Ancient building wall column defect grading method based on multi-source data fusion and machine learning

Abstract

The application discloses a method for grading defects of ancient building wall wood columns based on multi-source data fusion and machine learning, which comprises the following steps: building a wood column apparent feature information collection device to realize wood column defect image collection and enhancement processing; designing and building a wood column internal defect data collection device to collect micro drill resistance data and process the data; measuring wood column entity missing data and integrating the data into the micro drill resistance data; expanding the data set and fusing multi-source data sets; selecting wall wood column defect grading indexes, determining defect grading standards and grading defects of ancient building wall wood columns; taking the grading results as labels, establishing a GWO-RBF neural network model and grading wood column defects based on machine learning. The application fuses multi-source data, utilizes a machine learning neural network model and realizes accurate, rapid and non-destructive grading detection of wood column defects.

Description

Technical Field

[0001] This invention belongs to the field of ancient building timber testing technology, specifically involving a method for classifying defects in ancient building wall wooden columns based on multi-source data fusion and machine learning. Background Technology

[0002] Ancient wooden structures in my country are an important cultural heritage, and wooden pillars are crucial load-bearing components. As a biological material, wooden pillars are susceptible to erosion by various fungi and insects in the natural environment, causing various surface and internal defects. Wall-mounted wooden pillars, which are in contact with or partially or completely encased in walls, are often situated in dark and damp environments, further promoting biological erosion and exacerbating problems such as insect infestation, decay, and damage. These defects reduce the load-bearing capacity of the wooden pillars, seriously threatening the safety and stability of ancient buildings.

[0003] Non-destructive testing and defect classification of wooden pillars encased in ancient building walls are crucial for the early detection and assessment of defects and corrosion, and are essential for the preventative protection of these structures. Commonly used non-destructive testing methods for wood include stress wave testing, infrared thermography, X-ray testing, and machine vision testing, each with its own applicability and limitations. Due to the unique location of wooden pillars in ancient building walls and the difficulty of sampling, it is often necessary to combine multiple methods to comprehensively and reliably detect defects.

[0004] Combining machine vision with micro-drilling resistance methods can overcome the limitation of machine vision in detecting deep internal defects, which is limited to the surface defects of wooden pillars. Defect classification based on machine learning improves assessment efficiency and provides a faster and more accurate reference for the maintenance, repair and reinforcement of wooden pillars in ancient buildings. Summary of the Invention

[0005] To achieve the above objectives, this invention proposes a method for classifying defects in wooden columns of ancient building walls based on multi-source data fusion and machine learning. The method integrates images of wooden column defects with micro-drill resistance data, and combines machine vision with machine learning methods to achieve rapid identification of defect levels in actual inspections of wooden columns of ancient building walls.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A method for classifying defects in wooden columns of ancient building walls based on multi-source data fusion and machine learning includes the following steps:

[0008] S1. Collect multi-source data, use a camera to record images of wooden pillar defects containing surface defect information, use a micro-drill resistance detector to acquire internal defect data of the wooden pillar and import it into the entity missing data, including:

[0009] S11. Construct a device for acquiring surface feature information of wooden pillars to realize the acquisition and enhancement processing of images of defects in wooden pillars. The acquisition and processing of images of defects in wooden pillars includes:

[0010] Images of surface defects in the wooden pillars of ancient buildings were recorded using a digital camera. Information on these defects was collected, and the images were cropped to 1100×500 pixels based on the locations detected by a micro-drill resistance meter. Image enhancement processing was then performed, including:

[0011] Color transformation and normalization are performed, and Gaussian filtering and median filtering are used to smooth the image and remove noise.

[0012] S12. Design and construct a data acquisition device for internal defects of wooden pillars, collect and process micro-drilling resistance data. The collection and processing of micro-drilling resistance data includes:

[0013] A micro-drill resistance tester was used to collect micro-drill resistance data inside the wooden column, obtain information on internal defects, and analyze the types and distribution of defects inside the wooden column, including:

[0014] Driven by an electric motor, the drill bit of a micro-drill resistance meter is drilled into the interior of a wooden column at a constant rate. A microcomputer system collects the resistance experienced by the drill bit and displays the resistance data curve. Combining wood science knowledge, the micro-drill resistance data curve is analyzed to determine the types and distribution of defects inside the wooden column. Processing the micro-drill resistance data includes:

[0015] Adding missing data, deleting idle data, and adjusting sample length;

[0016] S13. Measure the missing data of the wooden pillar and import it into the micro-drill resistance data, including:

[0017] Use probes, rulers, and other tools to measure the missing data of the wooden pillar. Use the measured missing data as a supplement to incorporate the missing dimensions into the micro-drill resistance data curve.

[0018] S2. Expand the dataset and integrate multi-source data information, including:

[0019] The micro-drill resistance data after the missing entity data is imported is matched one-to-one with the image of the wooden column defect. Multi-source datasets are fused, including data-level fusion, feature-level fusion and decision-level fusion, to obtain fused data with correlation and integration, and to integrate the feature information of the wooden column surface and internal defects.

[0020] Furthermore, a device for collecting visual feature information of wooden pillars is constructed, including:

[0021] Taking the exposed part of the wooden columns in the ancient building walls as the collection samples, select a suitable light source. For bright sites, use natural light illumination; for dark places, use artificial supplementary lighting to determine the light intensity. Fix the camera stand and adjust the shooting angle and distance of the camera.

[0022] Furthermore, design and build a data collection device for internal defects of wooden columns, including:

[0023] The collection device consists of a support frame 1, a workbench 2, and a micro-drilling resistance instrument 3. The support frame 1 includes a base 11, a telescopic rod 12, and a support plate 13. The support plate 13 bears the loads of the workbench 2 and the micro-drilling resistance instrument 3 and transmits them to the ground through the base 11. The upper part of the support plate 13 is connected to the workbench 2 through a positioning pin; the workbench 2 includes a clamp 21, a positioning groove 22, and a limit stop 23. Spiral clamping devices are installed on both sides of the clamp 21, and the body of the micro-drilling resistance instrument 3 is clamped by adjusting the knob in the horizontal direction of the threaded rod. The positioning groove 22 is connected to the lower end of the body support of the micro-drilling resistance instrument 3, and the limit stop 23 is connected to the rear end of the body of the micro-drilling resistance instrument 3 to limit the forward and backward displacement of the body during operation.

[0024] S3. Select the defect grading indicators for the wooden columns in the wall, determine the defect grading standards, and conduct defect grading for the wooden columns in the ancient building walls, including:

[0025] According to national standards, select the external defect area, internal defect area, crack length, crack depth data and their proportions in the wooden column as evaluation indicators. The proportion of the cross-sectional area of the external defect in the cross-sectional area of the wooden column is the external defect area proportion coefficient k, the proportion of the cross-sectional area of the internal defect in the cross-sectional area of the wooden column is the internal defect area proportion coefficient p, the proportion of the crack length in the length of the wooden column is the crack length proportion coefficient m, and the proportion of the crack depth in the diameter of the wooden column is the crack depth proportion coefficient n;

[0026] According to the defect grading indicators of the wooden columns and combined with the on-site data accumulated by experts and the analysis results of laboratory data, the wooden columns are divided into three grades: A, B, and C:

[0027] Grade A: The range of the external defect area proportion coefficient is k ≤ 0.05, the internal defect area proportion coefficient p = 0, the range of the crack length proportion coefficient is m ≤ 0.1, and the range of the crack depth proportion coefficient is n ≤ 0.1. Wooden columns of Grade A are in good condition and do not require intervention;

[0028] Grade B: The range of the external defect area proportion coefficient is 0.05 < k < 0.50, the range of the internal defect area proportion coefficient is 0 < p < 0.15, the range of the crack length proportion coefficient is 0.1 < m < 0.8, and the range of the crack depth proportion coefficient is 0.1 < n < 0.8. Wooden columns of Grade B have defects and do not require intervention but need to be regularly rechecked;

[0029] Grade C: External defect area ratio coefficient ranges from k≥0.50, internal defect area ratio coefficient ranges from p≥0.15, crack length ratio coefficient ranges from m≥0.8, and crack depth ratio coefficient ranges from n≥0.8. Grade C wooden components have defects and require appropriate reinforcement of the wooden columns.

[0030] S4. Using the defect grading results of wooden columns as labels, establish a GWO-RBF neural network model to perform machine learning-based grading of wall wooden column defects, including:

[0031] The GWO-RBF neural network model is a radial basis function (RBF) neural network model optimized by the Grey Wolf Optimization Algorithm (GWO), consisting of an input layer, hidden layers, and an output layer.

[0032] The input vector is X = [x1, x2, ..., x...]. n ] T The output vector is Y = [y1, y2, ..., y]. m ] T The hidden layer uses radial basis functions for feature mapping and employs a Gaussian function:

[0033]

[0034] Among them, c i =[c i1 ,c i2 ,…,c ih ] T σ is the center of the Gaussian function. i Let be the variance of the Gaussian function;

[0035] The output of the hidden layer is:

[0036]

[0037] Among them, ||X i -c i || represents the distance between the input function vector and the center of the radial basis functions;

[0038] The output layer functions use a linear mapping relationship:

[0039]

[0040] Where, ω ij The weights connecting the hidden layer and the output layer;

[0041] The number of nodes in the input layer, hidden layer, and output layer corresponds to the dimension of the input vector, the dimension of the Gaussian function center, and the dimension of the output vector, respectively. That is, the number of nodes in the input layer is the number of features in the dataset, and the number of nodes in the output layer is the number of defect levels.

[0042] The Grey Wolf Optimization Algorithm (GWO) updates the target position by simulating the behavior of the leader, co-leader, and followers in a grey wolf pack, continuously optimizing the weights and biases of the Restricted Baseline Fiber (RBF) neural network, and using mean squared error to evaluate the performance of the GWO-RBF neural network.

[0043]

[0044] Among them, y i f(x) is the actual value. i The predicted value is given, and the position vector of the corresponding optimal solution is returned to the RBF network to update the original parameter values. The training is iterated until the termination condition is met.

[0045] Using the GWO-RBF neural network model, machine learning methods are used to achieve accurate, fast, and non-destructive graded detection of defects in wooden columns of ancient building walls.

[0046] Compared with existing technologies, the advantages of this invention are:

[0047] To address the shortcomings of traditional detection methods, such as their reliance on single methods, low efficiency, and poor robustness, this invention offers innovative improvements. By integrating multi-source datasets and utilizing machine vision and machine learning neural network models, it achieves accurate, rapid, and efficient classification of defects in wooden pillars. Compared to traditional models, this model achieves higher accuracy in detecting the level of defects in wooden pillars in walls, with an overall accuracy exceeding 98%. Attached Figure Description

[0048] Figure 1 This is a flowchart of a method for classifying defects in wooden pillars of ancient building walls based on multi-source data fusion and machine learning.

[0049] Figure 2 This is a schematic diagram of a micro-drill resistance data acquisition device inside a wooden pillar;

[0050] In the diagram, 1. Support frame; 11. Base; 12. Telescopic rod; 13. Support plate; 2. Workbench; 21. Fixture; 22. Positioning slot; 23. Limiting block; 3. Micro-drill resistance gauge;

[0051] Figure 3 This is a schematic diagram of the RBF neural network model in an embodiment of the method of the present invention. Detailed Implementation

[0052] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0053] like Figure 1 As shown, the present invention provides a method for classifying defects in wooden columns of ancient building walls based on multi-source data fusion and machine learning, comprising the following steps:

[0054] S1. Collect multi-source data, use a camera to record images of wooden pillar defects containing surface defect information, use a micro-drill resistance detector to acquire internal defect data of the wooden pillar and import it into the entity missing data, including:

[0055] S11. Construct a device for acquiring surface feature information of wooden pillars to realize the acquisition and enhancement processing of images of defects in wooden pillars. The acquisition and processing of images of defects in wooden pillars includes:

[0056] Images of surface defects in the wooden pillars of ancient buildings were recorded using a digital camera. Information on these defects was collected, and the images were cropped to 1100×500 pixels based on the locations detected by a micro-drill resistance meter. Image enhancement processing was then performed, including:

[0057] Color transformation and normalization are performed, and Gaussian filtering and median filtering are used to smooth the image and remove noise.

[0058] S12. Design and construct a data acquisition device for internal defects of wooden pillars, collect and process micro-drilling resistance data. The collection and processing of micro-drilling resistance data includes:

[0059] A micro-drill resistance tester was used to collect micro-drill resistance data inside the wooden column, obtain information on internal defects, and analyze the types and distribution of defects inside the wooden column, including:

[0060] Driven by an electric motor, the drill bit of a micro-drill resistance meter is drilled into the interior of a wooden column at a constant rate. A microcomputer system collects the resistance experienced by the drill bit and displays the resistance data curve. Combining wood science knowledge, the micro-drill resistance data curve is analyzed to determine the types and distribution of defects inside the wooden column. Processing the micro-drill resistance data includes:

[0061] Adding missing data, deleting idle data, and adjusting sample length;

[0062] S13. Measure the missing data of the wooden pillar and import it into the micro-drill resistance data, including:

[0063] Use probes, rulers, and other tools to measure the missing data of the wooden pillar. Use the measured missing data as a supplement to incorporate the missing dimensions into the micro-drill resistance data curve.

[0064] S2. Expand the dataset and integrate multi-source data information, including:

[0065] The micro-drill resistance data after the missing entity data is imported is matched one-to-one with the wooden column defect image, and multi-source datasets are fused, including data-level fusion, feature-level fusion and decision-level fusion, to obtain fused data with correlation and integration, and to integrate the feature information of the wooden column surface and internal defects.

[0066] In this embodiment, the apparatus for collecting the apparent feature information of wooden columns includes:

[0067] Taking the exposed part of the wooden columns in the ancient building wall as the collection sample, selecting a suitable light source. In a bright scene, natural light is used for illumination, and in a dark place, artificial supplementary light is used to determine the light intensity. Fix the camera bracket and adjust the shooting angle and distance of the camera.

[0068] In this embodiment, the apparatus for designing and building the internal defect data collection of wooden columns includes:

[0069] It consists of three parts: a support frame, a workbench, and a micro-drilling resistance instrument. The support frame includes a base, a telescopic rod, and a support plate, which can bear the loads of the workbench and the micro-drilling resistance instrument and adjust the height of the device. The workbench includes a fixture, a positioning groove, and a limit stop. The body of the micro-drilling resistance instrument is clamped by adjusting the screw clamping devices installed on both sides of the fixture, and the ineffective displacement of the body of the micro-drilling resistance instrument during operation is restricted by the positioning groove and the limit stop.

[0070] S3. According to the national standard, select the grading evaluation indexes of the defects of the wooden columns in the wall, and combine the on-site data accumulated by experts in the early stage and the analysis results of laboratory data to classify the defect conditions of the measured wooden columns, including:

[0071] Select the external defect area, internal defect area, crack length, crack depth data and their proportions in the wooden column as the evaluation indexes. The proportion of the cross-sectional area of the external defect in the cross-sectional area of the wooden column is the external defect area proportion coefficient k, the proportion of the cross-sectional area of the internal defect in the cross-sectional area of the wooden column is the internal defect area proportion coefficient p, the proportion of the crack length in the length of the wooden column is the crack length proportion coefficient m, and the proportion of the crack depth in the diameter of the wooden column is the crack depth proportion coefficient n;

[0072] In this embodiment, according to the wooden column defect grading index, the wooden columns are divided into three grades: A, B, and C:

[0073] Grade A: The range of the external defect area proportion coefficient is k ≤ 0.05, the internal defect area proportion coefficient p = 0, the range of the crack length proportion coefficient is m ≤ 0.1, and the range of the crack depth proportion coefficient is n ≤ 0.1. Wooden columns of grade A are in good condition and do not require intervention;

[0074] Grade B: The range of the external defect area proportion coefficient is 0.05 < k < 0.50, the range of the internal defect area proportion coefficient is 0 < p < 0.15, the range of the crack length proportion coefficient is 0.1 < m < 0.8, and the range of the crack depth proportion coefficient is 0.1 < n < 0.8. Wooden columns of grade B have defects and do not require intervention but need to be regularly rechecked;

[0075] Grade C: External defect area ratio coefficient ranges from k≥0.50, internal defect area ratio coefficient ranges from p≥0.15, crack length ratio coefficient ranges from m≥0.8, and crack depth ratio coefficient ranges from n≥0.8. Grade C wooden components have defects and require appropriate reinforcement of the wooden columns.

[0076] S4. Using the defect grading results of wooden columns as labels, establish a GWO-RBF neural network model to perform machine learning-based grading of wall wooden column defects, including:

[0077] The GWO-RBF neural network model is a radial basis function (RBF) neural network model optimized by the Grey Wolf Optimization Algorithm (GWO), consisting of an input layer, hidden layers, and an output layer.

[0078] The hidden layer uses a Gaussian function as the radial basis function for feature mapping. The output of the hidden layer depends on the distance between the input function vector and the center of the radial basis function. The output layer uses a linear mapping relationship.

[0079] The number of nodes in the input layer, hidden layer, and output layer correspond to the dimensions of the input vector, the center of the Gaussian function, and the output vector, respectively.

[0080] The Grey Wolf Optimization Algorithm (GWO) updates the target position by simulating the behavior of the leader, co-leader, and followers in a grey wolf pack, continuously optimizing the weights and biases of the Restricted Baseline Fiber (RBF) neural network, and using mean squared error to evaluate the performance of the GWO-RBF neural network.

[0081]

[0082] Among them, y i f(x) is the actual value. i The predicted value is given, and the position vector of the corresponding optimal solution is returned to the RBF network to update the original parameter values. The training is iterated until the termination condition is met.

[0083] Using the GWO-RBF neural network model, based on machine learning methods, we can achieve accurate, fast, and non-destructive graded detection of defects in the wooden columns of ancient building walls.

[0084] In this embodiment, the model achieved a high accuracy rate in detecting the defect level of wooden columns in the wall, with an overall accuracy rate of over 98%.

[0085] like Figure 2 As shown, in this embodiment, the design and construction of a data acquisition device for internal defects of wooden pillars includes:

[0086] The data acquisition device consists of a support frame 1, a worktable 2, and a micro-drill resistance meter 3. The support frame 1 includes a base 11, a telescopic rod 12, and a support plate 13. The support plate 13 bears the load of the worktable 2 and the micro-drill resistance meter 3 and transmits it to the ground through the base 11. The upper part of the support plate 13 is connected to the worktable 2 via a positioning pin. The worktable 2 includes a clamp 21, a positioning groove 22, and a limiting block 23. Screw clamping devices are installed on both sides of the clamp 21. The body of the micro-drill resistance meter 3 is clamped by adjusting the knob in the horizontal direction of the threaded rod. The positioning groove 22 is connected to the lower end of the support of the micro-drill resistance meter 3 body. The limiting block 23 is connected to the rear end of the micro-drill resistance meter 3 body, limiting the forward and backward displacement of the body during operation. This device has good stability, is simple to operate, and is easy to carry.

[0087] like Figure 3 As shown, the RBF neural network model includes an input layer, hidden layers, and an output layer. The principle of the RBF neural network model includes:

[0088] The input vector is X = [x1, x2, x3] T The output vector is Y = [y1, y2, y3]. T The hidden layer uses radial basis functions for feature mapping and employs a Gaussian function:

[0089]

[0090] Among them, c i =[c i1 ,c i2 ,…,c ih ] T σ is the center of the Gaussian function. i Let be the variance of the Gaussian function;

[0091] The output of the hidden layer is:

[0092]

[0093] Among them, ||X i -c i || represents the distance between the input function vector and the center of the radial basis functions;

[0094] The output layer functions use a linear mapping relationship:

[0095]

[0096] Where, ω ij The weights connecting the hidden layer and the output layer;

[0097] The number of nodes in the input layer, hidden layer, and output layer corresponds to the dimension of the input vector, the dimension of the Gaussian function center, and the dimension of the output vector, respectively. That is, the number of nodes in the input layer is the number of features in the dataset, and the number of nodes in the output layer is the number of defect levels.

Claims

1. A method for classifying defects in wooden columns of ancient building walls based on multi-source data fusion and machine learning, characterized in that, The method includes: A device for acquiring surface feature information of wooden pillars was constructed to realize the acquisition and enhancement processing of defect images of wooden pillars; Design and build a data acquisition device for internal defects of wooden pillars, collect and process micro-drill resistance data, including: The data acquisition device for internal defects of the wooden column consists of a support frame, a workbench and a micro-drill resistance meter. The support frame is equipped with a telescopic rod to adjust the working height of the micro-drill resistance meter. The workbench fixes the body of the micro-drill resistance meter to limit its ineffective displacement during operation. Measure the missing data of the wooden pillar and import the data into the micro-drill resistance data; The dataset was expanded and multi-source datasets were merged to integrate the feature information of surface and internal defects of the wooden pillars; The area of ​​external defects, the area of ​​internal defects, the length of cracks, the depth of cracks, and their proportions in the wooden columns were selected as defect classification indicators. Among them, the proportion of the cross-sectional area of ​​external defects to the cross-sectional area of ​​the wooden column was the external defect area ratio coefficient k, the proportion of the cross-sectional area of ​​internal defects to the cross-sectional area of ​​the wooden column was the internal defect area ratio coefficient p, the proportion of the crack length to the length of the wooden column was the crack length ratio coefficient m, and the proportion of the crack depth to the diameter of the wooden column was the crack depth ratio coefficient n. Based on the previously accumulated field data and laboratory analysis results, the defect classification standard was determined, and the defects of the wooden columns of ancient building walls were classified. Using the defect classification standard as a label, a GWO-RBF neural network model was established to classify defects in wooden wall columns based on machine learning. The GWO-RBF neural network model is a radial basis function (RBF) neural network model optimized by the Grey Wolf Optimization Algorithm (GWO). The RBF neural network model includes an input layer, a hidden layer, and an output layer, wherein the input vector is... The output vector is The hidden layer uses radial basis functions for feature mapping, and the method employs a Gaussian function: in, The center of the Gaussian function, Let be the variance of the Gaussian function; The output of the hidden layer is: in, The distance between the input function vector and the center of the radial basis functions; The output layer functions use a linear mapping relationship: in, The weights connecting the hidden layer and the output layer; The number of nodes in the input layer, hidden layer, and output layer corresponds to the dimension of the input vector, the dimension of the Gaussian function center, and the dimension of the output vector, respectively. That is, the number of nodes in the input layer is the number of features in the dataset, and the number of nodes in the output layer is the number of defect levels. The Gray Wolf Optimization Algorithm (GWO) updates the target position by simulating the behavior of the leader, co-leader, and followers in a gray wolf pack, continuously optimizing the weights and biases of the RBF neural network, and using mean squared error to evaluate the performance of the GWO-RBF neural network. in, This is the actual value. The predicted value is returned to the RBF network to update the original parameter values. The training continues iteratively until the termination condition is met. Using the GWO-RBF neural network model, a graded detection of defects in the wooden columns of ancient building walls is achieved based on machine learning methods.

2. The method based on claim 1, characterized in that, The construction of the wooden pillar appearance feature information acquisition device includes: The exposed wooden pillars of ancient building walls were used as the sampling points. Appropriate light sources were selected. Natural lighting was used in bright areas, and artificial lighting was used in dark areas to determine the light intensity. The camera bracket was fixed, and the shooting angle and distance of the camera were adjusted.

3. The method based on claim 1 or 2, characterized in that, The acquisition and processing of the images of defects in the wooden pillars includes: The image of the wooden pillar defect containing information about the surface defects of the wooden pillar is recorded using a camera, and the image of the wooden pillar defect is cropped into a size of 1100×500 pixels according to the detection position of the micro-drill resistance instrument; The image of the surface defects of the wooden pillar is enhanced by color transformation and normalization, and Gaussian filtering and median filtering are used to smooth the image and remove noise.

4. The method based on claim 1, characterized in that, The construction of the data acquisition device for internal defects of the wooden pillar includes: The support frame includes a base, a telescopic rod, and a support plate. The support plate bears the loads of the workbench and the micro-drilling resistance meter, and transmits them to the ground through the base. The upper part of the support plate is connected to the workbench by a positioning pin. The workbench includes a clamp, a positioning groove, and a limit stop. Spiral clamping devices are installed on both sides of the clamp, and the body of the micro-drilling resistance meter is clamped by adjusting the knob on the horizontal direction of the threaded rod. The positioning groove is connected to the lower end of the body support of the micro-drilling resistance meter, and the limit stop is connected to the rear end of the micro-drilling resistance meter to limit the forward and backward displacement of the body during operation.

5. The method based on claim 1, characterized in that, The acquisition and processing of the micro-drilling resistance data and the import of the missing data of the wooden column entity include: Under the drive of the motor, the drill bit of the micro-drilling resistance meter is drilled into the wooden column at a constant speed. The resistance data received by the drill bit is collected by the microcomputer system, the micro-drilling resistance data is processed, and combined with the knowledge of wood science to explore the types and distribution of defects inside the wooden column. Tools such as probes and measuring tapes are used to measure the missing data of the wooden column entity. Taking the measured missing data of the entity as a supplement, the missing dimensions of the entity are imported into the micro-drilling resistance data curve together, including adding missing data, deleting idling data, and adjusting the sample length of the micro-drilling resistance data.

6. The method based on claim 1 or 5, characterized in that, The fusion of the multi-source data sets includes: The micro-drilling resistance data after importing the missing data of the entity is corresponding to the wooden column defect images one by one to perform the fusion of the multi-source data sets, including data-level fusion, feature-level fusion, and decision-level fusion, and synthesizing the characteristic information of the surface and internal defects of the wooden column.

7. The method based on claim 1, characterized in that, The grading standard for the defects of the wall wooden columns is: According to the wooden column defect grading index, the wooden columns are divided into three grades A, B, and C, including: Grade A: The range of the external defect area proportion coefficient is k≤0.05, the internal defect area proportion coefficient p = 0, the range of the crack length proportion coefficient is m≤0.1, and the range of the crack depth proportion coefficient is n≤0.

1. Grade A wooden columns are in good condition and do not require intervention. Grade B: The range of the external defect area proportion coefficient is 0.05 < k < 0.50, the range of the internal defect area proportion coefficient is 0 < p < 0.15, the range of the crack length proportion coefficient is 0.1 < m < 0.8, and the range of the crack depth proportion coefficient is 0.1 < n < 0.

8. Grade B wooden columns have defects and do not require intervention but need to be regularly rechecked. Grade C: The range of the external defect area proportion coefficient is k≥0.50, the range of the internal defect area proportion coefficient is p≥0.15, the range of the crack length proportion coefficient is m≥0.8, and the range of the crack depth proportion coefficient n is n≥0.

8. Grade C wooden members have defects and need to appropriately reinforce the wooden columns.

Images