A multi-functional imaging method and system suitable for underwater unmanned platform

Abstract

The application discloses a kind of multi-functional imaging method and system suitable for underwater unmanned platform, it is related to the technical field of underwater navigation, the method comprises: constructing target echo model;The echo signal is carried out to get baseband signal by de-loading frequency and downsampling processing;Near-field echo is carried out time delay compensation;Three-dimensional data cube is generated by plane array two-dimensional beam forming algorithm;Three-dimensional point cloud data is extracted after filtering processing;Switch to forward-looking obstacle avoidance mode, based on point cloud data, seabed plane fitting is carried out to separate water point cloud, after clustering and evaluation, output obstacle detection result;Switch to the seabed imaging mode of downward-looking, the normal vector of fitted seabed plane is carried out coordinate conversion in combination with platform attitude information, and seabed inclination is calculated and output.The application realizes the multifunctional integration of obstacle avoidance and surveying by a set of hardware, reduces system redundancy, and improves the safety of platform navigation and bottom sitting.

Description

Technical Field

[0001] This invention relates to the technical field of underwater navigation, and more specifically, to a multifunctional imaging method and system suitable for underwater unmanned platforms. Background Technology

[0002] Underwater acoustic imaging is a key technology for underwater unmanned platforms to achieve environmental perception and autonomous navigation. In existing technologies, to meet the needs of obstacle avoidance and seabed topography mapping, underwater platforms typically need to carry two independent sonar systems simultaneously: forward-looking obstacle avoidance sonar and multibeam bathymetry sonar.

[0003] However, this traditional configuration has obvious drawbacks: First, the two independent systems result in heavy loads, high power consumption, and high costs for the underwater platform, and there is serious system redundancy; second, the forward obstacle avoidance information and seabed topography information come from different sensors, resulting in data fragmentation due to inconsistent spatiotemporal references; finally, the working principle of traditional multibeam bathymetry sonar dictates that it must rely on the continuous movement of the platform to perform mapping, and it cannot perform real-time online measurement and calculation of local seabed topographic dip angles when the platform is hovering and stationary (such as before anchoring or bottoming operations). Summary of the Invention

[0004] The purpose of this invention is to provide a multifunctional imaging method and system suitable for underwater unmanned platforms. It aims to achieve multifunctional integration of forward obstacle avoidance and downward seabed tilt angle detection through a hardware system and reconfigurable algorithm, thereby solving the problems of system redundancy, data fragmentation, and fixed working modes in the prior art.

[0005] The technical solution of this invention is: to provide a multifunctional imaging method suitable for underwater unmanned platforms, the method comprising:

[0006] S1. Construct a target echo model based on the incident field model, with the range-time scale and the spatial position of the receiving array elements as independent variables.

[0007] S2. The echo signal received by the receiving array element is digitally quadrature demodulated and downsampled through a cascaded integrator comb filter to convert the echo signal into a baseband signal.

[0008] S3. Determine whether the target is in the near field region based on the near field determination condition. If it is in the near field region, calculate the approximate time delay parameter and use the approximate time delay parameter to compensate for the time delay of the baseband signal.

[0009] S4. Using the baseband signal after time delay compensation, perform a two-dimensional discrete Fourier transform to generate two-dimensional multi-beam data. Perform modulus operation on the two-dimensional multi-beam data and combine it in order of distance dimension to generate three-dimensional cube data.

[0010] S5. Apply mean filtering and maximum value filtering to the three-dimensional cube data in sequence, and extract the three-dimensional point cloud data that meets the conditions by setting an adaptive intensity threshold.

[0011] S6. Switch to forward-looking obstacle avoidance mode for imaging acquisition. This mode uses the least squares method to fit the three-dimensional point cloud data to the seabed plane, calculates the vertical distance from the three-dimensional point cloud data to the fitted plane, and separates the seabed background from the water body point cloud by a distance threshold; performs cluster analysis on the water body point cloud to extract target features, and prioritizes obstacles based on a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section, and outputs the target detection results.

[0012] S7. Switch to the downward-looking seabed imaging mode for imaging acquisition. This mode obtains the normal vector in the platform coordinate system based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode. Combine the roll angle, pitch angle and heading angle of the underwater platform to construct a coordinate transformation matrix. The normal vector is transformed to the geographic coordinate system through the coordinate transformation matrix, and then the inclination angle of the seabed plane is calculated and output.

[0013] In any of the above technical solutions, further, in step S1, the constructed target echo model is represented in the frequency domain as:

[0014] ;

[0015] in, For the first The signal echo model received by each array element The total number of scattering points. The complex reflection coefficient of the target. The underwater sound absorption coefficient is... The distance from the sound source to the scattering target is denoted by . The spectrum of the transmitted signal. For wave number, For the scattered wave to propagate from the target point to the first The distance between each receiving array element.

[0016] In any of the above technical solutions, further, in step S2, after downsampling processing by a cascaded integrator-comb filter, the output baseband signal... for:

[0017] ;

[0018] in, The number of samples drawn is an integer multiple of the total number of samples drawn. The delay order of the comb-like portion, Let be the number of stages in the cascaded integrator-comb filter. This represents the Z-domain representation of a complex baseband signal.

[0019] In any of the above technical solutions, further, in step S3, the near-field determination condition is that the distance between the target and the sonar system is less than the working distance threshold, and the working distance threshold is equal to the square of the array aperture length divided by the wavelength of the sonar system's working signal.

[0020] Approximate time delay parameters The calculation formula is:

[0021] ;

[0022] in, For the target distance, Beam direction and Number the array elements. and These represent the element spacing in the horizontal and vertical directions, respectively. and These represent the angles of the target relative to the array element in the horizontal and vertical directions, respectively. The speed of sound.

[0023] In any of the above technical solutions, further, in step S5, the calculation formulas for mean filtering and maximum value filtering are as follows:

[0024] ;

[0025] ;

[0026] in, This is the data after mean filtering. The data after filtering for the maximum value. , , The size of the filter window in three dimensions. for Divide by the integer value of two, for Divide by the integer value of two, for Divide by the integer value of two, For three-dimensional cube data, This represents the filtering window.

[0027] In any of the above technical solutions, further, in step S6, when using the least squares method to fit the seabed plane to the three-dimensional point cloud data, let the seabed plane equation be... The optimal plane parameters are obtained by solving the following normal equations. , , :

[0028] ;

[0029] in, , , For point cloud coordinates, This represents the total number of data points.

[0030] Comprehensive evaluation function The calculation formula is:

[0031] ;

[0032] in, For the target size, The average intensity at each scattering point of the target. The scattering cross-section of the target. , , These are the weighting coefficients.

[0033] In any of the above technical solutions, further, in step S7, the constructed coordinate transformation matrix... for:

[0034] ;

[0035] in, The roll angle, For pitch angle, For heading angle;

[0036] The angle of inclination of the seabed The calculation formula is:

[0037] ;

[0038] in, and These are the horizontal and vertical components of the normal vector when transformed to the geographic coordinate system.

[0039] A multi-functional imaging system is also provided, which uses the multi-functional imaging method applicable to underwater unmanned platforms from any of the above technical solutions. The system includes:

[0040] The echo model construction module is used to construct a target echo model based on the incident field model, with the range-time scale and the spatial position of the receiving array elements as independent variables.

[0041] The carrier frequency removal and downsampling module is used to perform digital quadrature demodulation on the echo signal received by the receiving array element, and to perform downsampling processing through a cascaded integrator comb filter to convert the echo signal into a baseband signal;

[0042] The near-field compensation module is used to determine whether the target is in the near-field region based on the near-field determination conditions. If it is in the near-field region, it calculates the approximate time delay parameter and uses the approximate time delay parameter to compensate for the time delay of the baseband signal.

[0043] The two-dimensional beamforming module is used to perform windowing and zero-padding on the signals of each channel using the baseband signal after time delay compensation, perform two-dimensional discrete Fourier transform to generate two-dimensional multi-beam data, perform modulus operation on the two-dimensional multi-beam data and combine them in order of distance dimension to generate three-dimensional cube data.

[0044] The point cloud generation and filtering module is used to sequentially apply mean filtering and maximum value filtering to the 3D cube data, and extract the 3D point cloud data that meets the conditions by setting an adaptive intensity threshold.

[0045] The obstacle detection module is used to switch to the forward-looking obstacle avoidance mode. It uses the least squares method to fit the seabed plane to the 3D point cloud data, calculates the vertical distance from the point cloud data to the fitted plane, and separates the seabed background from the water body point cloud by a distance threshold. It performs cluster analysis on the water body point cloud to extract target features, and prioritizes obstacles based on a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section, and outputs the target detection results.

[0046] The seabed tilt detection module is used to switch to the downward-looking seabed imaging mode. Based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode, it obtains the normal vector in the platform coordinate system. It constructs a coordinate transformation matrix by combining the roll, pitch, and heading angles of the underwater platform. The normal vector is then transformed to the geographic coordinate system through the coordinate transformation matrix, and the tilt angle of the seabed plane is calculated and output.

[0047] The beneficial effects of this invention are:

[0048] This invention introduces a CIC filter for downsampling at the signal processing layer and combines full-array zero-padding with two-dimensional fast Fourier transform, which significantly reduces the computational complexity of multi-channel data and meets the real-time requirements of underwater platforms. At the same time, for close-range detection scenarios, a near-field Fresnel approximate time delay compensation is introduced to effectively overcome spherical wave distortion and provide high-precision true 3D point cloud data for subsequent target detection.

[0049] This invention does not simply physically superimpose two independent sonar systems. Instead, it uses the same set of 3D point cloud data, employing the least squares method to fit the seabed plane at the algorithm layer. By calculating the vertical distance threshold from the point cloud to the fitted plane, it achieves precise mathematical decoupling between the seabed background and suspended obstacles. This mechanism not only fundamentally eliminates the spatiotemporal reference fragmentation problem caused by multi-sensor fusion but also ensures the absolute homogeneity of obstacle avoidance and mapping data.

[0050] This invention breaks through the limitations of traditional multibeam echo sounders by extracting seabed normal vectors from a three-dimensional point cloud generated by a single acoustic pulse. Combined with a ZYX coordinate rotation matrix, it compensates in real-time for the roll, pitch, and heading disturbances of the underwater platform. This enables the underwater unmanned platform to instantly calculate high-precision seabed inclination angles in absolute geographic coordinates, even when hovering and stationary, filling the technological gap that prevents in-situ terrain assessment before anchoring and bottoming operations. Attached Figure Description

[0051] The advantages of the above and additional aspects of the present invention will become apparent and readily understood in the description of the embodiments in conjunction with the following drawings, wherein:

[0052] Figure 1 This is a schematic flowchart of a multi-functional imaging method applicable to an underwater unmanned platform according to an embodiment of the present invention;

[0053] Figure 2 This is an image showing the obstacle target imaging and detection results of a multifunctional imaging method and system applicable to underwater unmanned platforms according to an embodiment of the present invention.

[0054] Figure 3 This is an image showing the output of obstacle target information of a multifunctional imaging method and system applicable to underwater unmanned platforms according to an embodiment of the present invention.

[0055] Figure 4 This is an image of the original point cloud on the seabed, which is a result of a multifunctional imaging method and system applicable to underwater unmanned platforms according to an embodiment of the present invention.

[0056] Figure 5 This is a seabed plane fitting result diagram of a multifunctional imaging method and system applicable to underwater unmanned platforms according to an embodiment of the present invention;

[0057] Figure 6 This is a diagram showing the output of seabed tilt information of a multifunctional imaging method and system applicable to underwater unmanned platforms according to an embodiment of the present invention. Detailed Implementation

[0058] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0059] In the following description, many specific details are set forth in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0060] like Figure 1 As shown, this embodiment provides a multi-functional imaging method suitable for underwater unmanned platforms, the method comprising:

[0061] S1. Constructing the echo model: Based on the incident field model, construct a target echo model with the range-time scale and the spatial position of the receiving array elements as independent variables. The specific process is as follows:

[0062] Sonar transmitter radiates pulse signals into the water Its Fourier transform is The sound wave signal propagates in water, which has an absorption effect. For a spatial position vector... The first For a point-like scattering target, the incident sound field formed when sound waves reach the target from the emission source is represented in the frequency domain as:

[0063] ;

[0064] in, The spectrum of the transmitted signal; Indicates the underwater sound absorption coefficient; , Indicates the frequency of the transmitted signal; It indicates the speed of sound.

[0065] Constructing an echo model: The incident sound field excites the scattering point to generate a scattered wave, and the scattering intensity is determined by the target's complex reflection coefficient. Determined. This coefficient is a complex number, and its modulus is... The amplitude of sound wave reflection is characterized by its argument. The complex reflection coefficient characterizes the phase change introduced by the reflection process and is determined by the acoustic properties of the target itself, such as acoustic impedance and surface roughness. Subsequently, the scattered wave propagates from the target point to each element of the receiving array.

[0066] Assuming the planar receiver array is composed of Composed of array elements, the first The spatial position of each array element is Scattered waves from the target point Spread to the The distance between the receiving array elements is Assuming the actual underwater scenario includes... The scattering point, then the scattering point The signal echo model received by each array element is represented in the frequency domain as follows:

[0067] ;

[0068] in, For wave number, This indicates the distance from the sound source to the target scattering point.

[0069] The corresponding time-domain expression is as follows. This time-domain signal is the real-band communication signal obtained through the physical model, and its energy is concentrated in the carrier frequency of the transmitted signal. nearby:

[0070] .

[0071] S2. Carrier frequency removal and downsampling: To reduce the amount of data and computational complexity of subsequent 3D imaging processing, the time-domain echo signal received by the above-mentioned receiving array elements is processed. The signal is then decarrier-removed and downsampled to convert it into a baseband signal.

[0072] First, orthogonal demodulation is used to remove the carrier frequency: for each receiving channel... time domain signal or its sampled discrete signal Digital quadrature demodulation is performed. First, digital mixing is performed, mixing the signal with a pair of orthogonal local oscillator signals (frequency: Multiplying them together yields the in-phase (I) branch. Orthogonal (Q) branch Then, low-pass filtering is performed, passing the mixed I and Q signals separately through a low-pass filter with a cutoff frequency slightly greater than half the signal bandwidth B to filter out the interference. The high-frequency components centered on the baseband are used to obtain the baseband in-phase and quadrature components. and Thus, the complex baseband signal is obtained: .

[0073] Then downsampling is performed: because The highest frequency component has dropped to Downsampling is performed using a cascaded integrator-comb (CIC) filter. The CIC filter consists of paired integral and comb sections. The system transfer function of a single-stage integrator is... The system transfer function of the single-stage comb section is: ,in The number of samples drawn is an integer multiple of the total number of samples drawn. Let be the retardation order of the comb-like portion. One The transfer function of a cascaded CIC filter is:

[0074] ;

[0075] After CIC decimation filtering, the baseband signal output is:

[0076] ;

[0077] in, Complex baseband signal The Z-domain representation.

[0078] S3. Near-field Compensation: Based on the near-field determination criteria, time delay parameters are compensated for the echoes in the near field. In acoustic imaging, the wavefront in the near field region is a spherical wave, and the time delay difference between array elements has a non-linear relationship with distance; in the far field region, the wavefront can be approximated as a plane wave, and the time delay difference has a linear relationship with distance. The near-field determination criteria are set as follows:

[0079] ;

[0080] in, Indicates the working range of the sonar system; This represents the aperture length of the array. In the case of a square planar array, the aperture length represents the side length of the planar array. This indicates the wavelength of the sonar system's operating signal. The distance between the target and the sonar... satisfy When the target is determined to be in the near field, precise time delay compensation is required.

[0081] Under near-field conditions, the first The array element is at a distance of [missing information] from the target. Beam direction is Fresnel approximate time delay parameter Represented as:

[0082] ;

[0083] in, , They are respectively , Array element spacing in the direction, , The target is relative to the array element at... , The angle of direction. The baseband signal output in step S2 is compensated using the calculated time delay parameters to obtain time-delay-compensated multi-channel echo data. .

[0084] S4. Two-dimensional beamforming based on planar array: utilizing multi-channel echo data (i.e., the signal after time delay compensation in step S3), and generates two-dimensional multi-beam data through a planar array beamforming algorithm.

[0085] Channel data windowing processing: To suppress the sidelobe level of the beammap and improve imaging contrast, windowing is first applied to each range gate. The echo data from each channel is windowed. Let... It is a two-dimensional matrix representing the distance gate. The complex signals received by all array elements. Each element in the array is multiplied by a two-dimensional window function. Corresponding weights:

[0086]

[0087] Among them, symbols This indicates element-wise multiplication. The window function... Hanning window, Hamming window, or Taylor weighting can be used to achieve the best balance between main lobe width and side lobe suppression level.

[0088] Since real-world planar arrays may be sparse or have missing elements, zero-padding is required after windowing to facilitate efficient beamforming using the Fast Fourier Transform (FFT). Specifically, a matrix of size [size missing] is constructed. The zero matrix ,in and These represent the desired number of horizontal and vertical beams, respectively (usually powers of 2 for easier FFT calculation). The windowed data... Mapped to the actual physical location of the array elements The corresponding positions of the grid points not occupied by actual array elements are assigned a value of zero.

[0089] For each distance gate Full range of data Performing a two-dimensional discrete Fourier transform (2D-DFT) is typically achieved through a two-dimensional fast Fourier transform (2D-FFT). Horizontal beamforming corresponds to the pairing of matrices... FFT is performed on each row, and vertical beamforming corresponds to FFT on each column. The final two-dimensional beamforming result. for:

[0090] ;

[0091] in, It is a two-dimensional complex matrix, where the row indices correspond to different horizontal azimuth angles. The column indexes correspond to different vertical tilt angles. .

[0092] To obtain the energy value characterizing the acoustic scattering intensity, the two-dimensional beamforming results were analyzed. Perform modulo (absolute value) operation: The result That is, the target pixel in three-dimensional space The acoustic intensity value at the location. For all distance gates. Repeat steps S1 to S4 above to obtain all Combining the data according to the distance dimension, we finally obtain a three-dimensional cube data. .

[0093] S5. Generation and Filtering of 3D Point Cloud Data: The 3D acoustic intensity data obtained from beamforming in step S4, i.e., 3D cube data, is processed... Post-processing is performed, and effective 3D point cloud data is extracted using a filtering algorithm to generate a clear 3D acoustic image.

[0094] To suppress random noise and speckle noise, mean filtering (average filtering) is first applied to the three-dimensional sound intensity data. Mean filtering involves defining a sliding window in the three-dimensional spatial domain and replacing the center pixel value with the average value of the neighboring pixels within the window. Let the size of the filtering window be... The filtered data The calculation is as follows:

[0095] ;

[0096] in, , , .

[0097] To enhance the saliency of realistic targets while maintaining their edge sharpness, local maximum filtering is applied to the mean-filtered data. Local maximum filtering effectively enhances strong scattering points (such as target surfaces) while suppressing weak noise, helping to highlight the three-dimensional structure of realistic targets. Local maximum filtering uses the same or similar window size as mean filtering, and takes the maximum value within the window as the output value of the center pixel.

[0098] ;

[0099] After the above filtering process, 3D point cloud data is extracted using threshold detection. An adaptive intensity threshold is set. Extract the set of point cloud coordinates that meet the conditions:

[0100] ;

[0101] in, For distance coordinates, It is the horizontal azimuth. The vertical pitch angle. Each condition is satisfied. For a given point in three-dimensional space, its sound intensity value is .

[0102] S6. Obstacle Target Detection: Switch to forward-looking obstacle avoidance mode. Based on the 3D point cloud data obtained in step S5, through a series of processes such as seabed plane fitting, point cloud segmentation, cluster analysis, and target feature extraction, automatic detection and evaluation of underwater obstacles are achieved, and finally, a specified number of optimal target detection results are output. Details are as follows:

[0103] In forward-looking obstacle avoidance mode, the seabed plane is first fitted to the 3D point cloud data to separate the seabed background from underwater obstacles. The least squares method is used for plane fitting, and the equation of the seabed plane is given by: ,in For point cloud coordinates, , ,

[0104] Let be the parameters of the plane to be determined. The objective function of the least squares method is to minimize the sum of squared distances from all data points to the fitted plane:

[0105] ;

[0106] The optimal parameters are obtained by solving the normal equation:

[0107] ;

[0108] Through iterative calculations, the planar parameters are continuously optimized, and the optimal seabed plane fitting result is finally obtained.

[0109] Based on the optimal seabed plane obtained above, calculate the vertical distance from each point cloud data point to this plane:

[0110] ;

[0111] Set distance threshold , will satisfy The point cloud data is identified as water point cloud and retained, while the remaining point cloud data is removed as seabed background. This process effectively separates the seabed from underwater obstacles.

[0112] Point clouds of various water bodies were clustered based on Euclidean distance, resulting in multiple clusters. Specifically, these include:

[0113] Nearest neighbor distance calculation: for each point Find its nearest neighbor Calculate the average distance from each point. .

[0114] Cluster cluster generation: Initialize an empty cluster set; iterate through all unvisited points, and for the current point... ,like ,Will Add the points and their neighbors to the current cluster; otherwise, create a new cluster. Repeat until all points have been processed.

[0115] Cluster screening: Removing samples with too few samples ( ) or the average sound intensity is too low ( Clustering is used to eliminate noise interference; after completing Euclidean clustering, the point cloud can be divided into sub-clusters.

[0116] For each valid cluster, geometric features (point cloud density, spatial distribution, main dimensions), acoustic features (mean intensity, intensity variance, scattering characteristics), and spatial features (centroid location, eigenvalues ​​of the point cloud covariance matrix) are extracted. Based on the extracted features, a minimum bounding box is generated for each cluster to represent the approximate spatial occupancy of the target.

[0117] Design a comprehensive evaluation function to prioritize detected obstacles:

[0118]

[0119] in, The length / width / height of the target; The average intensity at each scattering point of the target; The scattering cross-section of the target; These are the weighting coefficients.

[0120] Based on the above priority scoring, the output should be... Each of the three optimal target detection results contains the target's three-dimensional position coordinates and target size information.

[0121] S7. Seabed Inclination Information Detection: In seabed imaging mode, by fitting the seabed plane and fusing platform information, the accurate calculation and output of seabed topographic inclination angles in both static and underway states are achieved.

[0122] Based on the seabed plane fitting results from step S6, the seabed plane equation in the platform coordinate system is obtained: The corresponding normal vector is After normalization, we get:

[0123] ;

[0124] Known platform attitude information includes roll angle Pitch angle Heading angle Using the ZYX rotation sequence (heading) Swaying (Roll), constructing a rotation matrix from the platform coordinate system to the geographic coordinate system. :

[0125] ;

[0126] The basic rotation matrix is:

[0127] ;

[0128] Then the complete transformation matrix for:

[0129] ;

[0130] Transform the seabed plane normal vector from the platform coordinate system to the geographic coordinate system:

[0131] ;

[0132] Transformed normal vector It indicates the direction of the seabed plane in the geographic coordinate system.

[0133] In a geographic coordinate system, the normal vector of the horizontal plane is assumed to be a vertically upward unit vector. Then the angle of inclination of the seabed surface The calculation is as follows:

[0134] ;

[0135] Ultimately, the seabed dip information is uploaded to the underwater platform, providing important topographic references for the platform's navigation decisions.

[0136] Based on the same inventive concept as the multifunctional imaging method in Embodiment 1 above, another embodiment of the present invention provides a multifunctional imaging system suitable for underwater unmanned platforms. This system aims to solve the problems of single-function, system redundancy, and data fragmentation in traditional sonar systems through a unified hardware platform and reconfigurable processing algorithms.

[0137] At the hardware architecture level, this system is mounted on an underwater unmanned platform (such as an autonomous underwater vehicle, AUV), and its main hardware components include: a sonar transmitter, and a... A planar receiving array composed of individual elements, an attitude sensor for real-time acquisition of the platform's roll, pitch, and yaw angles, and a core signal processor.

[0138] At the software and algorithm architecture level, the core signal processor is internally configured with multiple data processing modules that work collaboratively. Specifically, the data flow and module interaction mechanism during system operation is as follows:

[0139] First, the system simulates the underlying acoustic physics processes through an echo model construction module. This module controls the sonar transmitter to radiate pulse signals into the water and, by combining the underwater sound wave absorption coefficient, the target's complex reflection coefficient, and the spatial position of the planar receiving array, constructs a time-domain echo signal model containing multiple scattering points.

[0140] Subsequently, the actual echo signal acquired by the planar receiving array is sent to the carrier frequency removal and downsampling module. In order to reduce the massive amount of data and computational complexity of subsequent 3D imaging, this module performs digital quadrature demodulation (including digital mixing and low-pass filtering) on ​​the time-domain signal of each receiving channel to obtain the complex baseband signal, and uses a cascaded integrator comb (CIC) filter to downsample the complex baseband signal.

[0141] During the signal focusing phase, the system invokes the near-field compensation module. This module calculates the distance between the target and the sonar system in real time. When the target is determined to be in the near-field region, it calculates the Fresnel approximate time delay parameter and uses this parameter to accurately compensate for the time delay of the downsampled baseband signal.

[0142] Next, the compensated multi-channel echo data flows into the two-dimensional beamforming module. This module first multiplies each channel's data by a two-dimensional window function to suppress sidelobes, and then maps the data onto a regular two-dimensional grid for zero-padding. Based on this, a two-dimensional multi-beam data is generated by performing an efficient two-dimensional fast Fourier transform (2D-FFT), and after modulus taking and range dimension combination, a three-dimensional data cube containing the acoustic intensity of the target space is finally output.

[0143] To extract clear target structures from the massive data cube, the system is equipped with a point cloud generation and filtering module. This module defines a sliding window in the three-dimensional spatial domain and sequentially applies mean filtering and local maximum filtering to the three-dimensional sound intensity data to suppress random and speckle noise and enhance the saliency of the real target while maintaining edge sharpness. Finally, effective three-dimensional point cloud data is extracted through adaptive intensity threshold detection.

[0144] In particular, this system has a software-defined working mode switching function, which can expand two different working modes in parallel based on the generated 3D point cloud data:

[0145] When the system switches to forward-looking obstacle avoidance mode, the obstacle detection module is invoked. This module first uses the least squares method to fit the 3D point cloud data to the seabed plane, calculating the vertical distance from the point cloud to the fitted plane to remove the seabed background and extract the clean water point cloud. Next, it performs cluster analysis on the water point cloud using Euclidean distance as a feature, generating multiple obstacle clusters. For each cluster, geometric, acoustic, and spatial features are extracted to generate the minimum bounding box. Finally, this module uses a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section to prioritize detected obstacles and output the forward obstacle detection result to the underwater unmanned platform. Information on the distance, orientation, and size of the optimal obstacle avoidance target.

[0146] When the system switches to the downward-looking seabed imaging mode (e.g., when the platform is hovering or underway), the seabed tilt detection module is invoked. This module reuses the aforementioned seabed plane fitting results to obtain the seabed plane equation and normal vector in the platform coordinate system. Simultaneously, this module reads the roll, pitch, and heading angles transmitted from the attitude sensors in real time, constructs a coordinate transformation matrix using a ZYX rotation sequence, and transforms the seabed plane normal vector to the geographic coordinate system. Finally, by calculating the angle between this normal vector and the horizontal plane normal vector, the seabed topographic tilt angle is accurately calculated, providing decision support for the anchoring, bottom-sitting operations, or detailed geological exploration of the underwater unmanned platform.

[0147] In another embodiment of the present invention, a real-world sea trial was conducted to verify the effectiveness of the proposed multi-functional imaging method and system for underwater unmanned platforms. The sea data source was imaging data acquired by the multi-functional imaging system of the present invention mounted on an autonomous underwater vehicle (AUV).

[0148] In the experiment, the imaging and detection results of underwater scenes and obstacle targets were as follows: Figure 2 As shown. At the same time, as Figure 3 As shown, the system successfully outputs information about the obstacle target, including its distance, orientation, tilt, and dimensions.

[0149] In addition, the system acquires the raw seabed point cloud imaging results as follows: Figure 4 As shown in the figure. The original seabed point cloud was fitted to the seabed plane, and the result is shown in the figure. Figure 5 As shown. Combined with platform attitude information, such as... Figure 6 The system shown ultimately outputs accurate seabed plane tilt information. Experimental results demonstrate that this invention can effectively fit the seabed plane to raw seabed point cloud data and accurately output seabed tilt information, while simultaneously achieving high-precision obstacle detection.

[0150] In summary, this invention proposes a multifunctional imaging method suitable for underwater unmanned platforms, the method comprising:

[0151] S1. Constructing the echo model: Based on the incident field model, construct a target echo model with the range-time scale and the spatial position of the receiving array elements as independent variables.

[0152] S2. Carrier frequency removal and downsampling: The echo signal received by the receiving array element is digitally quadrature demodulated and downsampled through a cascaded integrator comb filter to convert the echo signal into a baseband signal.

[0153] S3. Near-field compensation: Determine whether the target is in the near-field region based on the near-field determination conditions. If it is in the near-field region, calculate the approximate time delay parameter and use the approximate time delay parameter to compensate for the time delay of the baseband signal.

[0154] S4. Two-dimensional beamforming based on planar array: Using the baseband signal after time delay compensation, the signals of each channel are windowed and zero-padding is performed to generate two-dimensional multi-beam data. The two-dimensional multi-beam data is then subjected to modulus operation and combined in order of distance dimension to generate three-dimensional cube data.

[0155] S5. Generation and Filtering of 3D Point Cloud Data: Mean filtering and maximum value filtering are applied sequentially to the 3D cube data, and 3D point cloud data that meets the conditions is extracted by setting an adaptive intensity threshold.

[0156] S6. Obstacle Target Detection: Switch to forward-looking obstacle avoidance mode, use the least squares method to fit the seabed plane to the 3D point cloud data, calculate the vertical distance from the point cloud data to the fitted plane, and separate the seabed background from the water body point cloud by a distance threshold; perform cluster analysis on the water body point cloud to extract target features, and prioritize obstacles based on a comprehensive evaluation function that includes target size, average intensity and scattering cross-section, and output the target detection results.

[0157] S7. Seabed Inclination Information Detection: Switch to the downward-looking seabed imaging mode. Based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode, obtain the normal vector in the platform coordinate system. Combine the roll angle, pitch angle and heading angle of the underwater platform to construct a coordinate transformation matrix. Transform the normal vector to the geographic coordinate system through the coordinate transformation matrix, and then calculate and output the inclination angle of the seabed plane.

[0158] A multifunctional imaging system using the above method is also provided, the system comprising:

[0159] The echo model construction module is used to construct a target echo model based on the incident field model, with the range-time scale and the spatial position of the receiving array elements as independent variables.

[0160] The carrier frequency removal and downsampling module is used to perform digital quadrature demodulation on the echo signal received by the receiving array element, and to perform downsampling processing through a cascaded integrator comb filter to convert the echo signal into a baseband signal.

[0161] The near-field compensation module is used to determine whether the target is in the near-field region based on the near-field determination conditions. If it is in the near-field region, it calculates the approximate time delay parameter and uses the approximate time delay parameter to compensate for the time delay of the baseband signal.

[0162] The two-dimensional beamforming module is used to perform windowing and zero-padding on the signals of each channel using the baseband signal after time delay compensation, perform two-dimensional discrete Fourier transform to generate two-dimensional multi-beam data, perform modulus operation on the two-dimensional multi-beam data and combine them in order of distance dimension to generate three-dimensional cube data.

[0163] The point cloud generation and filtering module is used to sequentially apply mean filtering and maximum value filtering to the 3D cube data, and extract the 3D point cloud data that meets the conditions by setting an adaptive intensity threshold.

[0164] The obstacle detection module is used to switch to the forward-looking obstacle avoidance mode. It uses the least squares method to fit the seabed plane to the 3D point cloud data, calculates the vertical distance from the point cloud data to the fitted plane, and separates the seabed background from the water body point cloud by a distance threshold. It performs cluster analysis on the water body point cloud to extract target features, and prioritizes obstacles based on a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section, and outputs the target detection results.

[0165] The seabed tilt detection module is used to switch to the downward-looking seabed imaging mode. Based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode, it obtains the normal vector in the platform coordinate system. It constructs a coordinate transformation matrix by combining the roll, pitch, and heading angles of the underwater platform. The normal vector is then transformed to the geographic coordinate system through the coordinate transformation matrix, and the tilt angle of the seabed plane is calculated and output.

[0166] The steps in this invention can be adjusted, combined, or deleted according to actual needs.

[0167] The units in the device of the present invention can be merged, divided, or reduced according to actual needs.

[0168] In this invention, the terms "installation," "connection," "linking," and "fixing" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; "linking" can be a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of these terms in this invention according to the specific circumstances.

[0169] The shapes of the components in the accompanying drawings are schematic and may differ from their actual shapes. The drawings are only used to illustrate the principles of the present invention and are not intended to limit the present invention.

[0170] Although the invention has been disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and not intended to limit the application of the invention. The scope of protection of the invention is defined by the appended claims and may include various modifications, alterations, and equivalents made to the invention without departing from the scope and spirit of the invention.

Claims

1. A multi-functional imaging method suitable for underwater unmanned platforms, characterized in that, The method includes: S1. Construct a target echo model based on the incident field model, with the range-time scale and the spatial position of the receiving array elements as independent variables. S2. The echo signal received by the receiving array element is digitally quadrature demodulated and downsampled through a cascaded integrator comb filter to convert the echo signal into a baseband signal. S3. Determine whether the target is in the near field region based on the near field determination condition. If it is in the near field region, calculate the approximate time delay parameter and use the approximate time delay parameter to compensate for the time delay of the baseband signal. S4. Using the baseband signal after time delay compensation, perform a two-dimensional discrete Fourier transform to generate two-dimensional multi-beam data. Perform modulus operation on the two-dimensional multi-beam data and combine it in order of distance dimension to generate three-dimensional cube data. S5. Apply mean filtering and maximum value filtering to the three-dimensional cube data in sequence, and extract the three-dimensional point cloud data that meets the conditions by setting an adaptive intensity threshold. S6. Switch to forward-looking obstacle avoidance mode for imaging acquisition. This mode uses the least squares method to fit the three-dimensional point cloud data to the seabed plane, calculates the vertical distance from the three-dimensional point cloud data to the fitted plane, and separates the seabed background from the water body point cloud by a distance threshold; performs cluster analysis on the water body point cloud to extract target features, and prioritizes obstacles based on a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section, and outputs the target detection results. S7. Switch to the downward-looking seabed imaging mode for imaging acquisition. This mode obtains the normal vector in the platform coordinate system based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode. Combine the roll angle, pitch angle and heading angle of the underwater platform to construct a coordinate transformation matrix. The normal vector is transformed to the geographic coordinate system through the coordinate transformation matrix, and then the inclination angle of the seabed plane is calculated and output.

2. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S1, the constructed target echo model is represented in the frequency domain as follows: ； in, For the first The signal echo model received by each array element The total number of scattering points. The complex reflection coefficient of the target. The underwater sound absorption coefficient is... The distance from the sound source to the scattering target is denoted by . The spectrum of the transmitted signal. For wave number, For the scattered wave to propagate from the target point to the first The distance between each receiving array element.

3. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S2, after downsampling processing by a cascaded integrator-comb filter, the output baseband signal... for: ； in, The number of samples drawn is an integer multiple. The retardation order of the comb-like portion is... Let be the number of stages in the cascaded integrator-comb filter. This represents the Z-domain representation of a complex baseband signal.

4. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S3, the near-field determination condition is that the distance between the target and the sonar system is less than the working distance threshold, and the working distance threshold is equal to the square of the array aperture length divided by the wavelength of the sonar system's working signal. Approximate time delay parameters The calculation formula is: ； in, For the target distance, Beam direction and Number the array elements. and These represent the element spacing in the horizontal and vertical directions, respectively. and These represent the angles of the target relative to the array element in the horizontal and vertical directions, respectively. The speed of sound.

5. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S5, the calculation formulas for mean filtering and maximum value filtering are as follows: ； ； in, This is the data after mean filtering. The data after maximum value filtering. , , The size of the filter window in three dimensions. for Divide by the integer value of two, for Divide by the integer value of two, for Divide by the integer value of two, For three-dimensional cube data, This represents the filtering window.

6. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S6, when fitting the seabed plane to the three-dimensional point cloud data using the least squares method, the equation of the seabed plane is set as follows: The optimal plane parameters are obtained by solving the following normal equations. , , : ； in, , , For point cloud coordinates, This represents the total number of data points. Comprehensive evaluation function The calculation formula is: ； in, For the target size, The average intensity at each scattering point of the target. The scattering cross-section of the target. , , These are the weighting coefficients.

7. The multi-functional imaging method for underwater unmanned platforms as described in claim 1, characterized in that, In step S7, the coordinate transformation matrix is ​​constructed. for: ； in, The roll angle, For pitch angle, For heading angle; The angle of inclination of the seabed The calculation formula is: ； in, and These are the horizontal and vertical components of the normal vector when transformed to the geographic coordinate system.

8. A multi-functional imaging system using the multi-functional imaging method for underwater unmanned platforms as described in any one of claims 1 to 7, characterized in that, The multifunctional imaging system includes: The echo model construction module is used to construct a target echo model based on the incident field model, with the range-time scale and the spatial position of the receiving array elements as independent variables. The carrier frequency removal and downsampling module is used to perform digital quadrature demodulation on the echo signal received by the receiving array element, and to perform downsampling processing through a cascaded integrator comb filter to convert the echo signal into a baseband signal; The near-field compensation module is used to determine whether the target is in the near-field region based on the near-field determination conditions. If it is in the near-field region, it calculates the approximate time delay parameter and uses the approximate time delay parameter to compensate for the time delay of the baseband signal. The two-dimensional beamforming module is used to perform windowing and zero-padding on the signals of each channel using the baseband signal after time delay compensation, perform two-dimensional discrete Fourier transform to generate two-dimensional multi-beam data, perform modulus operation on the two-dimensional multi-beam data and combine them in order of distance dimension to generate three-dimensional cube data. The point cloud generation and filtering module is used to sequentially apply mean filtering and maximum value filtering to the 3D cube data, and extract the 3D point cloud data that meets the conditions by setting an adaptive intensity threshold. The obstacle detection module is used to switch to the forward-looking obstacle avoidance mode. It uses the least squares method to fit the seabed plane to the 3D point cloud data, calculates the vertical distance from the point cloud data to the fitted plane, and separates the seabed background from the water body point cloud by a distance threshold. It performs cluster analysis on the water body point cloud to extract target features, and prioritizes obstacles based on a comprehensive evaluation function that includes target size, average intensity, and scattering cross-section, and outputs the target detection results. The seabed tilt detection module is used to switch to the downward-looking seabed imaging mode. Based on the seabed plane equation fitted in the forward-looking obstacle avoidance mode, it obtains the normal vector in the platform coordinate system. It constructs a coordinate transformation matrix by combining the roll, pitch, and heading angles of the underwater platform. The normal vector is then transformed to the geographic coordinate system through the coordinate transformation matrix, and the tilt angle of the seabed plane is calculated and output.

Images