A method for detecting corrosion targets of a space truss structure

Abstract

This application relates to a method for detecting corrosion targets in a space frame structure. The method acquires high-resolution image data, point cloud data, and thickness data of the surface of the space frame members to be tested, as well as the thickness data of the interior of the members. Channel features are extracted from the high-resolution image data for corrosion analysis to obtain a first corrosion detection result. Point cloud features are extracted from the preprocessed point cloud data to identify a second corrosion detection result. The difference between the preprocessed thickness data and normal thickness data is compared to obtain a third corrosion detection result. A multimodal data fusion strategy based on Bayesian inference fuses the channel features, point cloud features, and conditional probability distribution of the thickness data to obtain the corrosion probability of each detection area, thereby determining the corrosion area. The three corrosion detection results for the corrosion area are combined to obtain a comprehensive corrosion assessment report for the corrosion area. Finally, the comprehensive corrosion assessment report is combined with a future corrosion risk prediction based on an LSTM model to obtain a final corrosion detection report.

Description

Technical Field

[0001] This application relates to the field of corrosion detection technology for space frame structures, and in particular to a method for detecting corrosion targets in space frame structures. Background Technology

[0002] Currently, existing methods for detecting corrosion in space frame members have the following problems: 1. Traditional corrosion detection of space frame members mainly relies on single data sources such as image data or ultrasonic data, which makes it impossible to fully reflect the complex characteristics of corrosion (such as the joint assessment of surface corrosion and internal corrosion) and to efficiently and accurately identify the full picture of corrosion, especially corrosion in deep or hidden areas.

[0003] 2. Existing image processing methods often perform poorly against complex backgrounds and are easily affected by factors such as lighting, shadows, and stains, leading to inaccurate identification of rusted areas. Furthermore, traditional algorithms have low accuracy in feature extraction and classification, making it difficult to achieve efficient detection of minute rust.

[0004] 3. Current corrosion detection technologies mainly focus on single-modal or single-feature analysis in assessing corrosion, and cannot effectively combine data from different sensors (such as images, point clouds, ultrasonic thickness data, etc.) to comprehensively assess the morphology, depth, and impact on structural safety of corrosion.

[0005] 4. Existing detection methods mainly focus on static detection results, lacking dynamic prediction and assessment of rust development trends. They cannot detect potential corrosion risks in a timely manner, nor can they formulate effective maintenance and repair strategies based on rust development trends.

[0006] 5. Traditional corrosion detection methods are easily affected by external factors (such as environment, equipment performance, operators, etc.), resulting in low accuracy and reliability of the detection results, which in turn affects structural safety assessment and decision-making. Summary of the Invention

[0007] Therefore, it is necessary to provide a method for detecting corrosion targets in a space frame structure, the method comprising: S1: Acquire high-definition image data, point cloud data, and thickness data of the surface of the space frame member to be tested, and preprocess the point cloud data and the thickness data respectively; S2: A convolutional neural network model based on a deep residual network is used to extract texture features, color features, and edge features from high-definition image data. Corrosion analysis is performed based on these three channel features to obtain the first corrosion detection result, which includes the location, shape, and area of ​​corrosion. The preprocessed point cloud data was processed using a random sampling consistency plane fitting algorithm to extract point cloud features from the surface of the rod. Based on the point cloud features, local curvature changes were analyzed to identify the second corrosion detection results, including depressions and pits on the surface of the rod. By comparing the difference between the pre-processed thickness data and the normal thickness data, a third corrosion detection result including the degree of internal corrosion of the rod is obtained; S3: A multimodal data fusion strategy based on Bayesian inference is adopted to fuse the channel features, the point cloud features and the conditional probability distribution of the thickness data to obtain the corrosion probability of each detection area, and the corrosion area is determined based on the corrosion probability. The first corrosion detection result, the second corrosion detection result and the third corrosion detection result of the corrosion area are combined to obtain a comprehensive corrosion assessment report of the corrosion area. S4: Use an LSTM model to perform time-series analysis on channel features, point cloud features, and thickness data to predict future corrosion trends, and then conduct a corrosion risk assessment to obtain a future corrosion risk prediction; combine the comprehensive corrosion assessment report and the future corrosion risk prediction to obtain the final corrosion detection report.

[0008] Beneficial effects: This method acquires high-resolution image data, point cloud data, and thickness data of the surface of the space frame members to be tested, as well as the thickness data of the interior of the space frame members; extracts channel features from the high-resolution image data for corrosion analysis to obtain the first corrosion detection result; extracts point cloud features based on the preprocessed point cloud data to identify the second corrosion detection result; compares the difference between the preprocessed thickness data and the normal thickness data to obtain the third corrosion detection result; fuses the channel features, point cloud features, and conditional probability distribution of the thickness data based on a Bayesian inference multimodal data fusion strategy to obtain the corrosion probability of each detection area, thereby determining the corrosion area; and combines the three corrosion detection results of the corrosion area to obtain a comprehensive corrosion assessment report of the corrosion area; and combines the comprehensive corrosion assessment report with the future corrosion risk prediction based on LSTM model analysis to obtain the final corrosion detection report. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of the method for detecting corrosion targets in a space frame structure in this application embodiment. Detailed Implementation

[0011] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0012] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0013] like Figure 1 As shown in the figure, this embodiment provides a method for detecting corrosion targets in a space frame structure. The method includes: S1: Acquire high-definition image data, point cloud data, and thickness data of the surface of the space frame member to be tested, and preprocess the point cloud data and the thickness data respectively.

[0014] In this embodiment, high-definition image data is acquired using a high-definition camera, point cloud data is acquired using a 3D laser scanner, and thickness data is acquired using an ultrasonic thickness gauge.

[0015] In this embodiment, the preprocessing of the point cloud data and the thickness data includes: The point cloud data is denoised by employing statistical filtering or outlier removal algorithms (such as RANSAC algorithm), including removing isolated points, redundant points or error points. This process can remove irrelevant or erroneous points and improve the overall quality of the point cloud data. Signal filtering techniques (such as Kalman filtering and mean filtering) are used to filter and process the thickness data to improve the signal-to-noise ratio, making the processed data cleaner and facilitating subsequent feature extraction and distribution, thereby improving the accuracy of corrosion detection.

[0016] S2: A convolutional neural network model based on a deep residual network is used to extract texture features, color features, and edge features from high-definition image data. Corrosion analysis is performed based on these three channel features to obtain the first corrosion detection result, which includes the location, shape, and area of ​​corrosion.

[0017] Furthermore, the process of extracting texture features, color features, and edge features from high-resolution image data includes: Step 1: Denoise the high-definition image data using a Gaussian filter; The RGB values ​​of each pixel in the denoised high-definition image data are normalized. Data enhancement is performed on the high-definition image data; Step 2: Use ResNet as the backbone network to process the high-resolution image data after Step 1; ResNet is formed by stacking multiple residual blocks, and each residual block includes two or three convolutional operations; In ResNet, a small-sized convolutional kernel is defined, and this kernel is used to slide over high-resolution image data. By utilizing the local receptive field operation of convolutional kernels, texture features in high-resolution image data can be captured. Extract color features of RGB channels from high-resolution image data through multi-channel operations of convolution kernels; By embedding the Sobel operator into the convolution kernel, the boundary information of the rusted area in high-resolution image data is enhanced, and edge features are obtained.

[0018] Convolutional neural network models can effectively identify rusted areas and separate them from the background, outputting information such as the specific location, area, and shape of the rusted areas, providing accurate image feature input for subsequent multimodal data fusion.

[0019] Furthermore, the process of corrosion analysis based on texture features, color features, and edge features includes: Step 1: For the feature map output by ResNet, denoted as: Where X represents the feature map, H represents the feature map height, W represents the feature map width, and 3 is the number of channels; the feature map includes texture features, color features, and edge features; Step 2: Perform a global average on the three channels to obtain the global description vector of each channel. The calculation formula is: ; in, This represents the global description vector for the c-th channel; This represents the value of the c-th channel at position (i,j) in the feature map; Step 3: Input the global description vector into the fully connected layer to obtain the channel weights, calculated as follows: ; in, This represents the channel weight of the c-th channel; This represents the weight matrix of the first fully connected layer; This represents the weight matrix of the second fully connected layer; Represents the ReLU activation function; This represents the Sigmoid activation function; Step 4: Apply the channel weights to the corresponding channels to obtain the channel-weighted feature map; Step 5: Input the channel-weighted feature map into the classification network to obtain the first corrosion detection result, which includes the location, shape, and area of ​​corrosion.

[0020] The preprocessed point cloud data is processed using a random sampling consistency plane fitting algorithm to extract point cloud features from the surface of the rod. Based on the point cloud features, local curvature changes are analyzed to identify the second corrosion detection results, including depressions and pits on the surface of the rod.

[0021] Furthermore, the process of extracting point cloud features from the surface of the rod includes: Step 1: Set parameters, including maximum number of iterations, distance threshold, and interior point ratio threshold; Step 2: In each iteration, randomly select 3 points from the preprocessed point cloud data. A fitting plane is determined based on the three points, and the equation of the fitting plane is: ; ; ; Where a, b, and c are the three components of the normal vector n of the fitting plane; d represents the offset, which is the distance from the fitting plane to the origin, and determines the position of the fitting plane in three-dimensional space; x, y, and z are the x-coordinate, y-coordinate, and z-coordinate of any point in the point cloud data, respectively. Step 3: Calculate the distance from any point in the point cloud data to the fitting plane. The calculation formula is: ; in, This represents the distance from the i-th point to the fitted plane; , , Let x, y, and y represent the x-coordinate, y-coordinate, and y-coordinate of the i-th point, respectively. Step 4: Calculate the distance from all points in the point cloud data to the fitting plane and make it equal to the distance threshold. The points are compared, and those points whose distance is less than the distance threshold are identified as interior points. The number of interior points is then counted. Step 5: If the number of interior points in the current iteration is greater than the number of interior points in the previous iteration, then update the three components a, b, and c, as well as d, in the fitted plane equation; Step 6: Repeat steps 2-5 until the maximum number of iterations is reached or the proportion of interior points exceeds the interior point proportion threshold, to obtain the optimal plane (a, b, c, d) and the set of interior points belonging to the optimal plane; Step 7: Project the point cloud data onto the optimal plane, and use the normal vector of the optimal plane as the plane feature; Step 8: Based on the optimal plane, calculate the covariance matrix of each point in the point cloud data. The calculation formula is: ; Where C represents the covariance matrix; k represents the number of neighborhood points; The centroid of the neighborhood points is represented by T; T represents the transpose. Represents the i-th neighboring point; For any point, solving its covariance matrix yields three eigenvalues. , , ,and < < ; The curvature at corresponding points is calculated based on three eigenvalues, using the following formula: ; Wherein, K represents curvature; the planar feature and the curvature together constitute the point cloud feature.

[0022] Furthermore, analysis of local curvature changes based on point cloud features includes: Areas with high curvature correspond to depressions or pits on the surface of the rod; The distribution of curvature is used to identify the area, depth, and pattern of corrosion on the surface of the rod.

[0023] The space frame structure involved in this embodiment, due to prolonged exposure to the environment, may have rust, dirt, mechanical damage, etc. on its surface, resulting in a large number of outliers and noise points in the collected 3D point cloud data. Therefore, it is necessary to use a random sampling consistency plane fitting algorithm to denoise the point cloud data and extract planar features, thereby identifying clean surface features and providing a reliable data foundation for subsequent rust detection.

[0024] By comparing the pre-processed thickness data with the normal thickness data, a third corrosion detection result, including the degree of internal corrosion of the member, is obtained.

[0025] S3: A multimodal data fusion strategy based on Bayesian inference is adopted to fuse the channel features, the point cloud features, and the conditional probability distribution of the thickness data to obtain the corrosion probability of each detection area. Based on the corrosion probability, the corrosion area is determined. The first corrosion detection result, the second corrosion detection result, and the third corrosion detection result of the corrosion area are combined to obtain a comprehensive corrosion assessment report of the corrosion area.

[0026] In this embodiment, the fusion formula for the conditional probability distribution is: ; ; ; ; in, The probability of corrosion in the corresponding detection area is represented by F; F represents any one of the channel features, the point cloud features, and the thickness data; and H represents the hypothetical variable. The weights representing channel features; The weights represent the features of the point cloud. Indicates the weight of thickness data; The conditional probability distribution representing the channel characteristics; The conditional probability distribution representing the features of a point cloud; This represents the conditional probability distribution of thickness data; To express summation; This indicates that the data follows a normal distribution and is used to describe the probability density of the data. The mean difference represents the channel characteristics and is used to measure the central value of the channel characteristics in this region. The mean difference of point cloud features is represented by the center value of the point cloud features in that region. The mean difference of thickness data reflects the average thickness of the member in that region; The variance represents the channel characteristics and is used to describe the degree of fluctuation in high-definition image data; The variance represents the features of a point cloud and is used to reflect the discreteness of point cloud data. The variance of the thickness data is used to represent the variability of thickness measurements.

[0027] Furthermore, determining the rusted area based on the rust probability includes: If the probability of corrosion in the detected area is greater than the set threshold, the corresponding detected area is marked as a corrosion area; otherwise, it is not marked.

[0028] In this embodiment, the comprehensive corrosion assessment report also includes a distribution map of the corrosion probability of each detection area and marked corrosion areas.

[0029] Bayesian inference can establish the correlation between different modal data by combining the conditional probability distributions of different modal data, thereby fusing the detection results of different sensors into a comprehensive corrosion assessment model. Through the multimodal data fusion strategy of Bayesian inference, the method provided in this embodiment can more comprehensively assess the severity and distribution of corrosion and output a comprehensive corrosion assessment report.

[0030] S4: Use an LSTM model to perform time-series analysis on channel features, point cloud features, and thickness data to predict future corrosion trends, and then conduct a corrosion risk assessment to obtain a future corrosion risk prediction; combine the comprehensive corrosion assessment report and the future corrosion risk prediction to obtain the final corrosion detection report.

[0031] Specifically, the steps include: The channel features, point cloud features, and thickness data at each time point are fused and represented as a time-series vector, denoted as: ; ; Where L represents the time series vector; This represents the feature fused from the channel features, point cloud features, and thickness data at time point t. Indicates the number of historical time points; This represents the channel characteristics at time point t. Represents the point cloud features at time point t; This represents the thickness data at time point t. Normalize the values ​​in the time series vector to the [0,1] interval; The normalized time-series vector is input into the LSTM model to output the future corrosion development trend, which includes the corrosion area growth trend, the distribution evolution of high curvature areas, and the intensification of corrosion depth.

[0032] In this embodiment, the LSTM model includes an input gate, a forget gate, and an output gate; The formula for the forget gate (which determines how much historical information is retained) is as follows: ; in, This represents the hidden state at time point t-1; This represents the output value of the forget gate at time t, with a value range of [0,1], and if This indicates that most historical information is retained. This indicates that most historical information has been forgotten; This represents the sigmoid activation function; The weight matrix represents the forget gate, used to adjust the influence of the hidden state and the current input on the forget gate output; The bias term representing the forget gate; The formula for the input gate (which determines how much new information to add) is: ; ; in, Represents the hyperbolic tangent function; This represents the output value of the input gate at time t, with a value range of [0,1], and if This means that the information at the current time step will be added to memory. This means that no new information will be introduced; This represents the weight matrix of the inputs, used to adjust the hidden states in the LSTM. and input Impact on new information; This represents the bias term of the input gate; This indicates the candidate cell state, i.e., the new memory content at this time step; A weight matrix representing the candidate cell state, used to calculate the weights of new information; Bias terms representing the state of candidate cells; The state update formula is: ; in, This represents the cell state at time point t. This represents the cell state at time point t-1; The formula for the output gate (generating the hidden state) is: ; ; in, This represents the hidden state at time point t. This represents the output value of the output gate at time point t, which is used to control the final output of the LSTM at the current time step; This represents the weight matrix of the output gate, used to adjust the hidden state. and input Impact on output gates; This represents the bias term of the output gate.

[0033] The mean squared error is used as the loss function to measure the error between the predicted value and the true value. The loss function is calculated as follows: ; Where N represents the number of time points; This represents the predicted value at the i-th time point; This represents the true value at the i-th time point; Historical corrosion data was divided into a training set (e.g., the first 80% of time points) and a validation set (the last 20% of time points). The data format was as follows: Input: A time series of length T ; Output: Predict the features at N future time points. ; The Adam optimizer was used to adjust the model weights, with the initial learning rate set to 10. -3 .

[0034] The model's prediction accuracy is evaluated by calculating the error between the predicted and actual values ​​using a validation set.

[0035] The future corrosion development trend is visualized as a trend chart. When the corrosion area in the trend chart increases rapidly, it is considered that the corresponding area has a risk of corrosion in the future; otherwise, it is considered that there is no risk of corrosion in the future. The final corrosion detection report is obtained by combining the comprehensive corrosion assessment report and the future corrosion risk prediction.

[0036] The corrosion detection method for space frame structures provided in this embodiment acquires high-resolution image data, point cloud data, and thickness data of the surface of the space frame members to be tested, as well as the thickness data of the interior of the space frame members. Channel features are extracted from the high-resolution image data for corrosion analysis to obtain a first corrosion detection result. Point cloud features are extracted based on the preprocessed point cloud data to identify a second corrosion detection result. The difference between the preprocessed thickness data and the normal thickness data is compared to obtain a third corrosion detection result. A multimodal data fusion strategy based on Bayesian inference fuses the channel features, point cloud features, and conditional probability distribution of the thickness data to obtain the corrosion probability of each detection area, thereby determining the corrosion area. The three corrosion detection results of the corrosion area are combined to obtain a comprehensive corrosion assessment report for the corrosion area. Finally, the comprehensive corrosion assessment report is combined with a future corrosion risk prediction based on an LSTM model to obtain a final corrosion detection report. This method overcomes the limitations of single data source detection, the lack of efficient feature extraction and classification methods, the inability to comprehensively assess the impact of different types of corrosion, and the lack of ability to predict corrosion development trends. It also addresses the issues of accuracy and reliability, providing an efficient and accurate feasible solution for the detection of corrosion targets in grid structures.

[0037] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0038] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for detecting corrosion targets of a grid structure, characterized in that, include: S1: Acquire high-definition image data, point cloud data, and thickness data of the surface of the space frame member to be tested, and preprocess the point cloud data and the thickness data respectively; S2: A convolutional neural network model based on a deep residual network is used to extract texture features, color features, and edge features from high-definition image data. Corrosion analysis is performed based on these three channel features to obtain the first corrosion detection result, which includes the location, shape, and area of ​​corrosion. The preprocessed point cloud data was processed using a random sampling consistency plane fitting algorithm to extract point cloud features from the surface of the rod. Based on the point cloud features, local curvature changes were analyzed to identify the second corrosion detection results, including depressions and pits on the surface of the rod. By comparing the difference between the pre-processed thickness data and the normal thickness data, a third corrosion detection result including the degree of internal corrosion of the rod is obtained; S3: A multimodal data fusion strategy based on Bayesian inference is adopted to fuse the channel features, the point cloud features and the conditional probability distribution of the thickness data to obtain the corrosion probability of each detection area, and the corrosion area is determined based on the corrosion probability. The first corrosion detection result, the second corrosion detection result and the third corrosion detection result of the corrosion area are combined to obtain a comprehensive corrosion assessment report of the corrosion area. S4: The LSTM model is used to perform time series analysis on channel features, point cloud features and thickness data to predict future corrosion development trends, and then to conduct corrosion risk assessment and obtain future corrosion risk prediction. The final corrosion detection report is obtained by combining the comprehensive corrosion assessment report and the future corrosion risk prediction.

2. The method of claim 1, wherein the method further comprises: The preprocessing of the point cloud data and the thickness data includes: The point cloud data is subjected to noise reduction processing; The thickness data is then filtered and processed.

3. The method of claim 1, wherein the method further comprises: The process of extracting texture features, color features, and edge features from high-resolution image data includes: Step 1: Denoise the high-definition image data using a Gaussian filter; The RGB values ​​of each pixel in the denoised high-definition image data are normalized. Data enhancement is performed on the high-definition image data; Step 2: Use ResNet as the backbone network to process the high-resolution image data after Step 1; ResNet is formed by stacking multiple residual blocks, and each residual block includes two or three convolutional operations; In ResNet, a small-sized convolutional kernel is defined, and this kernel is used to slide over high-resolution image data. By utilizing the local receptive field operation of convolutional kernels, texture features in high-resolution image data can be captured. Extract color features of RGB channels from high-resolution image data through multi-channel operations of convolution kernels; By embedding the Sobel operator into the convolution kernel, the boundary information of the rusted area in high-resolution image data is enhanced, and edge features are obtained.

4. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, The process of corrosion analysis based on texture features, color features, and edge features includes: Step 1: For the feature map output by ResNet, denoted as: Where X represents the feature map, H represents the feature map height, W represents the feature map width, and 3 is the number of channels; the feature map includes texture features, color features, and edge features; Step 2: Perform a global average on the three channels to obtain the global description vector of each channel. The calculation formula is: ； in, Represents the global description vector of the c-th channel; This represents the value of the c-th channel at position (i,j) in the feature map; Step 3: Input the global description vector into the fully connected layer to obtain the channel weights, calculated as follows: ； in, This represents the channel weight of the c-th channel; This represents the weight matrix of the first fully connected layer; This represents the weight matrix of the second fully connected layer; Represents the ReLU activation function; This represents the Sigmoid activation function; Step 4: Apply the channel weights to the corresponding channels to obtain the channel-weighted feature map; Step 5: Input the channel-weighted feature map into the classification network to obtain the first corrosion detection result, which includes the location, shape, and area of ​​corrosion.

5. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, The process of extracting point cloud features from the surface of the rod includes: Step 1: Set parameters, including maximum number of iterations, distance threshold, and interior point ratio threshold; Step 2: In each iteration, randomly select 3 points from the preprocessed point cloud data. A fitting plane is determined based on the three points, and the equation of the fitting plane is: ； ； ； Where a, b, and c are the three components of the normal vector n of the fitting plane; d represents the offset; and x, y, and z are the x-coordinate, y-coordinate, and z-coordinate of any point in the point cloud data, respectively. Step 3: Calculate the distance from any point in the point cloud data to the fitting plane. The calculation formula is: ； in, This represents the distance from the i-th point to the fitted plane; , , Let x, y, and y represent the x-coordinate, y-coordinate, and y-coordinate of the i-th point, respectively. Step 4: Calculate the distance from all points in the point cloud data to the fitting plane and make it equal to the distance threshold. The points are compared, and those points whose distance is less than the distance threshold are identified as interior points. The number of interior points is then counted. Step 5: If the number of interior points in the current iteration is greater than the number of interior points in the previous iteration, then update the three components a, b, and c, as well as d, in the fitted plane equation; Step 6: Repeat steps 2-5 until the maximum number of iterations is reached or the proportion of interior points exceeds the interior point proportion threshold, to obtain the optimal plane (a, b, c, d) and the set of interior points belonging to the optimal plane; Step 7: Project the point cloud data onto the optimal plane, and use the normal vector of the optimal plane as the plane feature; Step 8: Based on the optimal plane, calculate the covariance matrix of each point in the point cloud data. The calculation formula is: ； Where C represents the covariance matrix; k represents the number of neighborhood points; The centroid of the neighborhood points is represented by T; T represents the transpose. Represents the i-th neighboring point; For any point, solving its covariance matrix yields three eigenvalues. , , ,and < < ; The curvature at corresponding points is calculated based on three eigenvalues, using the following formula: ； Wherein, K represents curvature; the planar feature and the curvature together constitute the point cloud feature.

6. The method for detecting corrosion targets in a space frame structure according to claim 5, characterized in that, Analysis of local curvature changes based on point cloud features includes: Areas with high curvature correspond to depressions or pits on the surface of the rod; The distribution of curvature is used to identify the area, depth, and pattern of corrosion on the surface of the rod.

7. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, The fusion formula for conditional probability distributions is: ； ； ； ； in, The value represents the corrosion probability of the corresponding detection area; F represents any one of the channel features, the point cloud features, and the thickness data; and H represents the hypothetical variable. The weights representing channel features; The weights represent the features of the point cloud. Indicates the weight of thickness data; The conditional probability distribution representing the channel characteristics; The conditional probability distribution representing the features of a point cloud; This represents the conditional probability distribution of thickness data; To express summation; This indicates that the data follows a normal distribution; The mean difference representing the channel characteristics; The mean difference of point cloud features; This represents the mean difference in thickness data; The variance representing the channel characteristics; The variance representing the features of the point cloud; This represents the variance of the thickness data.

8. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, The determination of the corrosion area based on the corrosion probability includes: When the probability of corrosion in the detected area is greater than the set threshold, the corresponding detected area is marked as a corrosion area. Otherwise, do not mark it.

9. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, The comprehensive corrosion assessment report also includes a distribution map of the corrosion probability of each detected area and the marked corrosion areas.

10. The method for detecting corrosion targets in a space frame structure according to claim 1, characterized in that, S4 includes: The channel features, point cloud features, and thickness data at each time point are fused and represented as a time-series vector, denoted as: ； ； Where L represents the time vector; This represents the feature fused from the channel features, point cloud features, and thickness data at time point t. Indicates the number of historical points in time; This represents the channel characteristics at time point t. Represents the point cloud features at time point t; This represents the thickness data at time point t. Normalize the values ​​in the time series vector to the [0,1] interval; The normalized time-series vector is input into the LSTM model to output the future corrosion development trend, which includes the corrosion area growth trend, the distribution evolution of high curvature areas, and the intensification of corrosion depth. The future corrosion development trend is visualized as a trend chart. When the corrosion area in the trend chart increases rapidly, it is considered that the corresponding area has a risk of corrosion in the future; otherwise, it is considered that there is no risk of corrosion in the future. The final corrosion detection report is obtained by combining the comprehensive corrosion assessment report and the future corrosion risk prediction.

Images