Underwater multi-modal identification method and system for high dam hidden micro-defects

Abstract

The application provides an underwater multi-modal identification method and system for high dam hidden micro defects, and the method comprises the following steps: acquiring displacement monitoring data, stress monitoring data and temperature monitoring data, generating a multi-source monitoring data matrix through time synchronization processing, extracting abnormal features by using an adaptive weight method to obtain a suspicious area list with priority; planning a sonar scanning path according to the suspicious area list, obtaining edge feature data through multi-level signal enhancement processing to determine an accurate defect area; collecting acoustic data and optical data, and obtaining a fusion feature matrix through time and space registration and feature fusion; performing feature discrimination on the fusion feature matrix to determine a defect type and grade, obtaining positioning results and reliability evaluation results through multi-level analysis, and finally generating a defect identification report. The application improves the accuracy and reliability of underwater hidden micro defect identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-modal data processing, in particular to an underwater multi-modal identification method and system for hidden micro-defects of high dams. BACKGROUND

[0002] With the continuous expansion of water conservancy construction scale, the safe operation of dams is facing more and more severe challenges. The underwater hidden micro-defects of high dams are one of the important factors threatening the safety of dams. These defects often show the characteristics of small size, non-obvious features and fast development. Timely detection and accurate identification of underwater hidden micro-defects are of great significance to ensure the safe operation of dams, prolong the service life and reduce the maintenance cost.

[0003] At present, underwater defect detection mainly adopts single modal detection means, such as sonar scanning, optical imaging or displacement monitoring. Traditional sonar detection detects underwater structures by emitting sound waves and receiving echo signals, but due to the influence of equipment performance and underwater environment, there are problems such as limited resolution and serious noise interference; although optical detection can obtain intuitive image information, the imaging quality is seriously degraded in turbid water, and the detection distance is limited; although displacement monitoring can reflect the structure deformation in real time, the monitoring points are limited and it is difficult to cover all possible defect positions. These single detection means have obvious limitations in practical application.

[0004] In actual engineering application, the current detection technology still has the following specific problems: first, the data collected by different sensors have inconsistent sampling frequencies and uneven data quality, which leads to difficulties in multi-source data fusion and poor abnormal feature extraction effect; second, in complex underwater environment, sonar signals are easily affected by temperature stratification, suspended solids and bubbles, etc., causing attenuation and distortion of echo signals, affecting the accurate extraction of defect features; third, due to the complex and variable defect morphology, the existing feature extraction methods are difficult to simultaneously consider the feature expression of different types of defects such as cracks, erosion and leakage, reducing the accuracy of identification; fourth, in the process of multi-modal data processing, the spatial and temporal alignment accuracy of different types of data is insufficient, affecting the feature fusion effect; fifth, there is a lack of scientific reliability evaluation mechanism, it is difficult to quantitatively evaluate the accuracy and stability of the detection results, affecting the credibility of the detection results. The existence of the above technical problems seriously restricts the effect and application promotion of underwater hidden micro-defect detection. SUMMARY

[0005] The present application provides an underwater multi-modal identification method and system for hidden micro-defects of high dams, which can solve at least one technical problem existing in the prior art.

[0006] Technical scheme, the underwater multi-modal identification method for hidden micro-defects of high dams, comprising the following steps:

[0007] Step S1, real-time acquisition of displacement monitoring data, stress monitoring data and temperature monitoring data from the sensor network database, and generation of a multi-source monitoring data matrix through time synchronization processing; extraction of abnormal features from the multi-source monitoring data matrix using an adaptive weight calculation method to generate an abnormal feature matrix; input of the abnormal feature matrix into a pre-configured multi-dimensional analysis model to calculate a suspicious region coordinate set and simultaneously generate a credibility score; based on the credibility score, priority sorting of the suspicious region coordinate set to output a suspicious region list;

[0008] Step S2, acquisition of the suspicious region list and planning of a scanning path according to the suspicious region list; based on the scanning path, acquisition of original sonar data, and through multi-level signal enhancement processing, obtaining enhanced sonar data; based on the enhanced sonar data, extraction of edge feature data, multi-feature analysis, determination of an accurate defect region and defect feature description;

[0009] Step S3, based on the accurate defect region and defect feature description, collection of acoustic data and optical data, and through space-time alignment, obtaining registration data; enhancement processing of the registration data to obtain enhanced feature data; based on the enhanced feature data, extraction and fusion of features to obtain multi-modal feature data; based on the multi-modal feature data, generation of a fusion feature matrix and a feature credibility index;

[0010] Step S4, based on the fusion feature matrix and the feature credibility index, through feature discrimination, obtaining a defect discrimination result; based on the defect discrimination result, multi-level analysis to determine a defect type and a defect level; based on the defect type and the defect level, calculation of an accurate spatial position to obtain a positioning result; based on the positioning result, comprehensive evaluation to obtain a reliability evaluation result, generation of a defect identification report and an early warning parameter update list.

[0011] An underwater multi-modal recognition system for hidden micro-defects of a high dam, comprising:

[0012] at least one processor; and,

[0013] a memory in communication connection with the at least one processor; wherein,

[0014] the memory stores instructions executable by the processor, the instructions being executed by the processor to implement the underwater multi-modal recognition method for hidden micro-defects of a high dam.

[0015] Beneficial effects, the present application realizes the rapid and accurate positioning of suspicious areas, improves the quality and reliability of detection data, and enhances the completeness and accuracy of feature expression; realizes the accurate recognition and quantitative evaluation of defects, and solves the technical problems of data heterogeneity, environmental interference, feature extraction difficulty and low recognition accuracy in underwater defect detection of a dam. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 is a flowchart of the present application.

[0017] Figure 2 is a flowchart of step S1 of the present application.

[0018] Figure 3 is a flowchart of step S2 of the present application.

[0019] Figure 4 is a flowchart of step S3 of the present application.

[0020] Figure 5 is a flowchart of step S4 of the present application. DETAILED DESCRIPTION

[0021] As shown in Figure 1 , the present application proposes an underwater multi-modal identification method for hidden micro-defects of high dams, comprising the following steps:

[0022] Step S1, real-time acquisition of displacement monitoring data, stress monitoring data and temperature monitoring data from a sensor network database, and generation of a multi-source monitoring data matrix through time synchronization processing; extraction of abnormal features from the multi-source monitoring data matrix using an adaptive weight calculation method to generate an abnormal feature matrix; input of the abnormal feature matrix into a pre-configured multi-dimensional analysis model to calculate a suspicious region coordinate set and simultaneously generate a credibility score; based on the credibility score, priority sorting of the suspicious region coordinate set to output a suspicious region list;

[0023] Step S2, acquisition of the suspicious region list and planning of a scanning path according to the suspicious region list; based on the scanning path, acquisition of original sonar data, and through multi-level signal enhancement processing, obtaining enhanced sonar data; based on the enhanced sonar data, extraction of edge feature data, multi-feature analysis, determination of an accurate defect region and defect feature description;

[0024] Step S3, based on the accurate defect region and defect feature description, collection of acoustic data and optical data, and through spatio-temporal alignment, obtaining registered data; enhancement processing of the registered data to obtain enhanced feature data; based on the enhanced feature data, extraction and fusion of features to obtain multi-modal feature data; based on the multi-modal feature data, generation of a fusion feature matrix and a feature credibility index;

[0025] Step S4, based on the fusion feature matrix and the feature credibility index, through feature discrimination, obtaining a defect discrimination result; based on the defect discrimination result, multi-level analysis to determine a defect type and a defect level; based on the defect type and the defect level, calculation of an accurate spatial position to obtain a positioning result; based on the positioning result, comprehensive evaluation to obtain a reliability evaluation result, generation of a defect identification report and an early warning parameter update list.

[0026] As Figure 2 shown, according to one aspect of the application, step S1 is further:

[0027] Step S11, read the displacement monitoring data corresponding to the three-dimensional space position from the displacement sensor acquisition node, read the stress monitoring data of the same position from the stress sensor acquisition node, and read the temperature monitoring data of the same position from the temperature sensor acquisition node; sample frequency detection is performed on the displacement monitoring data, stress monitoring data and temperature monitoring data respectively, and a sampling feature vector is generated; based on the sampling feature vector, a linear interpolation parameter is constructed, the displacement monitoring data, stress monitoring data and temperature monitoring data are resampled and time-aligned, and time-aligned data is generated; the time-aligned data is organized according to the space position and the sensor type, and a multi-source monitoring data matrix is output;

[0028] The process of time alignment is specifically: T(i,j) = Tb(i) + α [D(i,j) / Dmax] [1 + β Q(i,j)]; wherein T(i,j) is the aligned time point; Tb(i) is the reference time sequence; D(i,j) is the data delay; Dmax is the maximum allowed delay; Q(i,j) is the data quality factor; α is the delay compensation coefficient (0.5-1.0); β is the quality adjustment factor (0.2-0.4). S(i) = Σ[wk |T(i,k)-T(i,k-1)|] exp(-γ V(i)); wherein S(i) is the synchronization score; wk is the time weight; V(i) is the data variability; γ is the stability coefficient (0.3-0.6).

[0029] The process of organizing the space position includes: a space coordinate unified transformation algorithm: P'(x,y,z) = R P(x,y,z) + t + Δ(x,y,z); wherein P'(x,y,z) is the transformed coordinate; P(x,y,z) is the original coordinate; R is the rotation matrix; t is the translation vector; Δ(x,y,z) is the nonlinear correction term. Δ(x,y,z) = Σ[wi B(x,y,z)] [1 + α E(x,y,z)]; wherein B(x,y,z) is the basis function value; E(x,y,z) is the environmental influence factor; wi is the basis function weight; α is the environmental correction coefficient (0.1-0.3).

[0030] Step S12, divide the multi-source monitoring data matrix into a core data part representing the main change trend and a peripheral data part representing fluctuation changes, calculate adaptive weight coefficients of the core data part and the peripheral data part; add the core data part and the peripheral data part after multiplying the core data part and the peripheral data part by corresponding adaptive weight coefficients, to obtain weighted monitoring data; based on the weighted monitoring data, calculate distances between each monitoring point and its adjacent monitoring points; convert the distances into correlation weight values, the correlation weight values decreasing with the increase of the distances between the monitoring points; multiply the correlation weight values by the weighted monitoring data, to obtain an abnormal feature matrix;

[0031] wherein the multi-source monitoring data matrix is divided into a core data part representing the main change trend and a peripheral data part representing fluctuation changes, and the following algorithm is used: core data and peripheral data decomposition algorithm: U(x, y) =∑[wi Di(x, y)] [1+α F(x, y)]; wherein U(x, y) is a decomposition eigenvalue; Di(x, y) is an i-th layer decomposition result; wi is a hierarchical weight; F(x, y) is a frequency feature; α is a feature adjustment coefficient (0.2~0.4). Mc = U(x, y) [1 + β |gradU(x, y)|]; Mp = M is Mc; wherein Mc is the core data; Mp is the peripheral data; M is the original data; gradU(x, y) is a feature gradient; β is a gradient weight (0.1~0.3), and grad is a gradient operator.

[0032] Step S13, for each monitoring point in the abnormal feature matrix, calculate a local correlation value with surrounding monitoring points and a global correlation value with all monitoring points; based on a weighted combination of the local correlation value and the global correlation value, construct an enhanced feature matrix;

[0033] Step S14, based on the enhanced feature matrix, calculate a dynamic threshold value of each monitoring point, the threshold value changing with the increase of the distance from a reference point; record monitoring points exceeding the corresponding dynamic threshold value to form a suspicious region coordinate set;

[0034] Step S15, for each region in the suspicious region coordinate set, respectively calculate a displacement abnormality degree, a stress abnormality degree and a temperature abnormality degree; combine the displacement abnormality degree, the stress abnormality degree and the temperature abnormality degree through adaptive weights to obtain a credibility score; based on the credibility score, sort the suspicious region coordinate set to generate a suspicious region list with priorities.

[0035] The process of combining the three abnormality degrees by adaptive weight to obtain the reliability score is as follows: multi-index fusion calculation of reliability: C(r) = λ1P1(r) + λ2P2(r) + λ3P3(r) [1 + αS(r)]; wherein C(r) is the reliability score; P1, P2, and P3 are displacement, stress, and temperature indexes; λ1, λ2, and λ3 are dynamic weights; S(r) is spatial correlation; and α is a correlation coefficient (0.2-0.4). λi = exp(-|Pi-μi| / σi) [1 + βTi]; wherein μi is a historical mean value; σi is a standard deviation; Ti is a time stability; and β is a stability weight (0.1-0.3).

[0036] The embodiment obtains displacement, stress, and temperature monitoring data in real time, combines time synchronization processing and an adaptive weight calculation method, and solves the technical problems of strong heterogeneity of dam monitoring data, inconsistent sampling frequencies, and uneven data quality. Specifically, time synchronization is performed by linear interpolation, which can align data of different sampling frequencies to a unified time sequence, avoiding misjudgment caused by data time sequence misalignment; adaptive weight coefficients of core data and peripheral data are calculated, effectively balancing the influence of main trends and fluctuation changes, and improving the accuracy of anomaly detection; the distance correlation weight between monitoring points is calculated, fully considering the influence of spatial position relationship on abnormal features, so that the finally generated abnormal feature matrix not only reflects the change law of the time dimension, but also contains the correlation characteristics of the spatial dimension. The embodiment improves the accuracy of suspicious area identification, reduces the false positive rate, and can provide a reliable reliability score, providing effective priority guidance for subsequent accurate detection.

[0037] In an embodiment of the present application, an abnormal feature enhancement function can also be used to calculate and construct enhanced feature values, and form an enhanced feature matrix: F(x, y) = Fb(x, y) [1 + αL(x, y)] [1 + βG(x, y)]; wherein F(x, y) is an enhanced feature; Fb(x, y) is a basic feature; L(x, y) is a local abnormality degree, i.e., a local correlation value; G(x, y) is a global abnormality degree, i.e., a global correlation value; and α and β are weight coefficients (0-0.5). L(x, y) = Σ[wi|f(x, y)-f(xi, yi)| / d(i)]; wherein f(x, y) is a feature value; d(i) is a spatial distance; wi is a distance weight; and xi and yi are neighborhood point coordinates.

[0038] As shown in FIG. 2, according to one aspect of the present application, step S2 further includes: Figure 3

[0039] ​Step S21, based on the priority and spatial coordinates of each suspicious region in the suspicious region list, calculate the horizontal scanning angle and the vertical scanning angle, and generate a sonar scanning path; control the sonar device to collect echo signals of the target region according to the sonar scanning path, and obtain original sonar data containing distance values and signal strengths;

[0040] Step S22, based on the original sonar data, calculate the signal mean and variance of the local region; based on the mean, determine the signal gain coefficient; based on the variance, determine the noise suppression coefficient; multiply the signal gain coefficient with the original sonar data to obtain gain sonar data; multiply the noise suppression coefficient with the gain sonar data to obtain enhanced sonar data;

[0041] Step S23, based on the enhanced sonar data, calculate the gradient values in the horizontal and vertical directions; based on the gradient values, calculate the edge strength and direction angle, and determine the adaptive threshold combined with the statistical characteristics of the local region; record the edge points higher than the adaptive threshold as edge feature data;

[0042] Wherein the accurate extraction of edge features is: G(x, y) = sqrt(Gx 2 + Gy 2 ) [1 +α D(x,y)];wherein G(x, y) is the edge strength; Gx, Gy is the horizontal and vertical gradient; D(x, y) is the direction consistency; α is the direction weight (0.2~0.4). E(x, y) = G(x, y) exp(-β |θ(x, y)-θm|); wherein E(x, y) is the edge feature; θ(x, y) is the local direction; θm is the main direction; β is the direction weight (0.3~0.6).

[0043] Step S24, based on the edge feature data, calculate the intensity feature, shape feature and texture feature of each edge point; match the intensity feature, shape feature and texture feature with the preset defect feature template to obtain the feature matching probability; mark the region with a feature matching probability higher than a preset threshold as an accurate defect region;

[0044] Wherein the process of feature matching is: feature matching probability calculation: P(f, t) = exp(-||f t|| 2 / σ1 2 ) [1 + α S(f, t)]; wherein P(f, t) is the matching probability; f is the feature to be matched; t is the template feature; S(f, t) is the structural similarity; α is the similarity weight (0.2~0.4); σ1 is the scale parameter. M(i) =Σ[wj P(fi, tj)] [1 +β C(i)]; wherein M(i) is the comprehensive matching degree; wj is the feature weight; C(i) is the context consistency; β is the context weight (0.1~0.3).

[0045] Step S25, based on the accurate defect area, the geometric size, edge definition, echo intensity and spatial continuity of the area are calculated respectively; the geometric size, edge definition, echo intensity and spatial continuity are combined to form a defect feature description, and the reliability index of the feature is calculated to generate a defect detection report containing the defect position, feature and reliability.

[0046] The embodiment is based on the suspicious area list to plan the sonar scanning path and process data collection, solves the technical problems of complex underwater environment, sonar signal easy to be disturbed and low detection efficiency. By introducing the environmental parameters such as water depth, temperature and salinity to calculate the sound velocity profile, combining the target distance and water turbidity to optimize the transmission power and receiving gain, the adaptive adjustment of sonar parameters is realized, and the signal-to-noise ratio of echo signal is improved; the improved ant colony algorithm is used to optimize the detection sequence, and the scanning density is dynamically adjusted according to the area priority, which ensures the detection accuracy of key areas and improves the overall detection efficiency; in the signal processing link, the influence of platform swing and sound velocity change is eliminated through attitude compensation and sound path correction, the gain adjustment is carried out by using overlapping sliding window and segmented linear mapping, which effectively improves the quality and detail performance of sonar image. The embodiment can obtain clear and reliable sonar data in complex underwater environment, and provides high-quality basic data for defect recognition.

[0047] According to one aspect of the present application, step S21 further comprises:

[0048] Step S211, read the spatial coordinate information of the suspicious area list, based on the spatial coordinate information, use the adaptive grid decomposition algorithm to divide the detection area into a predetermined sub-area, generate grid division data; read the current flow data and obstacle data from the pre-stored environmental database, combine the grid division data to calculate the detection difficulty coefficient of each sub-area, output the difficulty coefficient matrix; combine the priority information in the suspicious area list with the difficulty coefficient matrix by weighting to generate an area priority matrix; based on the area priority matrix, use the improved ant colony algorithm for optimization to output the initial scanning path;

[0049] Wherein the adaptive grid resolution algorithm is specifically: C(i,j) = Cb [1 + alpha D(i,j)] [1 + beta O(i,j)] [1 + gamma P(i,j)]; Wherein: C(i,j) is the grid size; Cb is the basic grid size; D(i,j) is the depth variation rate; O(i,j) is the occlusion degree; P(i,j) is the priority coefficient; alpha, beta, gamma are adjustment factors (0~1). Q(i,j) = exp(-lambda1 V(i,j)) exp(-lambda2 R(i,j)); Wherein Q(i,j) is the grid quality score; V(i,j) is the visibility index; R(i,j) is the resolution requirement; lambda1, lambda2 are weight coefficients.

[0050] The improved ant colony algorithm (S211) is specifically: tauij(t+1) = (1-p) tauij(t) + delta tauij(t) [1 + alpha Q(i,j)]; Wherein tauij(t) is the pheromone concentration; p is the volatilization coefficient (0.10.3); delta tauij(t) is the pheromone increment; Q(i,j) is the path quality; alpha is the quality weight (0.20.4). P(i,j) = [tauij(t)] α [etaij(t)] β [1 + gamma V(i,j)] / Sigma [tauik(t)] α [etaik(t)] β ; Wherein P(i,j) is the transition probability; etaij(t) is the heuristic information; V(i,j) is the visibility adjustment term; alpha, beta are the pheromone and heuristic information weights; gamma is the adjustment coefficient (0.1~0.3). E(i,j) = w1 L(i,j) + w2 D(i,j) + w3 C(i,j); Wherein E(i,j) is the path evaluation function; L(i,j) is the path length; D(i,j) is the detection difficulty; C(i,j) is the coverage efficiency; w1, w2, w3 are evaluation weights. K(t) = Kb [1+exp(-lambda t)] [1 + mu S(t)]; Wherein K(t) is the convergence control factor; Kb is the basic convergence coefficient; S(t) is the search stagnation degree; lambda is the time scale parameter; mu is the stagnation weight (0.1~0.3).

[0051] Step S212, based on the initial scanning path and the pre-stored water depth data, generate the sound velocity profile data; based on the sound velocity profile data, calculate the attenuation coefficient of the sonar beam at different depths to generate an attenuation coefficient table; based on the attenuation coefficient table, calculate the optimal transmission power and receiving gain from the target area distance data; based on the optimal transmission power and receiving gain, combine the pre-stored turbidity data to perform parameter optimization to generate a parameter configuration table;

[0052] Step S213, based on the parameter configuration table and the area priority matrix, the optimal pitch angle and horizontal rotation angle of each detection position are calculated, and a multi-angle scanning sequence is generated; the sonar rotating mechanism is controlled to move according to the multi-angle scanning sequence, and echo signal data is collected in real time;

[0053] Step S214, attitude data is obtained from an inertial navigation system, and based on the attitude data, the echo signal data is compensated for attitude to generate compensated signal data; a sound range correction parameter is extracted from the sound velocity profile data, and the compensated signal data is corrected for sound range to generate corrected signal data; all the corrected signal data are time-aligned and spatially spliced to output original sonar data.

[0054] The embodiment solves the technical problems of strong blindness and low efficiency of underwater detection through multi-level sonar scanning path planning and parameter optimization. The detection area is divided by using an adaptive grid decomposition algorithm, the detection difficulty coefficient is calculated in combination with flow and obstacle data, so that the path planning is more in line with the actual working conditions; an adaptive adjustment mechanism of sonar parameters is established by analyzing the sound velocity profile and attenuation characteristics, the transmission power and reception gain are dynamically optimized according to the target distance and water turbidity, and the quality of the sonar signal is improved; the detection sequence is optimized based on an improved ant colony algorithm, and the scanning density is adjusted by using an area priority matrix, so that the detection resources are reasonably allocated; through attitude compensation and sound range correction, the influence of platform motion and sound velocity change is eliminated, and the accuracy of the sonar data is ensured. The embodiment not only improves the detection efficiency, but also ensures the data quality, and lays a solid foundation for subsequent defect identification.

[0055] In another embodiment of the present application, in step S211, the following method can also be used to optimize and calculate the scanning path to obtain an initial scanning path: R(θ, φ) = Rb [1 + α D(θ, φ)] [1 + β O(θ, φ)]; wherein R(θ, φ) is the scanning radius; Rb is the basic scanning radius; D(θ, φ) is the depth variation rate; O(θ, φ) is the obstruction degree; α and β are adjustment coefficients (0~0.5). E(p) = w1 L(p) + w2 T(p) + w3 C(p); wherein E(p) is the path evaluation value; L(p) is the path length; T(p) is the time consumption; C(p) is the coverage degree; w1, w2 and w3 are weight coefficients.

[0056] According to an aspect of the present application, step S22 further comprises:

[0057] Step S221, the original sonar data is divided into overlapping local analysis windows, the signal mean value, variance and peak value distribution in each local analysis window are calculated, and a statistical feature matrix is generated; based on the statistical feature matrix, an adaptive window algorithm is used to optimize the window size, and optimized statistical data are output.

[0058] wherein the adaptive window algorithm is specifically: W(x, y) = Wb [1 + a E(x, y)] [1 + b N(x, y)]; wherein W(x, y) is the window size; Wb is the basic window size; E(x, y) is the edge intensity factor; N(x, y) is the noise level; a is the edge adjustment coefficient (0.3-0.7); b is the noise suppression coefficient (0.2-0.5). H(x, y) = exp(-|G(x, y)| / s1) exp(-|L(x, y)| / s2); wherein H(x, y) is the window uniformity; G(x, y) is the gray gradient; L(x, y) is the local variance; s1, s2 are scale parameters.

[0059] Step S222, based on the optimized statistical data, calculating the signal dynamic range of each window to generate a dynamic range vector; based on the dynamic range vector, constructing a piecewise linear mapping function to generate a gain mapping table; extracting the gain control parameter of each signal segment from the gain mapping table, and performing segmented gain adjustment on the original sonar data according to the gain control parameter to output gain-adjusted data;

[0060] Step S223, performing time-frequency analysis on the gain-adjusted data to extract the frequency characteristics and energy distribution of the noise; based on the frequency characteristics and energy distribution, constructing a multi-layer noise model to identify system noise, environmental noise and interference noise, and generating a noise feature vector and a noise suppression coefficient;

[0061] Step S224, based on the noise feature vector and the noise suppression coefficient, constructing an adaptive filter bank to process the signal in different frequency bands to obtain processed data; based on the processed data, using a wavelet threshold method to eliminate noise, and reconstructing the signal to obtain enhanced sonar data.

[0062] wherein the wavelet threshold method is specifically: T(j, k) = s j [log(N) / N] a [1 + b |W(j, k)| / s j]; wherein T(j, k) is the adaptive threshold value of the jth layer and the kth wavelet coefficient; s j is the noise standard deviation of the jth layer; N is the signal length; W(j, k) is the wavelet coefficient; a is the scale adjustment factor (0.3-0.7); b is the coefficient adjustment factor (0.1-0.5).

[0063] The embodiment solves the technical problems of poor underwater sonar signal quality and unobvious detail features by multi-stage signal enhancement and adaptive parameter optimization. Local analysis is performed by using overlapping sliding windows, and the optimal window size is automatically determined by calculating statistical features, thereby avoiding the blind area and redundancy caused by fixed windows. A segmented linear mapping method based on signal dynamic range is designed, and the gain control parameter is adaptively calculated, which not only enhances the detail performance of the weak signal region, but also avoids the saturation distortion of the strong signal region. Time-frequency analysis means is introduced to identify different types of noise features, and a hierarchical noise suppression is performed by using a wavelet threshold method, thereby retaining the effective signal and reducing the background noise. A multi-layer filter bank is used to realize the signal processing in different frequency bands, and the signal is adaptively weighted and reconstructed according to the importance of the features in different frequency bands, thereby improving the signal-to-noise ratio and resolution of the signal. The embodiment not only improves the quality and clarity of the sonar image, but also maintains the authenticity and integrity of the defect features, thereby providing a high-quality data basis for subsequent feature extraction.

[0064] As shown in Figure 4 According to one aspect of the present application, step S3 further comprises:

[0065] Step S31, based on the accurate defect area and defect feature description, controlling the acoustic sensor to collect acoustic raw data of the target area at different angles, and controlling the underwater camera to collect optical raw data at the corresponding position, recording the collection time and spatial coordinates of each data; based on the collection time and spatial coordinates, aligning the acoustic raw data and the optical raw data to generate registration data;

[0066] The collection process of the optical and acoustic raw data is as follows: according to the spatial coordinates of the accurate defect area, reading the water quality parameter data and water flow velocity data of the current water area from the database, and dividing the detection space into working sub-regions by using an adaptive grid method. The water quality parameter data and the water flow velocity data are input into the obstacle avoidance evaluation model to calculate the passing difficulty coefficient of each sub-region, and a region difficulty matrix is generated. According to the region difficulty matrix, a dynamic obstacle avoidance path planning algorithm is used to generate a device initial path, and the device initial path is input into an energy consumption optimization model to generate a device operation path and device control parameters. According to the device operation path, the acoustic sensor collects acoustic raw data of the target area at different angles, and according to the device control parameters, the underwater camera collects optical raw data at the corresponding position.

[0067] The process of aligning the acoustic raw data and the optical raw data based on the collection time and the spatial coordinates comprises: constructing a time-space registration matrix: T(p, q) = Tb [1 + a D(p, q)] [1 + b V(p, q)]; wherein T(p, q) is a time-space transformation matrix; Tb is a basic transformation matrix; D(p, q) is a spatial displacement; V(p, q) is a velocity influence; a and b are adjustment factors (0~0.5). R(p, q) = exp(-|gradT(p, q)| / s1) exp(-|H(p, q)| / s2); wherein R(p, q) is a registration reliability; gradT is a transformation gradient; H(p, q) is a registration entropy; s1 and s2 are control parameters.

[0068] In step S32, the acoustic part in the registration data is read, and high-frequency components and low-frequency components are calculated; based on the high-frequency components and the low-frequency components, enhancement is performed respectively, and then weighted combination is performed to obtain enhanced acoustic data; the optical part in the registration data is read, and enhancement of contrast and texture features is performed respectively, and then weighted combination is performed to obtain enhanced optical data; based on a preset quality evaluation index, a complementary enhancement coefficient is calculated, the enhanced acoustic data and the enhanced optical data are combined, and enhanced feature data is obtained;

[0069] In step S33, based on the enhanced feature data, spectral features, time-domain features and energy features are extracted from the enhanced acoustic data, acoustic feature data is generated; edge features, texture features and shape features are extracted from the enhanced optical data, optical feature data is generated; similarity between the acoustic feature data and the optical feature data is calculated, and a feature mutual verification matrix is generated;

[0070] In step S34, based on the acoustic feature data and the optical feature data, a quality evaluation result is generated, a feature weight is calculated, and a weighted feature is obtained; the weighted feature is fused to obtain multi-modal feature data; based on the multi-modal feature data, filtering is performed according to a preset feature importance, and a fusion feature matrix is generated;

[0071] The process of calculating the feature weight and fusion comprises: calculating a feature fusion weight: W(i, j) = Wb(i, j) [1 + a Q(i, j)] [1 + b D(i, j)]; wherein W(i, j) is a fusion weight; Wb(i, j) is a basic weight; Q(i, j) is a feature quality; D(i, j) is a feature redundancy; a and b are adjustment coefficients (0~0.5). Q(i, j) = exp(-|F(i, j)-m| / s1) exp(-|gradF(i, j)| / s2); wherein F(i, j) is a feature value; m is a mean value; gradF is a feature gradient; s1 and s2 are scale parameters.

[0072] Step S35, calculate the geometric size, physical property, dynamic response and spatial distribution of each feature in the fused feature matrix to generate a feature description vector; based on the feature description vector, calculate the credibility of the acoustic feature data, the optical feature data and the feature mutual evidence matrix respectively, and combine them to obtain a feature credibility index.

[0073] The embodiment solves the technical problems of limited information and poor anti-interference ability of single detection means through multi-modal data acquisition and fusion processing. Path planning is performed based on water quality parameters and water flow velocity, a dynamic obstacle avoidance algorithm is used to generate an optimal operation path, and the efficiency and safety of data acquisition are improved; in the data acquisition link, the acoustic sensor and the low-light camera are cooperatively controlled, and the acquisition parameters are dynamically adjusted according to the environmental conditions to ensure the quality of acoustic and optical data; in the data processing aspect, a space-time registration algorithm is used to accurately align heterogeneous data, a quality evaluation model is used to calculate the reliability weight to ensure the accuracy of data fusion; in the feature extraction link, the feature performance of acoustic and optical data is enhanced respectively, and adaptive weight method is used for feature fusion to fully exert the complementary advantages of the two modal data. The embodiment not only improves the reliability of detection, but also enhances the environmental adaptability of the system.

[0074] According to one aspect of the present application, step S31 further comprises:

[0075] Step S311, read the spatial coordinates of the accurate defect area, and divide the detection space into working sub-regions using an adaptive grid method; based on the working sub-regions, the pre-stored water flow velocity and the obstacle distribution, calculate the passing difficulty coefficient; based on the passing difficulty coefficient, use a dynamic obstacle avoidance path planning algorithm to generate an initial path of the underwater vehicle; based on the initial path of the underwater vehicle, use a pre-configured energy consumption model for optimization to obtain an operation path of the underwater vehicle;

[0076] Wherein the adaptive grid method is specifically: S(i,j) = Sb [1 +α D(i,j)] [1 +β V(i,j)]; Wherein S(i,j) is the grid size; Sb is the basic grid size; D(i,j) is the depth variation rate; V(i,j) is the water flow velocity influence factor; α, β are adjustment coefficients (0~1). Q(i,j) = exp(-γ C(i,j)) exp(-η O(i,j)); Wherein Q(i,j) is the grid quality score; C(i,j) is the complexity index; O(i,j) is the obstacle density; γ, η are weight coefficients.

[0077] Step S312, based on the operation path of the underwater vehicle and the defect feature description, calculate the optimal working distance of the acoustic sensor and the low-light camera to generate a device arrangement scheme; based on the device arrangement scheme, use a multi-device cooperative control algorithm to calculate the working time sequence of each sensor and output a cooperative acquisition instruction;

[0078] Step S313, read real-time water quality parameters and ambient light intensity, dynamically adjust the exposure time, gain value and color enhancement parameter of the low-light camera, generate a camera parameter table; based on the camera parameter table, obtain the reflection characteristics of the target area, calculate the optimal supplementary light angle and intensity, and output the supplementary light control instruction;

[0079] Step S314, based on the cooperative acquisition instruction and the supplementary light control instruction, receive the original acoustic data and original image data collected by each sensor, extract the timestamp and spatial position marker of the acoustic data, and extract the acquisition time and camera pose information of the image data; based on the timestamp, spatial position marker, acquisition time and camera pose information, a space-time registration algorithm is used to generate a space-time correspondence table;

[0080] The space-time registration algorithm is specifically: E(R, t) = wsEs(R, t) + wtEt(R, t) + wcEc(R, t); wherein Es(R, t) is a spatial registration error term; Et(R, t) is a time synchronization error term; Ec(R, t) is a feature consistency error term; ws, wt, wc are weight coefficients and ws+wt+wc=1. Es(R, t) = Σ||ps(i)-(R pt(i)+t)|| 2 exp(-d(i) / ds); wherein ps(i), pt(i) are the spatial coordinates of the corresponding feature points; R is a rotation matrix; t is a translation vector; d(i) is the distance between feature points; ds is a spatial scale parameter.

[0081] Step S315, based on the space-time correspondence table, calculate the signal-to-noise ratio and clarity of the acoustic data to obtain the acoustic quality index; based on the space-time correspondence table, calculate the clarity and contrast of the image data to obtain the image quality index; based on the acoustic quality index and the image quality index, a multi-index fusion evaluation method is used to generate registration data and a data quality report.

[0082] In another embodiment of the present application, acoustic original data and optical original data collected by each sensor are received, the sound wave emission time, reception time and device pose data are read from the acoustic sensor to generate acoustic space-time markers. The image acquisition time, camera pose and GPS position information are read from the underwater camera to generate optical space-time markers. The acoustic space-time markers and the optical space-time markers are input into a space-time registration model to calculate the spatial mapping relationship and the time correspondence relationship of the two kinds of data, and a registration mapping matrix is generated. According to the registration mapping matrix, the space-time alignment of the acoustic original data and the optical original data is performed, and the space-time aligned data is output.

[0083] For the acoustic part in the spatio-temporal alignment data, the peak ratio, signal-to-noise ratio and spectral definition of the signal are calculated to generate an acoustic quality vector. For the optical part in the spatio-temporal alignment data, the definition, contrast and illumination uniformity of the image are calculated to generate an optical quality vector. The acoustic quality vector and the optical quality vector are input into a quality evaluation model to calculate the reliability weight of each data and generate a data weight matrix. The spatio-temporal alignment data are weighted and fused according to the data weight matrix to output registration data and a registration quality report.

[0084] The embodiment solves the technical problems of great difficulty and unstable quality in underwater multi-modal data acquisition through precise device arrangement and cooperative control. An adaptive grid division method based on water quality parameters and obstacle distribution is proposed, the path of the underwater vehicle is optimized by calculating the difficulty coefficient of passing, and the operation efficiency and safety are improved; a calculation model of the optimal working distance of the acoustic sensor and the low-light camera is designed, the precise arrangement of the acquisition devices is realized by combining the multi-device cooperative control algorithm, and the spatial coverage and time synchronization of data acquisition are ensured; the exposure parameters and light compensation strategy of the camera are dynamically adjusted according to the real-time water quality parameters and environmental light intensity, and the image quality is effectively improved; the real-time evaluation mechanism of data quality is established by analyzing the signal-to-noise ratio of acoustic data and the definition of image data, and the usability of the acquired data is ensured. The embodiment not only ensures the integrity and synchronization of the data, but also improves the reliability and efficiency of the acquisition process.

[0085] According to an aspect of the present application, step S32 further comprises:

[0086] In step S321, the acoustic signals in the registration data are subjected to multi-scale wavelet decomposition, the feature components in different frequency bands are extracted, the energy distribution of each feature component is calculated, and a frequency domain feature matrix is generated; based on the frequency domain feature matrix, the frequency band correlation is analyzed to obtain frequency domain correlation data; wherein the feature components in different frequency bands include high frequency components, medium frequency components and low frequency components;

[0087] In step S322, based on the frequency domain feature matrix and the frequency domain correlation data, the high frequency components are subjected to directional enhancement, the medium frequency components are subjected to selective retention, and the low frequency components are subjected to noise suppression, the components in each frequency band after fusion processing are generated, and enhanced acoustic data are generated;

[0088] In step S323, the optical images in the registration data are subjected to geometric correction and illumination equalization, an adaptive histogram model is constructed, and the non-uniform illumination effect is compensated to generate corrected image data;

[0089] In step S324, a multi-scale contrast enhancement algorithm is applied to the corrected image data, the enhancement parameters are adaptively adjusted in combination with the statistical features of the local region, the performance of details and textures is optimized, and enhanced optical data are output;

[0090] Wherein the multi-scale contrast enhancement algorithm is specifically: E(x, y) = I(x, y) + λ Σ[wk Dk(x, y)]; wherein E(x, y) is the enhanced image; I(x, y) is the original image; Dk(x, y) is the detail map of the kth scale; wk is the scale weight coefficient; λ is the enhancement intensity factor (0.5~2.0). Dk(x, y) = Gk(x, y)-Gk+1(x, y); wherein: Gk(x, y) is the kth layer Gaussian pyramid image; wk = exp(-|Dk(x, y)| / σk); σk is the kth layer scale parameter.

[0091] Or, E(x, y) = I(x, y) [1 + α C(x, y)] [1 + β T(x, y)]; wherein E(x, y) is the enhancement result; I(x, y) is the original image; C(x, y) is the contrast term; T(x, y) is the texture enhancement term; α, β are enhancement coefficients (0.3~0.6). T(x, y) = Σ[wi Li(x, y)] exp(-γ N(x, y)); wherein Li(x, y) is the i-th layer Laplace feature; N(x, y) is the noise evaluation; wi is the hierarchical weight; γ is the noise suppression coefficient (0.2~0.4).

[0092] Step S325, calculate the complementary features of the enhanced acoustic data and the enhanced optical data, and construct the feature fusion weight; based on the feature fusion weight, the adaptive weighting method is used to combine the enhanced acoustic data and the enhanced optical data to generate the enhanced feature data.

[0093] Wherein the process of generating the enhanced feature data is specifically: M(i, j) =λ1 F1(i, j) +λ2 F2(i, j) +θ I(F1, F2); wherein M(i, j) is the fusion feature; F1, F2 is two kinds of features; λ1, λ2 is the feature weight; I(F1, F2) is the mutual information term; θ is the mutual information weight (0.1~0.3). I(F1, F2) =Σ[P(x, y) log(P(x, y) / P(x)P(y))] [1 + α C(x, y)]; wherein P(x, y) is the joint probability; P(x), P(y) is the marginal probability; C(x, y) is the local correlation; α is the correlation coefficient (0.2~0.4).

[0094] The embodiment solves the technical problems of unobvious features and poor complementarity of underwater heterogeneous data by independent enhancement and synergistic fusion of acoustic and optical data. The frequency domain features of acoustic signals are extracted by multi-scale wavelet decomposition, the differentiated enhancement of different frequency bands is realized by analyzing the frequency band correlation, and the information quantity of acoustic data is improved. An adaptive geometric correction and illumination equalization algorithm is designed to effectively compensate the influence of uneven illumination and improve the contrast and detail performance of the image. A multi-scale contrast enhancement algorithm is introduced to adaptively adjust the enhancement parameters according to the local region features and optimize the texture performance of the image. By analyzing the complementary features of acoustic and optical data, an adaptive feature fusion weight is designed to maintain the respective advantages and realize effective information complementarity. The embodiment improves the expression ability and distinguishability of the features and provides high-quality data support for subsequent feature fusion.

[0095] As shown in Figure 5 According to one aspect of the present application, step S4 further comprises:

[0096] Step S41, based on the feature credibility index, the fusion feature matrix is compared and calculated with the pre-stored structural similarity data, time sequence correlation data and spatial position correlation data to obtain a feature discrimination matrix; based on the feature weight in step S34, the feature discrimination matrix is weighted and calculated to output a defect discrimination result;

[0097] Step S42, based on the defect discrimination result, the matching degree with the preset crack feature library, the erosion feature library, the leakage feature library and the deformation feature library is calculated respectively to obtain feature matching data; based on the feature matching data and the defect discrimination result, the defect type is determined and the defect level is divided;

[0098] Step S43, based on the defect type and the defect level, combined with the pre-stored sonar positioning data, optical positioning data and monitoring positioning data, the spatial positioning data is calculated according to the respective reliability weight; based on the spatial positioning data, the positioning error analysis is carried out to generate error evaluation data; based on the error evaluation data, the final positioning result and the positioning confidence interval are determined;

[0099] Step S44, the recognition accuracy index of the defect discrimination result, the spatial precision index of the positioning result, the feature time stability index and the spatial consistency index are calculated to generate evaluation index data; based on the evaluation index data, the comprehensive reliability score is obtained by weighted combination to form the reliability evaluation result;

[0100] Step S45, based on the reliability evaluation result, the defect type, the defect level and the pre-stored historical evolution trend data, a warning index value is calculated, a warning level data is generated; based on the warning level data, the key monitoring point list, the monitoring parameter set and the warning threshold are updated to form a warning parameter update list; based on the warning parameter update list, a defect identification report is generated.

[0101] The embodiment solves the technical problems of low underwater defect identification accuracy, poor positioning accuracy and difficult reliability evaluation through multi-level feature analysis and comprehensive evaluation. In the feature discrimination link, the fusion feature matrix is compared with the pre-stored structure similarity, time correlation and spatial position correlation data in multiple dimensions, and the robustness of defect discrimination is improved through adaptive adjustment of feature weights; in the defect classification process, multi-feature library matching and multi-level fuzzy evaluation method are adopted, which can not only accurately identify the defect type, but also give reasonable grade division; in the spatial positioning aspect, the positioning accuracy is improved through weighted fusion and error compensation of multi-source positioning data, and a reliable positioning confidence interval can be given; in the reliability evaluation link, a complete evaluation system is established by analyzing multi-dimensional indexes such as identification accuracy, spatial accuracy, time stability and spatial consistency, which provides a quantitative basis for the credibility of the detection result; at the same time, the warning parameters are dynamically updated based on the evaluation result, and the adaptive optimization of the detection system is realized. The embodiment realizes accurate identification and reliable positioning of underwater defects, and establishes a scientific reliability evaluation mechanism.

[0102] According to one aspect of the present application, step S42 further comprises:

[0103] Step S421, read the edge feature data from the defect discrimination result, extract the crack area by using the adaptive threshold segmentation algorithm based on region growing, and generate crack contour data. Input the crack contour data into the improved Hough transform model, calculate the main direction and branch direction of the crack, and generate crack trend data. Apply sub-pixel edge positioning algorithm to the crack contour data, calculate the accurate coordinates of each contour point, and generate contour accuracy data. Extract the depth feature from the fusion feature matrix output from S3, calculate the crack width and length in combination with the contour accuracy data, and output the crack size data.

[0104] Wherein the improved Hough transform is A(ρ, θ) = Σ w(i) Δ(ρ-xi cos(θ)-yi sin(θ)); wherein w(i) = exp(-|gradI(i)| / σ1) exp(-|κ(i)| / σ2); gradI(i) is the image gradient intensity; κ(i) is the local curvature; σ1, σ2 are scale parameters; xi, yi are edge point coordinates.

[0105] Step S422, a region growing algorithm is applied to the fusion feature matrix to divide the damage area, a boundary curve of each damage area is extracted, a modified area calculation model is used to obtain damage area data, and damage depth data is calculated based on image gray gradient analysis.

[0106] The region growing algorithm is specifically: T(x, y) = μ(x, y) + α σ(x, y) [1 + β g(x, y)]; wherein T(x, y) is a local adaptive threshold; μ(x, y) is a local area mean; σ(x, y) is a local area standard deviation; g(x, y) is a local gradient intensity; α is a mean adjustment coefficient (1.02.0); β is a gradient influence factor (0.20.8). If |I(x, y)-I(s, t)|< T(x, y) and C(x, y) > Tc, then R(x, y) = 1; otherwise R(x, y) = 0; wherein R(x, y) is a region growing result; I(x, y) is a target pixel value; I(s, t) is a seed point pixel value; C(x, y) is a connectivity constraint; Tc is a connectivity threshold.

[0107] The modified area calculation model is specifically: A(R) = Σ[w(i, j) a(i, j)] [1 + α K(i, j)]; wherein A(R) is a region area; w(i, j) is a pixel weight; a(i, j) is a pixel area; K(i, j) is a curvature correction term; α is a curvature influence coefficient (0.1~0.4). w(i, j) = exp(-|gradI(i, j)| / σ1) exp(-|θ(i, j)| / σ2); wherein gradI(i, j) is a gradient intensity; θ(i, j) is a direction difference; σ1, σ2 are adjustment parameters.

[0108] Step S423, in combination with the crack width data, the crack length data and the damage depth data, a defect three-dimensional reconstruction model is established, a defect volume is calculated to obtain defect volume data, and a defect development prediction is generated by analyzing a defect extension trend.

[0109] The process of establishing the three-dimensional reconstruction model of the defect can be: performing three-dimensional defect reconstruction calculation: V(x, y, z) = Vb(x, y, z) [1 + a H(x, y, z)] [1 + b S(x, y, z)]; wherein V(x, y, z) is a reconstructed voxel value; Vb(x, y, z) is a basic voxel value; H(x, y, z) is a depth feature; S(x, y, z) is a surface feature; a, b are feature weights (0.2-0.4). R(x, y, z) = exp(-|gradV(x, y, z)| / s1) exp(-|K(x, y, z)| / s2); wherein R(x, y, z) is a reconstruction reliability; grad V is a voxel gradient; K(x, y, z) is a local curvature; s1, s2 are control parameters.

[0110] Step S424, matching the defect volume data and the damage area data with the pre-stored feature template library, identifying the defect type by using a multi-feature fusion classification method, and outputting the defect type and the type confidence.

[0111] Step S425, based on the defect type, the defect volume data and the pre-stored rating standard, calculating a geometric feature damage index and a structure influence index, and determining the defect grade and the grade confidence by using a multi-level fuzzy evaluation method.

[0112] The embodiment solves the technical problems of low accuracy of underwater defect type identification and unreasonable grade division by multi-feature library matching and multi-level fuzzy evaluation. The adaptive threshold segmentation algorithm is used to extract the crack contour, the improved Hough transform is used to realize accurate calculation of the crack direction, and the accuracy of feature extraction is improved. The sub-pixel edge positioning technology and depth information fusion are used to realize high-precision measurement of crack width and length, and provide reliable quantitative basis for defect evaluation. The region growing algorithm is used to accurately divide the damage area, and the improved area calculation model and gray gradient analysis are used to realize accurate measurement of damage area and depth. Based on the three-dimensional reconstruction technology of the defect, not only the defect volume is calculated, but also the extension trend of the defect is analyzed, which provides a scientific basis for predicting the development of the defect. The multi-feature fusion classification method is used for defect identification, and the multi-level fuzzy evaluation is used to determine the defect grade, which not only ensures the accuracy of classification, but also realizes the rationality of grade division. The embodiment improves the accuracy and reliability of defect identification.

[0113] According to one aspect of the present application, step S43 further comprises:

[0114] Step S431, real-time sonar positioning data is received from the sonar device, optical positioning data is received from the optical system, and structural monitoring data at the corresponding time is read from the monitoring system database. The three types of data are input into the time synchronization model, aligned according to the respective timestamps, and synchronized positioning data is generated. For each data source in the synchronized positioning data, a reliability weight is calculated according to its acquisition conditions and historical stability, and a positioning weight matrix is generated. The synchronized positioning data is weighted and fused according to the positioning weight matrix, and an initial fusion position is output.

[0115] Step S432, system errors and random errors in the synchronized positioning data are analyzed, an error compensation model is established, the initial fusion position is corrected using an adaptive Kalman filtering algorithm, and a compensated position is output.

[0116] Step S433, the compensated position is converted from the device coordinate system to the dam structure coordinate system, considering the attitude error in the coordinate transformation process, the conversion accuracy is improved using a multi-source data constraint method, and structure coordinate data is generated.

[0117] Step S434, a positioning accuracy evaluation index is calculated for the structure coordinate data, a three-dimensional error ellipsoid model is established, the positioning error distribution in each direction is analyzed, and the positioning result and the corresponding confidence interval are output.

[0118] The embodiment solves the technical problems of low underwater defect positioning accuracy and poor reliability through multi-source positioning data fusion and error compensation. The timestamp alignment method is used to synchronize the multi-source positioning data, and the reliability weight calculation ensures the accuracy of data fusion. The analysis model of system errors and random errors is designed, and the adaptive Kalman filtering algorithm is combined to dynamically correct the fusion position, improving the positioning accuracy. In the coordinate conversion process, the influence of attitude error is considered, and the accuracy of coordinate transformation is improved through multi-source data constraint. A three-dimensional error ellipsoid model is established to analyze the positioning error distribution, not only the positioning result is given, but also a reliable confidence interval is provided. The embodiment not only improves the accuracy of defect spatial positioning, but also quantifies the uncertainty of the positioning result, providing accurate spatial guidance for defect repair.

[0119] According to one aspect of the present application, step S44 further comprises:

[0120] Step S441, key indicators such as recognition accuracy, feature completeness, and boundary clarity are extracted from the defect discrimination result to generate a recognition quality vector. Key indicators such as spatial accuracy, repeatability, and stability are extracted from the positioning result to generate a positioning quality vector. The recognition quality vector and the positioning quality vector are normalized according to the preset evaluation rules, and multi-dimensional evaluation indicators are output.

[0121] Step S442, read the historical evaluation indicators in the historical database, time sequence comparison between the current multi-dimensional evaluation indicators and the historical evaluation indicators, calculate the short-term fluctuation indicators by using the sliding window method, and generate the fluctuation feature data. Apply the trend analysis algorithm to the fluctuation feature data, calculate the change rate and acceleration of each indicator, and generate the change feature data. Input the fluctuation feature data and the change feature data into the time stability evaluation model, and output the time stability data.

[0122] Wherein the sliding window method is specifically: F(t) = Σ[w(k) x(t-k)] [1 + a D(t)]; Wherein F(t) is the filter output; w(k) is the window weight; x(t) is the input sequence; D(t) is the dynamic adjustment term; a is the adaptive coefficient (0.1~0.3). w(k) = exp(-k 2 / 2σ 2 ) [1 + β E(k)]; Wherein k is the time delay; σ is the window width; E(k) is the edge retention factor; β is the edge weight (0.2~0.5).

[0123] The trend analysis algorithm is specifically: P(t) = a exp(-λ t) sin(ωt + φ) + b t + c; Wherein P(t) is the prediction trend value; t is the time variable; a is the amplitude coefficient; λ is the attenuation coefficient; ω is the period factor; φ is the phase angle; b is the linear trend coefficient; c is the bias constant. R(t) = |P(t)-M(t)| / σ(t) [1 + a V(t)]; Wherein R(t) is the reliability indicator; M(t) is the measurement value; σ(t) is the standard deviation; V(t) is the variation coefficient; a is the adjustment factor (0.1~0.5).

[0124] Step S443, read the spatial correlation template from the database, and perform spatial mapping on the current multi-dimensional evaluation indicators according to the spatial correlation template to generate spatial distribution data. Calculate the correlation coefficient and consistency indicator of the indicator values of adjacent regions in the spatial distribution data to generate regional correlation data. Input the regional correlation data into the spatial consistency evaluation model, combine the preset evaluation standard, and output the spatial consistency data.

[0125] Step S444, input the multi-dimensional evaluation indicators, the time stability data and the spatial consistency data into the multi-level fuzzy evaluation model, calculate the fuzzy membership degree according to the importance of each layer indicator, and generate the fuzzy evaluation matrix. Apply the comprehensive calculation method to the fuzzy evaluation matrix to obtain the final score of each evaluation indicator, and generate the comprehensive score data. Compare the comprehensive score data with the preset reliability level standard, and output the reliability evaluation result.

[0126] Wherein the application of comprehensive calculation method to fuzzy evaluation matrix is specifically: M(i, j) = μ(i, j) [1 + α S(i, j)] [1 + β T(i, j)]; wherein M(i, j) is fuzzy evaluation value; μ(i, j) is basic membership degree; S(i, j) is spatial correlation; T(i, j) is time correlation; α, β are correlation weights (0~0.5). R = Σ [wi M(i)] exp(-γ V(i)); wherein R is comprehensive evaluation result; wi is index weight; M(i) is each index evaluation value; V(i) is evaluation volatility; γ is stability coefficient (0.2~0.6).

[0127] The process of outputting reliability evaluation result is specifically: R(i) = Rb(i) [1 + α T(i)] [1 + β S(i)] [1+ γ V(i)]; wherein R(i) is comprehensive reliability score; Rb(i) is basic reliability; T(i) is time stability; S(i) is spatial consistency; V(i) is fluctuation degree; α, β, γ are weight coefficients (0~0.5). C(i) = Σ [wj R(i, j)] exp(-λ D(i)); wherein C(i) is confidence degree; wj is index weight; R(i, j) is sub-item reliability; D(i) is data deviation; λ is deviation influence factor (0.2~0.4).

[0128] The embodiment solves the technical problems of difficult detection result reliability evaluation and poor stability by multi-dimensional index evaluation and space-time characteristic analysis. A multi-dimensional evaluation system including recognition accuracy, feature integrity, boundary definition, etc. is constructed, scientific quality evaluation is realized through index normalization processing; a time sequence analysis method based on sliding window is designed, short-term volatility and long-term stability are calculated to realize dynamic evaluation of the index; spatial correlation analysis technology is used to evaluate the spatial distribution characteristics of the index, and the reliability of the evaluation is improved by calculating the regional consistency; based on the multi-level fuzzy comprehensive evaluation model, the importance of each index is considered, and scientific comprehensive evaluation is realized. The embodiment not only provides a quantitative standard for the credibility of the detection result, but also establishes a complete quality evaluation system.

[0129] According to one aspect of the application, step S45 further comprises:

[0130] Step S451, extracting the defect development trend in the historical monitoring data, establishing a time sequence prediction model, analyzing the change law combined with the reliability evaluation result, and generating trend prediction data.

[0131] Step S452, based on the trend prediction data and the current defect type and defect level, calculating the current value and predicted value of each early warning index, evaluating the development risk, and outputting early warning index data.

[0132] The early warning index is also dynamically updated, specifically: P(t) = Pb(t) [1 + a H(t)] [1 + b F(t)]; wherein P(t) is the early warning index value; Pb(t) is the baseline index value; H(t) is the historical trend term; F(t) is the future prediction term; a, b are time weights (0.2~0.4). U(t) = exp(-|dP(t) / dt| / s1) exp(-|d 2 P(t) / dt 2 | / s2); wherein U(t) is the update coefficient; dP(t) / dt is the change rate; d 2 P(t) / dt 2 is the acceleration; s1, s2 are control parameters.

[0133] Step S453, according to the early warning index data, an adaptive threshold update algorithm is used to dynamically adjust the threshold values of each early warning index, and an updated threshold list is generated.

[0134] The adaptive threshold update algorithm is specifically: T(k+1) = T(k) + η [μ(k)-T(k)] + b [(s(k) / s(k-1))-1]; wherein T(k+1) is the updated threshold value; T(k) is the current threshold value; μ(k) is the current mean value; s(k) is the current standard deviation; η is the learning rate (0.1~0.3); b is the volatility adjustment coefficient (0.2~0.4). W(k) = exp(-|T(k)-T(k-1)| / D) exp(-|s(k)-s(k-1)| / g); wherein W(k) is the update weight; D, g are stability parameters; T(k-1) is the threshold value of the last period; s(k-1) is the standard deviation of the last period.

[0135] Step S454, based on the updated threshold list, the sampling frequency, measurement accuracy and other parameters of the key monitoring points are optimized, and a monitoring parameter optimization table is generated.

[0136] Step S455, the early warning index data and the early warning threshold are comprehensively determined to determine the early warning level, update the monitoring control parameters, and generate an early warning parameter update list and a defect identification report.

[0137] The embodiment solves the technical problems of non-timely updating of early warning parameters and poor optimization effect through trend prediction and parameter adaptive optimization. A time series prediction model based on historical data is constructed, the change law is analyzed in combination with reliability evaluation results, and the accuracy of trend prediction is improved. An adaptive threshold updating algorithm is designed, the threshold is dynamically adjusted according to the change of the early warning index, and the sensitivity and reliability of early warning are improved. By optimizing the sampling parameters of the key monitoring points, the monitoring density of the key areas is ensured, and the system operation efficiency is improved. A dynamic evaluation mechanism of early warning levels is established, and the adaptive optimization of the system is realized by updating the monitoring control parameters. The embodiment not only improves the accuracy and real-time performance of the early warning system, but also realizes the continuous optimization of the detection system.

[0138] According to one aspect of the present application, an underwater multi-modal identification system for hidden micro-defects of high dams comprises:

[0139] at least one processor; and

[0140] a memory communicatively connected with the at least one processor; wherein

[0141] The memory stores instructions executable by the processor, and the instructions are executed by the processor to implement the multi-modal identification method for hidden micro-defects of high dams described in any of the above embodiments.

[0142] The present application generates a sampling feature vector by detecting the sampling frequency of displacement, stress and temperature data; then designs linear interpolation parameters according to the sampling feature vector to realize data resampling and time alignment; and finally generates a standardized multi-source monitoring data matrix through the organization of spatial position and sensor type. By dividing core data and peripheral data and calculating adaptive weight coefficients, the problem of uneven data quality is effectively solved. This processing mechanism ensures the time synchronization and quality consistency of data from different sources, and realizes synchronous data processing.

[0143] The sound velocity profile is calculated by obtaining water depth, temperature and salinity data from the environmental database; then the attenuation coefficients at different depths are calculated based on the sound velocity profile, and the transmission power and receiving gain are optimized in combination with the target distance and water turbidity; finally, the influence of platform motion and sound velocity variation is eliminated through attitude compensation and sound path correction. The gain is adjusted by using overlapping sliding window and piecewise linear mapping, and the frequency band processing is realized by using multi-layer filter bank, which effectively improves the quality and reliability of the sonar signal and constructs a complete environmental adaptation mechanism.

[0144] For crack defects, the self-adaptive threshold segmentation and improved Hough transform are used to accurately extract the strike characteristics; for damaged areas, the area and depth are calculated through the region growing algorithm and gray gradient analysis; through the establishment of a three-dimensional reconstruction model of defects, not only the defect volume is calculated, but also the extension trend is analyzed. Finally, a multi-feature fusion classification method is adopted to realize the accurate identification of different types of defects, and a multi-level defect feature extraction system is constructed.

[0145] By extracting the timestamp and spatial position mark of acoustic data, the acquisition time and camera pose information of image data are extracted; then an improved space-time registration algorithm is used to generate a registration mapping matrix; finally, through the calculation of the quality indicators of acoustic and image data, a multi-index fusion evaluation method is used to ensure the reliability of the registration result, and an accurate space-time registration mechanism is constructed.

[0146] By extracting multi-dimensional evaluation indicators from defect discrimination results and positioning results; then analyzing the time stability and spatial consistency of the indicators; finally, through a multi-level fuzzy comprehensive evaluation model, the comprehensive reliability score is calculated. Based on the reliability evaluation results, the early warning parameters are dynamically updated, the continuous optimization of the detection system is realized, the long-term reliability of the detection results is guaranteed, and a complete reliability evaluation system is established.

[0147] The present application solves the key technical problems of data heterogeneity, environmental interference, feature extraction difficulty and low recognition accuracy in dam underwater defect detection by constructing a complete technical chain of "monitoring and early warning-accurate detection-feature extraction-defect recognition". Based on the abnormal early warning mechanism of multi-source monitoring data, through time synchronization processing and adaptive weight calculation, the suspicious area is quickly and accurately positioned; an underwater detection method of sound and light cooperation is proposed, through multi-parameter optimization and adaptive control, the quality and reliability of the detection data are improved; a multi-modal feature extraction and fusion algorithm is designed, which fully utilizes the complementary advantages of acoustic and optical data, enhances the integrity and accuracy of feature expression; a multi-level defect recognition and evaluation system is established, through feature matching, spatial positioning and reliability evaluation, the accurate identification and quantitative evaluation of defects are realized. The present application has strong practicability and adaptability, and can realize the automatic and intelligent detection of dam defects in complex underwater environment, providing reliable technical support for dam safety monitoring.

[0148] It should be noted that various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, various possible combinations are not described again by the present application.

Claims

1. An underwater multi-modal identification method for hidden micro-defects of high dams, characterized in that, The method comprises the following steps: Step S1, real-time acquisition of displacement monitoring data, stress monitoring data and temperature monitoring data from a sensor network database, and generation of a multi-source monitoring data matrix through time synchronization processing; extraction of abnormal features from the multi-source monitoring data matrix using an adaptive weight calculation method to generate an abnormal feature matrix; input of the abnormal feature matrix into a pre-configured multi-dimensional analysis model to calculate a suspicious region coordinate set and simultaneously generate a credibility score; priority sorting of the suspicious region coordinate set based on the credibility score, and output of a suspicious region list; Step S2, acquisition of the suspicious region list and planning of a scanning path according to the suspicious region list; Based on the scanning path, original sonar data is acquired, and enhanced sonar data is obtained through multi-level signal enhancement processing; Based on the enhanced sonar data, edge feature data is extracted, multi-feature analysis is performed, and an accurate defect region and defect feature description are determined; Step S3, based on the accurate defect region and defect feature description, acoustic data and optical data are collected, and registration data is obtained through space-time alignment; enhanced feature data is obtained through enhancement processing of the registration data; Based on the enhanced feature data, multi-modal feature data is extracted and fused; Based on the multi-modal feature data, a fusion feature matrix and a feature credibility index are generated; Step S4, based on the fusion feature matrix and the feature credibility index, a defect discrimination result is obtained through feature discrimination; Based on the defect discrimination result, a multi-level analysis is performed to determine the defect type and defect level; Based on the defect type and defect level, an accurate spatial position is calculated to obtain a positioning result; Based on the positioning result, a comprehensive evaluation is performed to obtain a reliability evaluation result, and a defect identification report and a pre-warning parameter update list are generated.

2. The method according to claim 1, wherein, Step S1 is further divided into: Step S11, reading of displacement monitoring data corresponding to a three-dimensional space position from a displacement sensor collection node, reading of stress monitoring data at the same position from a stress sensor collection node, and reading of temperature monitoring data at the same position from a temperature sensor collection node; sampling frequency detection of the displacement monitoring data, the stress monitoring data and the temperature monitoring data to generate a sampling feature vector; based on the sampling feature vector, a linear interpolation parameter is constructed to resample and time-align the displacement monitoring data, the stress monitoring data and the temperature monitoring data to generate time-aligned data; the time-aligned data is organized according to the space position and the sensor type to output a multi-source monitoring data matrix; Step S12, division of the multi-source monitoring data matrix into a core data part representing the main trend and a peripheral data part representing fluctuation changes, calculation of adaptive weight coefficients of the core data part and the peripheral data part; addition of the core data part and the peripheral data part after being multiplied by the corresponding adaptive weight coefficients to obtain weighted monitoring data; based on the weighted monitoring data, the distance between each monitoring point and its adjacent monitoring point is calculated; the distance is converted into a correlation weight value; the correlation weight value is multiplied by the weighted monitoring data to obtain an abnormal feature matrix; Step S13, for each monitoring point in the abnormal feature matrix, calculate its local correlation value with the surrounding monitoring points and the global correlation value with all monitoring points; Based on the weighted combination of the local correlation value and the global correlation value, construct an enhanced feature matrix; Step S14, based on the enhanced feature matrix, calculate the dynamic threshold value of each monitoring point; Record the monitoring points that exceed the corresponding dynamic threshold value to form a suspicious region coordinate set; Step S15, for each region in the suspicious region coordinate set, respectively calculate its displacement anomaly degree, stress anomaly degree and temperature anomaly degree; combine the displacement anomaly degree, stress anomaly degree and temperature anomaly degree through adaptive weight combination to obtain the credibility score; based on the credibility score, sort the suspicious region coordinate set to generate a suspicious region list with priority.

3. The method for underwater multi-modal identification of micro-defects of high dams according to claim 2, characterized in that, Step S2 is further: Step S21, based on the priority and spatial coordinates of each suspicious region in the suspicious region list, calculate the horizontal scanning angle and the vertical scanning angle to generate a sonar scanning path; control the sonar device to collect the echo signal of the target region according to the sonar scanning path to obtain the original sonar data containing the distance value and the signal intensity; Step S22, based on the original sonar data, calculate the signal mean value and variance of the local region; based on the mean value, determine the signal gain coefficient; based on the variance, determine the noise suppression coefficient; multiply the signal gain coefficient with the original sonar data to obtain the gain sonar data; multiply the noise suppression coefficient with the gain sonar data to obtain the enhanced sonar data; Step S23, based on the enhanced sonar data, calculate the gradient value in the horizontal direction and the vertical direction; based on the gradient value, calculate the edge strength and direction angle, and combine the statistical characteristics of the local region to determine the adaptive threshold; record the edge points higher than the adaptive threshold as edge feature data; Step S24, based on the edge feature data, calculate the intensity feature, shape feature and texture feature of each edge point; match the intensity feature, shape feature and texture feature with the preset defect feature template to obtain the feature matching probability; Mark the region with a feature matching probability higher than a preset threshold as an accurate defect region; Step S25, based on the accurate defect region, respectively calculate the geometric size, edge sharpness, echo intensity and spatial continuity of the region; combine the geometric size, edge sharpness, echo intensity and spatial continuity to form a defect feature description, and calculate the reliability index of the feature to generate a defect detection report containing the defect position, feature and reliability.

4. The method according to claim 3, wherein, Step S3 is further: Step S31, based on the accurate defect region and the defect feature description, control the acoustic sensor to collect acoustic original data of the target region at different angles, and control the underwater camera to collect optical original data at the corresponding position at the same time, record the acquisition time and spatial coordinates of each data; based on the acquisition time and spatial coordinates, align the acoustic original data and the optical original data to generate registration data; Step S32, read the acoustic part in the registration data, calculate the high frequency component and the low frequency component; Based on the high frequency component and the low frequency component, respectively enhance and then weightedly combine to obtain enhanced acoustic data; Read the optical part in the registration data, respectively, enhance the contrast and texture features, and then weighted combination to obtain enhanced optical data; Based on the preset quality evaluation index, calculate the complementary enhancement coefficient, combine the enhanced acoustic data and the enhanced optical data to obtain the enhanced feature data; Step S33, based on the enhanced feature data, extract the frequency spectrum feature, the time domain feature and the energy feature from the enhanced acoustic data to generate the acoustic feature data; extract the edge feature, the texture feature and the shape feature from the enhanced optical data to generate the optical feature data; calculate the similarity between the acoustic feature data and the optical feature data to generate the feature mutual verification matrix; Step S34, based on the acoustic feature data and the optical feature data, generate the quality evaluation result, calculate the feature weight to obtain the weighted feature; Fuse the weighted feature to obtain the multi-modal feature data; Based on the multi-modal feature data, filter according to the preset feature importance to generate the fusion feature matrix; Step S35, calculate the geometric size, physical property, dynamic response and spatial distribution of each feature in the fusion feature matrix to generate the feature description vector; Based on the feature description vector, calculate the credibility of the acoustic feature data, the optical feature data and the feature mutual verification matrix respectively, and then weighted combination to obtain the feature credibility index.

5. The method for underwater multi-modal identification of micro-defects of high dams according to claim 4, characterized in that, Step S4 is further: Step S41, based on the feature credibility index, compare and calculate the fusion feature matrix with the pre-stored structure similarity data, time sequence correlation data and spatial position correlation data to obtain the feature discrimination matrix; Based on the feature weight in step S34, weighted calculation is performed on the feature discrimination matrix, and the defect discrimination result is output; Step S42, based on the defect discrimination result, calculate the matching degree with the preset crack feature library, the erosion feature library, the leakage feature library and the deformation feature library respectively to obtain the feature matching data; Based on the feature matching data and the defect discrimination result, determine the defect type and divide the defect level; Step S43, based on the defect type and the defect level, combine the pre-stored sonar positioning data, optical positioning data and monitoring positioning data to calculate the spatial positioning data; Based on the spatial positioning data, perform positioning error analysis to generate error evaluation data; based on the error evaluation data, determine the final positioning result and the positioning confidence interval; Step S44, calculate the recognition accuracy index of the defect discrimination result, the spatial precision index of the positioning result, the feature time stability index and the spatial consistency index to generate the evaluation index data; Based on the evaluation index data, obtain the comprehensive reliability score through weighted combination to form the reliability evaluation result; Step S45, based on the reliability evaluation result, the defect type, the defect level and the pre-stored historical evolution trend data, calculate the early warning index value to generate the early warning level data; Based on the early warning level data, update the key monitoring point list, the monitoring parameter set and the early warning threshold to form the early warning parameter update list; Based on the early warning parameter update list, generate the defect identification report.

6. The method for underwater multi-modal identification of micro-defects of high dams according to claim 5, characterized in that, Step S21 is further: Step S211, read the spatial coordinate information of the suspicious area list, based on the spatial coordinate information, divide the detection area into a predetermined sub-area using an adaptive grid decomposition algorithm, and generate grid division data; read the current water flow data and obstacle data from the pre-stored environmental database, combine the grid division data to calculate the detection difficulty coefficient of each sub-area, and output the difficulty coefficient matrix; Combine the priority information in the suspicious area list with the difficulty coefficient matrix by weighting, and generate an area priority matrix; Based on the area priority matrix, an improved ant colony algorithm is used for optimization, and an initial scanning path is output; Step S212, based on the initial scanning path and the pre-stored water depth data, generate the sound velocity profile data; Based on the sound velocity profile data, calculate the attenuation coefficient of the sonar beam at different depths, and generate an attenuation coefficient table; Based on the attenuation coefficient table, calculate the optimal transmission power and reception gain from the target area distance data; based on the optimal transmission power and reception gain, combine the pre-stored turbidity data for parameter optimization, and generate a parameter configuration table; Step S213, based on the parameter configuration table and the area priority matrix, calculate the optimal pitch angle and horizontal turning angle of each detection position, and generate a multi-angle scanning sequence; control the sonar rotating mechanism to move according to the multi-angle scanning sequence, and real-time collect echo signal data; Step S214, obtain the attitude data from the inertial navigation system, based on the attitude data, perform attitude compensation on the echo signal data to generate compensated signal data; extract the sound path correction parameter from the sound velocity profile data, and perform sound path correction on the compensated signal data to generate corrected signal data; time-align and space-splice all the corrected signal data, and output the original sonar data.

7. The method for underwater multi-modal identification of micro-defects of high dams according to claim 5, characterized in that, Step S22 is further provided: Step S221, divide the original sonar data into overlapping local analysis windows, calculate the signal mean, variance and peak value distribution in each local analysis window, and generate a statistical feature matrix; Based on the statistical feature matrix, an adaptive window algorithm is used to optimize the window size, and optimized statistical data is output; Step S222, based on the optimized statistical data, calculate the signal dynamic range of each window, and generate a dynamic range vector; based on the dynamic range vector, construct a piecewise linear mapping function, and generate a gain mapping table; extract the gain control parameter of each signal segment from the gain mapping table, and perform segmented gain adjustment on the original sonar data according to the gain control parameter, and output the gain adjusted data; Step S223, perform time-frequency analysis on the gain adjusted data, extract the frequency characteristics and energy distribution of the noise; based on the frequency characteristics and energy distribution, construct a multi-layer noise model, identify system noise, environmental noise and interference noise, and generate a noise feature vector and a noise suppression coefficient; Step S224, based on the noise feature vector and the noise suppression coefficient, construct an adaptive filter bank to process the signal in different frequency bands, and obtain processed data; based on the processed data, use the wavelet threshold method to eliminate noise, and reconstruct the signal to obtain enhanced sonar data. Step S31 is further provided:

8. The method for underwater multi-modal identification of micro-defects of high dams according to claim 5, characterized in that, ​ Step S311, read the spatial coordinates of the precise defect area, divide the detection space into working sub-regions using an adaptive grid method; based on the working sub-regions, pre-stored water flow velocity and obstacle distribution, calculate the difficulty coefficient of passing; based on the difficulty coefficient of passing, use a dynamic obstacle avoidance path planning algorithm to generate an initial path of the underwater vehicle; Based on the initial path of the underwater vehicle, use the pre-configured energy consumption model to optimize and obtain the operation path of the underwater vehicle; Step S312, based on the operation path of the underwater vehicle and the defect feature description, calculate the best working distance of the acoustic sensor and the low-light camera, and generate a device arrangement scheme; based on the device arrangement scheme, use a multi-device cooperative control algorithm to calculate the working time sequence of each sensor, and output a cooperative acquisition instruction; Step S313, read the real-time water quality parameters and environmental light intensity, dynamically adjust the exposure time, gain value and color enhancement parameters of the low-light camera, and generate a camera parameter table; Based on the camera parameter table, obtain the reflection characteristics of the target area, calculate the best light supplement angle and intensity, and output a light supplement control instruction; Step S314, based on the cooperative acquisition instruction and the light supplement control instruction, receive the original acoustic data and original image data collected by each sensor, extract the timestamp and spatial position marker of the acoustic data, and extract the acquisition time and camera pose information of the image data; Based on the timestamp, spatial position marker, acquisition time and camera pose information, use a space-time registration algorithm to generate a space-time correspondence table; Step S315, based on the space-time correspondence table, calculate the signal-to-noise ratio and clarity of the acoustic data to obtain an acoustic quality index; based on the space-time correspondence table, calculate the clarity and contrast of the image data to obtain an image quality index; based on the acoustic quality index and the image quality index, use a multi-index fusion evaluation method to generate registration data and a data quality report.

9. The method for underwater multi-modal identification of micro-defects of high dams according to claim 5, characterized in that, Step S32 further comprises: Step S321, multi-scale wavelet decomposition is performed on the acoustic signals in the registration data to extract feature components of different frequency bands, the energy distribution of each feature component is calculated, and a frequency domain feature matrix is generated; Based on the frequency domain feature matrix, analyze the frequency band correlation to obtain frequency domain correlation data; wherein the feature components of different frequency bands include high frequency components, medium frequency components and low frequency components; Step S322, based on the frequency domain feature matrix and the frequency domain correlation data, the high frequency components are directionally enhanced, the medium frequency components are selectively maintained, and the low frequency components are noise suppressed, and the processed frequency band components are fused to generate enhanced acoustic data; Step S323, geometric correction and illumination equalization are performed on the optical image in the registration data, an adaptive histogram model is constructed, and the non-uniform illumination effect is compensated to generate corrected image data; Step S324, a multi-scale contrast enhancement algorithm is applied to the corrected image data, the enhancement parameters are adaptively adjusted combined with the statistical characteristics of the local area, the detail and texture performance are optimized, and enhanced optical data are output; Step S325, complementary features of the enhanced acoustic data and the enhanced optical data are calculated, and a feature fusion weight is constructed; Based on the feature fusion weight, the enhanced acoustic data and the enhanced optical data are combined using an adaptive weighting method to generate enhanced feature data.

10. An underwater multi-modal identification system for micro-defects of high dams, characterized in that, It comprises: at least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the processor for execution by the processor to implement the method of any one of claims 1 to 9.

Images