Clothing damage detection method and system based on flexible electric wire grid
By using multimodal data processing and feature fusion technology with flexible wire mesh, the problems of low efficiency, strong subjectivity and poor accuracy in traditional clothing damage detection have been solved, realizing real-time, automatic and accurate detection and assessment of clothing damage.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-15
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional methods for detecting damage to clothing are inefficient, subjective, and difficult to detect minor damage. Furthermore, real-time detection is cumbersome and destructive, making it impossible to achieve precise location.
A clothing damage detection method based on flexible electric wire mesh is adopted. By simultaneously collecting multimodal data, bias correction and combined preprocessing are performed to construct feature signals and use attention mechanism for dynamic weighted fusion. After dimensionality reduction, the data is input into a pre-trained damage quantization model for evaluation.
It enables real-time, automated, and accurate detection and assessment of clothing damage, overcoming the high false alarm problem of traditional methods and improving detection efficiency and accuracy.
Smart Images
Figure CN121834432A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of garment damage detection, and particularly relates to a garment damage detection method and system based on a flexible electric wire grid. BACKGROUND
[0002] Special protective clothing is extremely prone to damage due to wear and tear, tearing, puncture, burning, etc. during task execution or daily use. If these damages are not discovered and repaired in time, not only will it seriously affect the service life of the clothing, but more importantly, it will greatly weaken its protective performance, directly threatening the personal safety of the wearer. Therefore, it is of great practical significance and application value to develop efficient and reliable automatic garment damage detection technology.
[0003] Traditional garment damage detection mainly relies on two traditional methods. One is manual visual inspection, that is, the wearer or professional personnel observes the surface of the garment for damage, stains or deformation by naked eye. This method is highly dependent on the experience and attention of the inspector, and has problems such as low efficiency, strong subjectivity and difficulty in reproduction, especially for damages occurring in the inner layer, seams or small size of the garment.
[0004] The second is a simple physical test, for example, through air tightness test, water pressure test or tensile strength sampling test at specific positions to indirectly evaluate the overall integrity of the garment. Although this method has a certain objectivity, it is usually not real-time in daily use, the detection process is tedious and destructive, and it is difficult to achieve precise positioning of the damage.
[0005] Therefore, the present application provides a garment damage detection method and system based on a flexible electric wire grid, which can detect and evaluate garment damage in real time, automatically and accurately, to improve the support level of equipment and the safety factor of personnel. SUMMARY
[0006] In order to solve the above technical problems, the purpose of the present application is to provide a garment damage detection method and system based on a flexible electric wire grid, which can detect and evaluate garment damage in real time, automatically and accurately, to improve the support level of equipment and the safety factor of personnel.
[0007] In order to achieve the above purpose, the present application provides the following technical solution: a garment damage detection method based on a flexible electric wire grid, the method comprising: Based on the flexible electric wire grid pre-set in the inner layer of the garment, the multi-modal data corresponding to each grid node is synchronously collected, the collected multi-modal data is subjected to bias correction and combination preprocessing, and the preprocessed multi-dimensional data is obtained. Based on the preprocessed multi-dimensional data, feature signals are constructed, and time-domain features, frequency-domain features, and the rate of change of feature signals are extracted from the feature signals as basic features. The basic features are dynamically weighted and fused using an attention mechanism, and the fused features are then subjected to dimensionality reduction processing to obtain the core damage feature set. Based on the core damage feature set and the pre-trained damage quantification model, damage quantification assessment results are output.
[0008] Preferably, the bias correction of the acquired multimodal data includes: The sensor data acquired continuously under non-destructive conditions were selected as the initial dataset. Calculate the sum of squares of various types of sensor data for each grid node in the initial dataset to form a node data sum of squares vector; The normalization standard is calculated based on the sum of squares of the node data vector, and the corresponding bias correction coefficient is determined based on the normalization standard and the sum of squares of each node. The bias correction coefficient is used to correct all multimodal data of the corresponding node to eliminate the initial sensing differences between nodes.
[0009] Preferably, the combined preprocessing of the acquired multimodal data includes: Based on the bias-corrected multimodal data, gain-adjusted data is obtained through adaptive gain adjustment; Based on the gain-adjusted data, the denoised data is obtained by improving wavelet threshold denoising. Based on the denoised data, smoothed data is obtained by empirical mode decomposition smoothing. Based on the smoothed data, standardized data with unified dimensions is obtained through Z-Score standardization, which serves as the preprocessed multidimensional data.
[0010] Preferably, constructing feature signals based on preprocessed multi-dimensional data includes: Based on the standardized data of each category, at each sampling time, extract the maximum, minimum and median values of all grid node data at that time. Based on the maximum, minimum and median values at each sampling time, the maximum value vector, minimum value vector and median vector are constructed in chronological order. Based on the maximum value vector, minimum value vector, and median vector corresponding to each type of standardized data, and based on the preset feature signal calculation formula, a feature signal corresponding to each type of standardized data is generated.
[0011] Preferably, the step of extracting time-domain features, frequency-domain features, and the rate of change of the feature signal as basic features from the feature signal includes: Based on the aforementioned feature signals, calculations are performed on the time series to extract time-domain features; Time-frequency analysis is performed based on the feature signal, and it is converted into the frequency domain by short-time Fourier transform, thereby extracting frequency domain features; Based on the characteristic signal, the numerical change of the characteristic signal at consecutive sampling times is calculated along the time series to obtain the rate of change of the characteristic signal.
[0012] Preferably, the step of dynamically weighting and fusing the basic features using an attention mechanism, and then performing dimensionality reduction on the fused features to obtain the core damage feature set includes: Based on a pre-set sample dataset with damage labels, the correlation measure between each basic feature and the damage label is calculated in the basic feature set consisting of the time-domain features, frequency-domain features, and the rate of change of the feature signals. Based on the correlation metric, a weight coefficient is assigned to each basic feature; After standardizing the basic features based on the assigned weight coefficients, a high-dimensional fusion feature vector is generated through linear combination. Based on preset temporal correlation constraints and local linear embedding algorithms, the high-dimensional fused feature vector is subjected to dimensionality reduction processing to generate a low-dimensional feature set, which serves as the corresponding core damage feature set.
[0013] Preferably, the method further includes: Based on the flexible electric wire mesh, a two-dimensional reference coordinate system is constructed, and the physical position of each mesh node of the flexible electric wire mesh is mapped to the absolute coordinates in the two-dimensional reference coordinate system. Based on the core damage feature set and the preprocessed multi-dimensional data of each grid node, feature vector similarity matching is performed to determine the matching coefficient corresponding to each grid node. Based on the matching coefficients corresponding to each grid node, construct the corresponding diagonal weighted matrix; The coordinates of each grid node are used to form a coordinate vector. The coordinates of the damaged area are then calculated by fusing the coordinate vector with a diagonal weighted matrix.
[0014] A second aspect of the present invention also provides a clothing damage detection system based on a flexible electrical wire mesh, comprising: The acquisition module, based on a flexible electric wire mesh pre-set in the inner layer of the garment, synchronously acquires multimodal data corresponding to each mesh node, performs bias correction and combination preprocessing on the acquired multimodal data, and obtains preprocessed multidimensional data. The feature extraction module constructs feature signals based on preprocessed multi-dimensional data, extracts time-domain features, frequency-domain features, and the rate of change of feature signals as basic features from the feature signals, uses an attention mechanism to dynamically weight and fuse the basic features, and performs dimensionality reduction processing on the fused features to obtain the core damage feature set. The evaluation module, based on the core damage feature set and a pre-trained damage quantification model, outputs damage quantification evaluation results data.
[0015] Compared with the prior art, the beneficial effects of the present invention are: By synchronously acquiring multimodal data from each node using a flexible wire mesh, a comprehensive physical information source is provided for damage identification. Furthermore, through a "bias correction and combined preprocessing" technical workflow, the initial data deviation and non-stationary signal interference caused by individual sensor differences, environmental noise, and wearer dynamics are systematically eliminated for the first time in the industry, generating unified, clean, and standardized multidimensional data. This fundamentally overcomes the high false alarm problem caused by reference drift and noise sensitivity in traditional flexible sensing systems, ensuring the high reliability of the basis for subsequent analysis.
[0016] This invention constructs feature signals from preprocessed data and extracts fundamental features from three complementary dimensions: time domain, frequency domain, and rate of change, enabling a more comprehensive capture of the dynamic patterns during damage occurrence. Furthermore, an attention mechanism is introduced to dynamically weight and fuse these fundamental features, allowing the model to automatically focus on the key features most relevant to the current damage, suppressing redundancy and resulting in a highly discriminative fused feature set. Finally, dimensionality reduction is used to generate a compact, low-redundancy "core damage feature set." This series of operations enables the invention to intelligently extract and condense the essential characteristics of damage from massive amounts of raw data, improving the input quality of subsequent damage quantification models and reducing their computational complexity, thus providing the technical prerequisite for real-time, automated damage analysis.
[0017] The data is input into a pre-trained damage quantification model, which, through learning from a large amount of sample data, can establish a precise mapping from complex feature patterns to specific damage types and degrees. The entire process of "data acquisition, preprocessing, feature extraction, and model evaluation" is highly automated, overcoming the problems of traditional methods that rely on manual labor and are highly subjective. It can quickly, objectively, and accurately output damage quantification assessment results, achieving real-time, automatic, and accurate detection and assessment of clothing damage. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0019] Figure 1 This is a schematic diagram of a clothing damage detection method based on flexible electric wire mesh.
[0020] Figure 2 This is a schematic diagram of a clothing damage detection system based on a flexible electric wire mesh. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0023] Example 1 like Figure 1 As shown in the figure, this embodiment discloses a method for detecting clothing damage based on a flexible electrical wire mesh. The method includes: Based on the flexible electric wire mesh pre-set in the inner layer of the garment, multimodal data corresponding to each mesh node are collected synchronously. The collected multimodal data is subjected to bias correction and combination preprocessing to obtain preprocessed multidimensional data. It should be noted that the bias correction of the acquired multimodal data includes: The sensor data acquired continuously under non-destructive conditions were selected as the initial dataset. Calculate the sum of squares of various types of sensor data for each grid node in the initial dataset to form a node data sum of squares vector; The normalization standard is calculated based on the sum of squares of the node data vector, and the corresponding bias correction coefficient is determined based on the normalization standard and the sum of squares of each node. The bias correction coefficient is used to correct all multimodal data of the corresponding node to eliminate the initial sensing differences between nodes.
[0024] In this embodiment, n sets of continuously acquired sensor data under non-damage conditions are selected as the initial dataset (n is the preset number of initial data). This dataset contains four types of multimodal data (dynamic resistance change of grid nodes, distributed capacitance change of nodes to ground, strain tensor signal of grid elements, and damage electrical noise spectrum signal) of m grid nodes in the flexible wire mesh (m is the total number of grid nodes). Calculate the sum of squares of the node data for each grid node. ( For each type of sensor data, calculate the sum of squares of its value across the n initial data sets. : In the formula, Let be the multimodal sensing data value of a certain type at the j-th sampling point of the i-th node; The sum of squares of the same type of sensor data from m nodes is calculated in this way, forming a vector of node data sum of squares. : ; Calculate the normalization standard based on the sum of squares of the node data vector A. : ; Based on normalization standard Sum of squares of each grid node Determine the corresponding bias correction coefficient. : ; Using the above bias correction coefficients Correct all multimodal data of the corresponding node i to obtain the bias-corrected multimodal data of node i. The correction formula is: ;in, This represents the original multimodal data for grid node i. The correction of the four types of multimodal data from all m nodes is performed in this manner, eliminating the initial sensing differences between nodes.
[0025] It should be noted that the combined preprocessing of the collected multimodal data includes: Based on the bias-corrected multimodal data, gain-adjusted data is obtained through adaptive gain adjustment; Based on the gain-adjusted data, the denoised data is obtained by improving wavelet threshold denoising. Based on the denoised data, smoothed data is obtained by empirical mode decomposition smoothing. Based on the smoothed data, standardized data with unified dimensions is obtained through Z-Score standardization, which serves as the preprocessed multidimensional data.
[0026] In this embodiment, an adaptive gain control (AGC) algorithm is used to match signals of different amplitudes in real time. The gain coefficient is dynamically adjusted through a preset amplitude threshold range. The gain is amplified for weak damaged signals with amplitudes below the threshold and attenuated for saturated signals with amplitudes above the threshold. This ensures that the amplitudes of all signals are within the preset effective processing range, preventing weak damaged signals from being overwhelmed by subsequent processing steps, and obtaining data after gain adjustment. Based on the gain-adjusted data, an improved wavelet thresholding denoising algorithm with an exponential decay term is employed. First, the signal undergoes multi-scale wavelet decomposition to separate high-frequency noise components from low-frequency effective components. Then, an exponential decay term is introduced to optimize the threshold function. Compared to traditional hard / soft thresholding, this reduces distortion at signal abrupt changes, performs threshold filtering on high-frequency noise components, and preserves the signal abrupt change characteristics caused by damage in the low-frequency effective components, resulting in denoised data. Based on the denoised data, an improved wavelet thresholding algorithm with an exponential decay term is used. The algorithm achieves accurate separation of the signal trend term and residual noise term. It constructs upper and lower envelopes by selecting local extrema of the signal, calculates the mean of the envelopes, and subtracts it from the original signal to obtain the intrinsic mode function (EMF). Repeat this process until the remaining signal exhibits a monotonic trend, eliminating high-frequency components containing residual noise. Components, retaining low-frequency components containing core damage information The components and trend terms are reconstructed to obtain smoothed data. Based on the smoothed data, the Z-Score normalization algorithm is used to unify data of different dimensions. Using the overall mean and standard deviation of various types of multimodal data as a benchmark, the numerical range deviation caused by the difference in dimensions of different types of signals (such as resistance change and capacitive coupling strength) is eliminated. All data are uniformly mapped to the [0,1] normalization interval to ensure the fairness of the weight ratio of various types of data in the subsequent feature extraction stage, and the normalized data is obtained. This normalized data is the preprocessed multidimensional data.
[0027] Based on the preprocessed multi-dimensional data, feature signals are constructed. It should be noted that constructing feature signals based on preprocessed multi-dimensional data includes: Based on the standardized data of each category, at each sampling time, extract the maximum, minimum and median values of all grid node data at that time. Based on the maximum, minimum and median values at each sampling time, the maximum value vector, minimum value vector and median vector are constructed in chronological order. Based on the maximum value vector, minimum value vector, and median vector corresponding to each type of standardized data, and based on the preset feature signal calculation formula, a feature signal corresponding to each type of standardized data is generated.
[0028] In this embodiment, for each type of standardized data, at each sampling time... ( (where n is the total number of sampling points), extract the maximum value from all m grid node data at that moment. Minimum value With median ,in The set of all node sample values for this type of data at the t-th sampling point. , This represents the standardized data value of this type at the t-th sampling point of the k-th node.
[0029] Based on the maximum, minimum, and median values at each sampling time, vectors representing the maximum, minimum, and median values are constructed sequentially over time: Maximum vector ; Minimum value vector ; Median vector ; The preset feature signal calculation formula is as follows: ;in, , , For the corresponding preset weight coefficients, + + =1; Extract time-domain features, frequency-domain features, and the rate of change of the feature signal as basic features from the feature signal; It should be noted that the extraction of time-domain features, frequency-domain features, and the rate of change of the feature signal as basic features from the feature signal includes: Based on the aforementioned feature signals, calculations are performed on the time series to extract time-domain features; Time-frequency analysis is performed based on the feature signal, and it is converted into the frequency domain by short-time Fourier transform, thereby extracting frequency domain features; Based on the characteristic signal, the numerical change of the characteristic signal at consecutive sampling times is calculated along the time series to obtain the rate of change of the characteristic signal.
[0030] In this embodiment, based on the feature signal The peak value, peak-to-peak value, mean, variance, kurtosis, waveform factor, and impulse factor are calculated and extracted from the time series as time-domain features; time-frequency analysis is performed based on the feature signal S, using a short-time Fourier transform with a 50% overlap sliding time window. The time-domain signal is converted into a frequency-domain signal, and the center frequency, peak frequency, spectral energy, and bandwidth entropy are extracted as frequency-domain features. Based on the feature signal S, the numerical change at consecutive sampling times is calculated along the time series. The formula for calculating the rate of change of the feature signal is: ,in, The preset sliding window length is n, where n is the total number of sampling points.
[0031] The basic features are dynamically weighted and fused using an attention mechanism, and the fused features are then subjected to dimensionality reduction to obtain a core damage feature set. It should be noted that the dynamic weighted fusion of the basic features using the attention mechanism, followed by dimensionality reduction of the fused features, yields the core damage feature set, which includes: Based on a pre-set sample dataset with damage labels, the correlation measure between each basic feature and the damage label is calculated in the basic feature set consisting of the time-domain features, frequency-domain features, and the rate of change of the feature signals. Based on the correlation metric, a weight coefficient is assigned to each basic feature; After standardizing the basic features based on the assigned weight coefficients, a high-dimensional fusion feature vector is generated through linear combination. Based on preset temporal correlation constraints and local linear embedding algorithms, the high-dimensional fused feature vector is subjected to dimensionality reduction processing to generate a low-dimensional feature set, which serves as the corresponding core damage feature set.
[0032] In this embodiment, a sample dataset with damage labels is pre-constructed (this dataset contains basic feature samples and corresponding damage labels under different damage types and degrees). The basic feature set (including time-domain features, frequency-domain features, and feature signal change rates) is then used to... Each basic feature in ) ( , Calculate the Pearson correlation coefficient between the total number of basic features and the damage label. Then, the Pearson correlation coefficient was normalized to form the attention weight coefficients for each basic feature.
[0033] When generating high-dimensional fusion features, the basic features after assigning weight coefficients are first processed. The standardization process maps all basic feature values to the [0,1] interval to eliminate the influence of numerical range differences. Then, based on the weight coefficients and standardized values of each basic feature, a high-dimensional fusion feature vector F is generated through linear combination.
[0034] In detail, based on preset temporal correlation constraints and a local linear embedding algorithm, the dimensionality reduction of the high-dimensional fused feature vector is performed as follows: First, the number of neighbors for each high-dimensional fused feature vector is determined based on a preset K-nearest neighbor adaptive selection mechanism (the local density of the feature vector space is calculated, and the number of neighbors is dynamically set based on the local density, which is calculated by the reciprocal of the mean of the K-nearest neighbor distances of the feature vectors). Second, when constructing the local neighborhood of each feature vector, a temporal correlation constraint is introduced, namely, the Euclidean distance between fused feature vectors at adjacent sampling times is constrained to be ≤ a threshold, to ensure that the temporal correlation of features is preserved during the dimensionality reduction process (avoiding the loss of temporal change information of the damaged signal). Finally, through the core steps of neighborhood reconstruction weight calculation and low-dimensional embedding optimization of the LLE algorithm, the K-dimensional high-dimensional fused feature vector is mapped to a low-dimensional space.
[0035] Based on the core damage feature set and a pre-trained damage quantification model, damage quantification assessment results are output. In this embodiment, the damage quantification model can be a neural network model, such as a hybrid model of an improved lightweight convolutional neural network and a long short-term memory network, or a backpropagation neural network model.
[0036] The method further includes: Based on a flexible electric wire mesh, a two-dimensional reference coordinate system is constructed, mapping the physical position of each mesh node of the flexible electric wire mesh to absolute coordinates within the two-dimensional reference coordinate system. ; , This represents the total number of grid nodes.
[0037] Based on the core damage feature set and the preprocessed multi-dimensional data of each grid node, feature vector similarity matching is performed to determine the matching coefficient corresponding to each grid node. Based on the matching coefficients corresponding to each grid node, construct the corresponding diagonal weighted matrix; The coordinates of each grid node are used to form a coordinate vector. The coordinates of the damaged area are then calculated by fusing the coordinate vector with a diagonal weighted matrix.
[0038] In this embodiment, the matching coefficient is calculated using the cosine similarity algorithm. .
[0039] Match coefficients of all nodes Construct as a diagonal weighted matrix The matrix dimension is The diagonal elements are the matching coefficients of each node. Off-diagonal elements are 0, that is: Construct a coordinate vector from the absolute coordinates of each node. The precise coordinates of the damaged area are obtained by fusing the weighted matrix and coordinate vector. The corresponding calculation formula is: ; ; like Figure 2 As shown, this embodiment also discloses a clothing damage detection system based on a flexible electric wire mesh, comprising: The acquisition module, based on a flexible electric wire mesh pre-set in the inner layer of the garment, synchronously acquires multimodal data corresponding to each mesh node, performs bias correction and combination preprocessing on the acquired multimodal data, and obtains preprocessed multidimensional data. The feature extraction module constructs feature signals based on preprocessed multi-dimensional data, extracts time-domain features, frequency-domain features, and the rate of change of feature signals as basic features from the feature signals, uses an attention mechanism to dynamically weight and fuse the basic features, and performs dimensionality reduction processing on the fused features to obtain the core damage feature set. The evaluation module, based on the core damage feature set and a pre-trained damage quantification model, outputs damage quantification evaluation results data.
[0040] Optionally, in this embodiment, those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.
[0041] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0042] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more electronic devices to execute all or part of the steps of the methods described in the various embodiments of this application.
[0043] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0044] In the several embodiments provided in this application, it should be understood that the disclosed application can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.
[0045] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0046] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0047] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for detecting clothing damage based on flexible electrical wire mesh, characterized in that, The method includes: Based on the flexible electric wire mesh pre-set in the inner layer of the garment, multimodal data corresponding to each mesh node are collected synchronously. The collected multimodal data is subjected to bias correction and combination preprocessing to obtain preprocessed multidimensional data. Based on the preprocessed multi-dimensional data, feature signals are constructed, and time-domain features, frequency-domain features, and the rate of change of feature signals are extracted from the feature signals as basic features. The basic features are dynamically weighted and fused using an attention mechanism, and the fused features are then subjected to dimensionality reduction processing to obtain the core damage feature set. Based on the core damage feature set and the pre-trained damage quantification model, damage quantification assessment results are output.
2. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 1, characterized in that, The bias correction of the acquired multimodal data includes: The sensor data acquired continuously under non-destructive conditions were selected as the initial dataset. Calculate the sum of squares of various types of sensor data for each grid node in the initial dataset to form a node data sum of squares vector; A normalization standard is calculated based on the sum of squares of the node data vector, and a corresponding bias correction coefficient is determined based on the normalization standard and the sum of squares of each node. The bias correction coefficient is then used to correct all multimodal data of the corresponding node to eliminate initial sensing differences between nodes.
3. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 2, characterized in that, The combined preprocessing of the acquired multimodal data includes: Based on the bias-corrected multimodal data, gain-adjusted data is obtained through adaptive gain adjustment; Based on the gain-adjusted data, the denoised data is obtained by improving wavelet threshold denoising. Based on the denoised data, smoothed data is obtained by empirical mode decomposition smoothing. Based on the smoothed data, standardized data with unified dimensions is obtained through Z-Score standardization, which serves as the preprocessed multidimensional data.
4. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 3, characterized in that, Constructing feature signals based on preprocessed multi-dimensional data includes: Based on the standardized data of each category, at each sampling time, extract the maximum, minimum and median values of all grid node data at that time. Based on the maximum, minimum and median values at each sampling time, the maximum value vector, minimum value vector and median vector are constructed in chronological order. Based on the maximum value vector, minimum value vector, and median vector corresponding to each type of standardized data, and based on the preset feature signal calculation formula, a feature signal corresponding to each type of standardized data is generated.
5. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 4, characterized in that, The extraction of time-domain features, frequency-domain features, and the rate of change of the feature signal as basic features from the feature signal includes: Based on the aforementioned feature signals, calculations are performed on the time series to extract time-domain features; Time-frequency analysis is performed based on the feature signal, and it is converted into the frequency domain by short-time Fourier transform, thereby extracting frequency domain features; Based on the characteristic signal, the numerical change of the characteristic signal at consecutive sampling times is calculated along the time series to obtain the rate of change of the characteristic signal.
6. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 5, characterized in that, The process of dynamically weighting and fusing the basic features using an attention mechanism, followed by dimensionality reduction of the fused features, yields a core damage feature set including: Based on a pre-set sample dataset with damage labels, the correlation measure between each basic feature and the damage label is calculated in the basic feature set consisting of the time-domain features, frequency-domain features, and the rate of change of the feature signals. Based on the correlation metric, a weight coefficient is assigned to each basic feature; After standardizing the basic features based on the assigned weight coefficients, a high-dimensional fusion feature vector is generated through linear combination. Based on preset temporal correlation constraints and local linear embedding algorithms, the high-dimensional fused feature vector is subjected to dimensionality reduction processing to generate a low-dimensional feature set, which serves as the corresponding core damage feature set.
7. The method for detecting clothing damage based on a flexible electrical wire mesh according to claim 6, characterized in that, The method further includes: Based on the flexible electric wire mesh, a two-dimensional reference coordinate system is constructed, and the physical position of each mesh node of the flexible electric wire mesh is mapped to the absolute coordinates in the two-dimensional reference coordinate system. Based on the core damage feature set and the preprocessed multi-dimensional data of each grid node, feature vector similarity matching is performed to determine the matching coefficient corresponding to each grid node. Based on the matching coefficients corresponding to each grid node, construct the corresponding diagonal weighted matrix; The coordinates of each grid node are used to form a coordinate vector. The coordinates of the damaged area are then calculated by fusing the coordinate vector with a diagonal weighted matrix.
8. A clothing damage detection system based on a flexible electric wire mesh, implementing the clothing damage detection method based on a flexible electric wire mesh as described in any one of claims 1 to 7, characterized in that, include: The acquisition module, based on a flexible electric wire mesh pre-set in the inner layer of the garment, synchronously acquires multimodal data corresponding to each mesh node, performs bias correction and combination preprocessing on the acquired multimodal data, and obtains preprocessed multidimensional data. The feature extraction module constructs feature signals based on preprocessed multi-dimensional data, extracts time-domain features, frequency-domain features, and the rate of change of feature signals as basic features from the feature signals, uses an attention mechanism to dynamically weight and fuse the basic features, and performs dimensionality reduction processing on the fused features to obtain the core damage feature set. The evaluation module, based on the core damage feature set and a pre-trained damage quantification model, outputs damage quantification evaluation results data.