point clouds and sediment classification point clouds based on the comparison results. The multi-level engine module merges the mechanical classification point cloud, the geological classification point cloud, and the sediment classification point cloud to generate the classification point cloud.

9. A multibeam sonar dynamic three-dimensional visualization system for underwater construction according to claim 8, characterized in that, The multi-level engine module establishes an asynchronous rendering queue including a high-frequency rendering queue, a mid-frequency rendering queue, and a low-frequency rendering queue. The multi-level engine module pushes the mechanical classification point cloud into the high-frequency rendering queue; The multi-level engine module pushes the sediment classification point cloud into the mid-frequency rendering queue; The multi-level engine module pushes the geological classification point cloud into the low-frequency rendering queue; The multi-level engine module counts the total amount of local point cloud data in the asynchronous rendering queue; The multi-level engine module combines the baseline refresh rate parameter, the normalization constant value, and the total amount of local point cloud data to generate a video memory weight value. The multi-level engine module uses the memory weight values ​​to allocate independent memory resources in the global memory allocation; The multi-level engine module sends refresh commands to the asynchronous rendering queue to generate output frames.

Images