A glass curtain wall damage positioning method and system based on binoculars and wavelet analysis

Abstract

The application provides a glass curtain wall damage positioning method and system based on binoculars and wavelet analysis, and belongs to the field of building structure health monitoring, and is particularly suitable for structural adhesive damage positioning of over-service hidden frame glass curtain walls. The application first collects curtain wall vibration video sequences through binocular cameras, and enhances micro-vibration features by using phase motion amplification technology; then, sub-pixel level displacement tracking is realized by using an adaptive KLT algorithm, and a three-dimensional displacement field is constructed in combination with stereo calibration parameters; secondly, wavelet packet decomposition is performed on the displacement signal, energy features of multiple frequency bands are extracted, a mean energy vector is calculated, and a health benchmark model is generated; thirdly, a quantitative damage index of the to-be-tested signal is calculated based on the health benchmark model, and a spatial distribution cloud diagram of the quantitative damage index is obtained; finally, a position area of the curtain wall structural adhesive damage is obtained according to the spatial distribution cloud diagram of the quantitative damage index and a damage threshold.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of building structure health monitoring, and particularly relates to a glass curtain wall damage positioning method and system based on binoculars and wavelet analysis. BACKGROUND

[0002] As the mainstream facade form of modern buildings, the hidden frame glass curtain wall has a core safety risk that the load bearing between the glass panel and the metal frame completely relies on the structural adhesive bonding, and the physical properties of the silicone structural adhesive will irreversibly degrade over time. A large number of accident cases show that when the structural adhesive bonding area loss reaches 5%, the shear strength decreases by more than 40%; if the damage area exceeds 10%, the risk of curtain wall panel falling will increase exponentially.

[0003] In the measurement aspect, although traditional contact sensors such as piezoelectric accelerometers can obtain structural vibration responses, their deployment process needs to destroy the air tightness of the curtain wall - each measurement point needs to drill a fixed hole with a diameter of 8 mm, resulting in at least 96 square centimeters of permanent damage to a single standard curtain wall unit (1.2 m x 1.5 m); more seriously, such devices have insufficient sensitivity to micro-vibration less than 0.2 mm, and their signal-to-noise ratio is often less than 4 decibels in the key frequency band (1-5 Hz), which cannot capture the characteristic signals of early structural adhesive damage. In the analysis method aspect, the detection technology based on frequency response has a principle defect: experimental data show that when the structural adhesive damage area is less than 3%, the inherent frequency offset of the curtain wall system is only 0.07% to 0.12%, while the frequency fluctuation caused by environmental temperature changes can reach 0.25%, and the noise completely overwhelms the damage signal. Even the binocular vision detection method that has emerged in recent years encounters insurmountable technical barriers in practical application: the root mean square error of the conventional displacement measurement algorithm in the curtain wall micro-vibration monitoring is as high as 0.58 mm, which is more than 5 times the safety threshold; the feature point tracking failure rate exceeds 30% during overcast or dusk periods (illuminance less than 300 lux); for complex conditions such as bilateral symmetric damage, the misjudgment rate of existing solutions even reaches 41%. The technical bottlenecks of the binocular vision detection method lie in two unsolved core problems: first, the stable capture of sub-pixel level micro-vibration, the displacement amplitude of the curtain wall under wind load or human-induced vibration is usually less than 0.1 mm, corresponding to an image displacement of less than 0.5 pixels, and the conventional algorithm completely fails under the interference of image noise and motion blur; second, the specific extraction of damage features, the vibration energy change caused by structural adhesive damage is concentrated in a specific frequency band (12-25 Hz), but the energy in this frequency band accounts for only 8% to 12% of the total vibration energy, and the traditional frequency spectrum analysis method cannot effectively separate it from the environmental vibration noise. This leads to the lack of effective detection means for damage areas below 5% so far, forming a major public safety hazard of "sick running" of high-risk curtain walls. SUMMARY

[0004] To solve the above problems in the prior art, the application provides a glass curtain wall damage positioning method and system based on binoculars and wavelet analysis. First, the binocular camera is used to collect the curtain wall vibration video sequence, and the phase motion amplification technology is used to enhance the micro-vibration characteristics. Then, the adaptive KLT algorithm is used to realize sub-pixel level displacement tracking, and a three-dimensional displacement field is constructed by combining the stereo calibration parameters. Secondly, the displacement signal is decomposed by wavelet packet, the energy features of multiple frequency bands are extracted, the mean energy vector is calculated, and the health benchmark model is generated. Thirdly, the quantitative damage index of the to-be-measured signal is calculated based on the health benchmark model, and the spatial distribution cloud map of the quantitative damage index is obtained. Finally, the damage threshold is set, and the position and area boundary of the curtain wall structure adhesive damage are obtained through damage gradient detection. To achieve the above purpose, the technical scheme is as follows:

[0005] On the one hand, the application provides a glass curtain wall damage positioning method based on binoculars and wavelet analysis, which comprises the following steps:

[0006] S1, two cameras are arranged on both sides of the glass curtain wall, the cameras are adjusted so that the optical axes of the cameras are parallel to each other and perpendicular to the plane where the glass curtain wall is located, and the vibration video sequence of the surface of the glass curtain wall is collected in a synchronous triggering mode;

[0007] S2, according to the vibration video sequence of the curtain wall surface, the enhanced vibration video sequence is obtained through the phase motion amplification technology;

[0008] S3, according to the enhanced vibration video sequence, the feature points are selected, and the high-precision displacement measurement sequence of the feature points is obtained through the adaptive KLT algorithm;

[0009] S4, according to the high-precision displacement measurement sequence of the feature points, the three-dimensional displacement field of the curtain wall surface is obtained through the three-dimensional displacement reconstruction algorithm;

[0010] S5, according to the three-dimensional displacement field of the curtain wall surface, the energy sequence of each frequency band is extracted through wavelet packet energy decomposition, the energy mean vector representing the damage characteristics is calculated, and the health benchmark model is obtained;

[0011] S6, according to the health benchmark model, the damage index spatial distribution cloud map is obtained through the damage quantification algorithm;

[0012] S7, according to the damage index spatial distribution cloud map, the position and area of the glass curtain wall structure adhesive damage are obtained by setting the damage threshold.

[0013] Optionally, in the S2, the enhanced vibration video sequence is obtained through the phase motion amplification technology according to the vibration video sequence of the curtain wall surface, which comprises the following steps:

[0014] S21, obtaining a multi-layer sub-band image pyramid through Laplacian pyramid decomposition according to the vibration video sequence of the curtain surface;

[0015] S22, extracting a target level sub-band image through setting a target frequency band [f L ,f H ] according to the multi-layer sub-band image pyramid,

[0016] wherein f L is a low frequency cutoff frequency of a filter, and f H is a high frequency cutoff frequency of the filter;

[0017] S23, obtaining a second stage target level sub-band image through the target frequency band [f L ,f H ] according to the target level sub-band image;

[0018] S24, obtaining a motion phase difference through same layer comparison of continuous frames of the vibration video sequence of the curtain surface according to the second stage target level sub-band image;

[0019] S25, obtaining an amplified image value through formula (1) according to the motion phase difference;

[0020] I out (x,y,t)=I in (x,y,t+β·Δφ(x,y,t)) (1)

[0021] wherein I in is an image pixel value of a current frame in the continuous frames, I oit is an image pixel value of a next frame in the continuous frames, x is a horizontal coordinate, y is a vertical coordinate, t is a time sequence, Δφ is a motion phase difference corresponding to a spatial position change between the continuous frames at coordinate (x,y) at time t, and β is an amplification ratio.

[0022] S26, obtaining a third stage target level sub-band image through superimposing back the second stage target level sub-band image according to the amplified image value;

[0023] S27, obtaining an enhanced vibration video sequence through pyramid reconstruction according to the third stage target level sub-band image.

[0024] Optionally, the S3 comprises: selecting a feature point according to the enhanced vibration video sequence, and obtaining a high precision displacement measurement sequence of the feature point through an adaptive KLT algorithm, comprising:

[0025] S31, selecting a feature point according to the enhanced vibration video sequence, and obtaining a vibration video sequence with feature point information;

[0026] S32, according to the feature point information with the video sequence, through the adaptive KLT algorithm tracking the feature point, get the high precision displacement measurement sequence of feature points.

[0027] Optionally, the three-dimensional displacement reconstruction algorithm comprises:

[0028]

[0029] d = u l -u r (3)

[0030] Wherein, B is the baseline distance, f is the focal length, is the parallax, u l is the horizontal coordinate of the feature point in the left camera, u r is the horizontal coordinate of the feature point in the right camera, d is the parallax of the feature point, v l is the vertical coordinate of the feature point in the left camera, c x is the horizontal position of the left camera principal point in the image coordinate system, c y is the vertical position of the left camera principal point in the image coordinate system, Z is the vibration displacement of the feature point, X is the horizontal coordinate of the feature point in the camera coordinate system, Y is the vertical coordinate of the feature point in the camera coordinate system.

[0031] Optionally, S5 according to the three-dimensional displacement field of the curtain wall surface, through wavelet packet energy decomposition, extract the energy sequence of each frequency band, calculate the energy mean vector representing the damage characteristics, get the health reference model, comprising:

[0032] S51, according to the three-dimensional displacement field of the curtain wall surface, through wavelet packet energy decomposition formula (4), get the coefficient sequence vector of different frequency bands,

[0033] Wherein,

[0034] X j =[X j1 ,X j2 ,…X jM ] (4)

[0035] Wherein, X j represent the coefficient sequence vector of the jth frequency band range, M represents the coefficient length, j represents the jth frequency band range.

[0036] S52, according to the coefficient sequence vector of different frequency bands, through formula (5), calculate the energy sequence vector of each frequency band

[0037]

[0038] Wherein, M represents the coefficient length, j represents the jth frequency band range, X jia sequence vector representing the i-th coefficient of the j-th frequency band, representing the energy value of the j-th frequency band, N representing the number of displacement signal groups of the three-dimensional displacement field of the curtain wall surface;

[0039] S53, according to the energy sequence vector of each frequency band, the energy mean value of each frequency band is calculated by formula (6), and the energy mean value vector representing the damage characteristics μ=[μ1, μ2, … μN] is obtained. j

[0040]

[0041] wherein, representing the energy value of the j-th frequency band of the k-th group of displacement signals of the three-dimensional displacement field, μ j is the energy mean value of the j-th frequency band;

[0042] S54, according to the energy mean value vector representing the damage characteristics, N≥50 groups of three-dimensional displacement fields of the curtain wall surface in the healthy state of the glass curtain wall are obtained, and the energy mean value vector representing the damage characteristics in the healthy state is calculated by repeating S51-S53, to obtain a healthy baseline model.

[0043] Optionally, S6 obtains a damage index spatial distribution cloud map according to the healthy baseline model by a damage quantification algorithm, including:

[0044] S61, according to the healthy baseline model, a quantified damage index SSD is obtained by formula (7),

[0045]

[0046] wherein, SSD represents the quantified damage index, n represents the total number of frequency bands after wavelet decomposition, j represents the j-th frequency band, representing the energy value of the j-th frequency band collected, j representing the energy mean value of the j-th frequency band in the healthy state of the glass curtain wall;

[0047] S62, according to the quantified damage index, a damage index spatial distribution cloud map is obtained by spatial calculation.

[0048] On the other hand, the application provides a glass curtain wall damage positioning system based on binoculars and wavelet analysis, which is applied to a glass curtain wall damage positioning method based on binoculars and wavelet analysis, and the system comprises:

[0049] A binocular vision acquisition module is used to deploy two cameras on both sides of the glass curtain wall, adjust the cameras so that the optical axes of the cameras are parallel to each other and perpendicular to the plane where the glass curtain wall is located, and acquire the vibration video sequence of the glass curtain wall surface by using a synchronous triggering mode. ​

[0050] a phase motion amplification module, configured to obtain an enhanced vibration video sequence through a phase motion amplification technique according to the vibration video sequence of the curtain wall surface;

[0051] a displacement tracking module, configured to select feature points and obtain a high-precision displacement measurement sequence of the feature points through an adaptive KLT algorithm according to the enhanced vibration video sequence;

[0052] a three-dimensional displacement reconstruction module, configured to obtain a three-dimensional displacement field of the curtain wall surface through a three-dimensional displacement reconstruction algorithm according to the high-precision displacement measurement sequence of the feature points;

[0053] a wavelet packet analysis module, configured to extract an energy sequence of each frequency band, calculate an energy mean vector representing damage characteristics, and obtain a health benchmark model through wavelet packet energy decomposition according to the three-dimensional displacement field of the curtain wall surface;

[0054] a damage positioning module, configured to obtain a damage index spatial distribution cloud map through a damage quantification algorithm according to the health benchmark model;

[0055] a position determination module, configured to obtain a position area of the structural adhesive damage of the glass curtain wall structure through setting a damage threshold according to the damage index spatial distribution cloud map.

[0056] Compared with the prior art, the technical scheme of the present application has at least the following beneficial effects:

[0057] (1) The above scheme can realize high-precision measurement of the micro-vibration signal of the hidden-frame glass curtain wall by adopting a binocular vision system combined with a phase motion amplification technique and a sub-pixel level displacement tracking algorithm, and the displacement measurement precision can reach 0.14 mm (RMSE), which effectively avoids physical damage to the curtain wall structure in the installation process of the traditional contact sensor, reduces the detection cost, and at the same time, the damage identification model based on wavelet packet energy analysis can sensitively identify early structural adhesive damage, and when the SSD value exceeds 0.001, the damage can be determined, which has significant identification ability for early damage with a damage area less than 5%, significantly improves the safety and reliability of the curtain wall, and further has high-precision damage positioning capability, with a positioning error of only 0.32% under single-side damage working conditions and a maximum positioning error of 14.287% under double-side damage working conditions, which can accurately position the damage area through the SSD spatial distribution cloud map, and provides strong support for curtain wall maintenance.

[0058] (2) The above scheme can keep stable operation under different ambient light conditions (300-1000 lux), the correlation coefficient R2 of displacement measurement and the laser displacement meter is 0.8719, which fully meets the engineering detection requirements, and the stability and applicability of the above scheme make it not only suitable for laboratory environment, but also can be reliably applied in actual engineering scene, which provides strong technical support for safety evaluation of over-service hidden frame glass curtain wall, effectively solves the major defects of existing technology in early damage identification and positioning, and significantly improves the efficiency and accuracy of curtain wall damage detection. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0060] Figure 1 is a flowchart of the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0061] Figure 2 is a schematic diagram of the arrangement of binocular vision in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0062] Figure 3 is a camera shooting calibration board and test glass curtain wall diagram in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0063] Figure 4 is a flowchart of enhancing micro-vibration in the vibration video sequence in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0064] Figure 5 is a flowchart of sub-pixel displacement tracking in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0065] Figure 6 is a comparison diagram of displacement map after selecting feature points and three-dimensional displacement reconstruction and displacement map collected by displacement meter in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis, figure (a) is the vibration displacement measured in the whole experiment process, and figure (b) is the vibration displacement map in one vibration stage;

[0066] Figure 7 is a flowchart of obtaining a healthy benchmark model in the glass curtain wall damage positioning method embodiment of the present application based on binocular and wavelet analysis;

[0067] Figure 8 is a flowchart of 3-layer db5 wavelet packet decomposition in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0068] Figure 9 is a flowchart of obtaining damage index spatial distribution cloud chart in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0069] Figure 10 is the SSD damage index distribution cloud chart in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0070] Figure 11 is the single-side damage positioning chart in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0071] Figure 12 is the horizontal positioning chart in the double-side damage of the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0072] Figure 13 is the vertical positioning chart in the double-side damage of the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0073] Figure 14 is the verification hidden frame glass curtain wall damage working condition chart of the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application, wherein (a) chart shows the hidden frame glass curtain wall without structural adhesive damage, (b) chart shows the hidden frame glass curtain wall without structural adhesive single-side damage 50%, and (c) chart shows the hidden frame glass curtain wall without structural adhesive double-side damage 50%;

[0074] Figure 15 is the layout chart of the verification experiment system of the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the present application;

[0075] Figure 16 is the system block diagram of the glass curtain wall damage positioning system embodiment based on binoculars and wavelet analysis of the present application.

[0076] The following are the labeling instructions in the diagram: 1. Binocular vision acquisition module; 2. Phase motion amplification module; 3. Displacement tracking module; 4. Three-dimensional displacement reconstruction module; 5. Wavelet packet analysis module; 6. Damage localization module; 7. Position determination module; 8. Computer; 9. Glass curtain wall; 10. First light source; 11. Second light source; 12. Second laser displacement meter; 13. First laser displacement meter; 14. Wall; 15. First accelerometer; 16. Second accelerometer; 17. Third accelerometer; 18. Fourth accelerometer; 19. Fifth accelerometer; 20. Sixth accelerometer; 21. Calibration plate; 101. Second high-resolution camera; 102. First high-resolution camera. Detailed Implementation

[0077] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0078] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0079] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0080] like Figure 1 The flowchart shown is an embodiment of the glass curtain wall damage localization method based on binocular and wavelet analysis of the present invention. The present invention provides a glass curtain wall damage localization method based on binocular and wavelet analysis, which is implemented by a glass curtain wall damage localization system based on binocular and wavelet analysis. The method includes:

[0081] S1. Deploy two cameras on both sides of the glass curtain wall, adjust the cameras so that their optical axes are parallel to each other and perpendicular to the plane of the glass curtain wall, and use a synchronous triggering method to collect the vibration video sequence of the glass curtain wall surface.

[0082] Specifically, such as Figure 2The arrangement schematic diagram of binocular vision in the glass curtain wall damage positioning method based on binocular and wavelet analysis is shown, the first high-resolution camera 102 is arranged on the left side of the glass curtain wall 9, the second high-resolution camera 101 is arranged on the right side of the glass curtain wall 9, the first high-resolution camera 102 is consistent with the second high-resolution camera 101 in type, the focal length is 16mm, the frame rate is greater than or equal to 100fps, the baseline length is 0.4m, the parallel arrangement is ensured by the synchronous triggering device, the distance from the glass curtain wall 9 is 3m, and the optical axis is perpendicular to the plane of the glass curtain wall 9; the first light source 10 and the second light source 11 are arranged on the two sides of the glass curtain wall respectively, and the computer 8 is used for receiving the collected image and performing signal processing and analysis.

[0083] Specifically, as shown in the camera shooting calibration board and test glass curtain wall diagram in the glass curtain wall damage positioning method based on binocular and wavelet analysis of the application, Figure 3 the camera shooting calibration board 21 and the test glass curtain wall 9 are arranged, the calibration board 7 image is a 100*100mm square grid, and the camera intrinsic parameter is calculated.

[0084] S2, according to the vibration video sequence of the surface of the glass curtain wall 9, the enhanced vibration video sequence is obtained through the phase motion amplification technology,

[0085] Specifically, as shown in the flow chart of enhancing micro-vibration in the vibration video sequence in the glass curtain wall damage positioning method based on binocular and wavelet analysis of the application, Figure 4 the flow chart comprises:

[0086] S21, according to the vibration video sequence of the surface of the glass curtain wall 9, the multi-layer sub-band image pyramid is obtained through Laplace pyramid decomposition;

[0087] S22, according to the multi-layer sub-band image pyramid, the target frequency band [f L ,f H ] is set, and the target level sub-band image is extracted,

[0088] Wherein, f L is the low-frequency cutoff frequency of the filter, and f H is the high-frequency cutoff frequency of the filter;

[0089] S23, according to the target level sub-band image, the second stage target level sub-band image is obtained through the target frequency band [f L ,f H ];

[0090] S24, according to the second stage target level sub-band image, the motion phase difference is obtained through the same layer comparison of the continuous frames of the vibration video sequence of the curtain wall surface.

[0091] S25. Based on the motion phase difference, the magnified image value is obtained using formula (1);

[0092] I out (x,y,t)=I in (x,y,t+β·Δφ(x,y,t)) (1)

[0093] Among them, I in It is the image pixel value of the current frame in the continuous frame, I out t is the image pixel value of the next frame in the continuous frame, x is the horizontal coordinate, y is the vertical coordinate, t is the time series, Δφ is the motion phase difference corresponding to the change in spatial position of coordinate (x,y) between the continuous frames at time t, and β is the magnification.

[0094] S26. Based on the magnified image value, the third-stage target layer sub-band image is obtained by overlaying the corresponding second-stage target layer sub-band image.

[0095] S27. Based on the target-level sub-band image of the third stage, an enhanced vibration video sequence is obtained through pyramid reconstruction.

[0096] S3. Based on the enhanced vibration video sequence, feature points are selected, and a high-precision displacement measurement sequence of the feature points is obtained through the adaptive KLT algorithm.

[0097] Specifically, such as Figure 5 The flowchart of sub-pixel displacement tracking in an embodiment of the glass curtain wall damage localization method based on binocular and wavelet analysis of the present invention, shown below, includes:

[0098] S31. Based on the enhanced vibration video sequence, select feature points to obtain a vibration video sequence with feature point information;

[0099] S32. Based on the vibration video sequence with feature point information, the feature point is tracked using the adaptive KLT algorithm to obtain a high-precision displacement measurement sequence of the feature point.

[0100] S4. Based on the high-precision displacement measurement sequence of the feature point, the three-dimensional displacement field of the surface of the curtain wall 9 is obtained through a three-dimensional displacement reconstruction algorithm.

[0101] The specific formulas for this three-dimensional displacement reconstruction algorithm are formula (2) and formula (3):

[0102]

[0103] d = u l -u r (3)

[0104] Where B is the baseline distance, f is the focal length, and u is the parallax.l is the horizontal coordinate of the feature point in the left camera, u r is the horizontal coordinate of the feature point in the right camera, d is the disparity of the feature point, v l is the vertical coordinate of the feature point in the left camera, c x is the horizontal position of the left camera principal point in the image coordinate system, c y is the vertical position of the left camera principal point in the image coordinate system, Z is the vibration displacement of the feature point, X is the horizontal coordinate of the feature point in the camera coordinate system, and Y is the vertical coordinate of the feature point in the camera coordinate system.

[0105] Specifically, Figure 6 The comparison diagram of the displacement graph of the selected feature point after three-dimensional displacement reconstruction and the displacement graph collected by the displacement meter is shown, Fig. (a) is the vibration displacement measured in the entire experimental process, and Fig. (b) is the vibration displacement graph in one vibration stage.

[0106] S5, according to the three-dimensional displacement field of the surface of the curtain wall 9, the energy sequences of each frequency band are extracted through wavelet packet energy decomposition, the energy mean vector representing the damage characteristics is calculated, and a healthy reference model is obtained;

[0107] Specifically, as Figure 7 shown in the flow chart for obtaining a healthy reference model in the glass curtain wall damage positioning method based on binoculars and wavelet analysis of the application, and as Figure 8 shown in the flow chart for 3-layer db5 wavelet packet decomposition in the glass curtain wall damage positioning method based on binoculars and wavelet analysis of the application, it includes:

[0108] S51, according to the three-dimensional displacement field, 3-layer db5 wavelet packet decomposition is performed to obtain 8 coefficient sequence vectors X j of different frequency bands, after 3-layer decomposition of the original time sequence signal S, 8 different frequency bands AA3, DAA3, ADA3, DDA3, AAD3, DAD3, ADD3, and DDD3 are obtained, and after arrangement, the coefficient sequence vector is obtained:

[0109] Among them,

[0110] X j = [X j1 , X j2 , … X jM ] (4)

[0111] Among them, X j represents the coefficient sequence vector of the jth frequency band range, M represents the coefficient length, and j represents the jth frequency band range;

[0112] S52, according to the coefficient sequence vector X jThe energy sequence vector of each frequency band is calculated using formula (5).

[0113]

[0114] Where M represents the coefficient length, j represents the j-th frequency band range, and X ji A sequence vector representing the i-th coefficient of the j-th frequency band. The value represents the energy of the j-th frequency band, and N represents the number of displacement signal groups in the three-dimensional displacement field of the curtain wall surface.

[0115] S53. Based on the energy sequence of each frequency band, calculate the average energy of each frequency band according to formula (6) to obtain the average energy vector μ = [μ1, μ2, ... μ1] representing the damage characteristics. j ],

[0116]

[0117] In the formula, μ represents the energy value of the j-th frequency band of the k-th group of displacement signals in this three-dimensional displacement field. j The average energy of the j-th frequency band;

[0118] S54. Based on the energy mean vector representing the damage characteristics, obtain N≥50 sets of three-dimensional displacement fields on the surface of the glass curtain wall under healthy conditions. By repeating S51-S53, calculate the energy mean vector representing the damage characteristics under healthy conditions to obtain a healthy baseline model.

[0119] S6. Based on the health benchmark model, the spatial distribution cloud map of damage indicators is obtained through the damage quantification algorithm;

[0120] Specifically, such as Figure 9 The flowchart shown in the embodiment of the glass curtain wall damage localization method based on binocular and wavelet analysis of the present invention, illustrating the acquisition of spatial distribution cloud maps of damage indicators, includes:

[0121] S61. Based on this health benchmark model, the quantitative damage index is obtained through formula (7).

[0122]

[0123] In the formula, SSD represents the quantization impairment index, n represents the total number of frequency bands after wavelet decomposition, and j represents the j-th frequency band. μ represents the energy value of the j-th frequency band collected. j This represents the average energy of the j-th frequency band of the glass curtain wall under healthy conditions;

[0124] S62. Based on the quantitative damage index, a spatial distribution cloud map of the damage index is obtained through spatial calculation.

[0125] Further, as Figure 10 A cloud map of the overall glass curtain wall SSD damage index is shown, wherein figure (a) is a single-side structural adhesive damage SSD cloud map, and figure (b) is a double-side structural adhesive damage SSD cloud map, which is used to roughly judge the position of the curtain wall structural adhesive damage.

[0126] S7, according to the damage index spatial distribution cloud map, a damage threshold is set to obtain the position area of the glass curtain wall structural adhesive damage;

[0127] Specifically, as Figure 11 The single-side damage positioning map in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the application is shown. Under the condition of single-side structural adhesive damage, the measured damage positioning map is obtained. According to the SSD curve fitted by Gaussamp, eight SSD values are fitted, wherein the points on the left and right sides are the starting points, i.e. the positions of 0 mm and 1000 mm, and the left side above is the boundary structural adhesive position of the glass curtain wall. The horizontal horizontal line represents the damage threshold 0.001. When the point on the fitting curve is greater than 0.001, it is defined as the damage starting position from the position. X represents the intersection of the fitting curve and the damage threshold 0.001, i.e. the damage starting position. As can be seen from the figure, the damage starting position is 249.23-747.600, the damage length is 498.377 mm, the actual damage length is 500 mm, and the error is 0.3246%.

[0128] Specifically, as Figure 12 The double-side damage transverse positioning map in the glass curtain wall damage positioning method embodiment based on binoculars and wavelet analysis of the application is shown. Under the condition of double-side structural adhesive damage, the measured transverse edge damage positioning map is obtained. According to the SSD curve fitted by Gaussamp, eight SSD values are fitted, wherein the points on the left and right sides are the starting points, i.e. the positions of 0 mm and 1000 mm, and the left side above is the boundary structural adhesive position of the glass curtain wall. The horizontal horizontal line represents the damage threshold 0.001. When the point on the fitting curve is greater than 0.001, it is defined as the damage starting position from the position. X represents the intersection of the fitting curve and the damage threshold 0.001, i.e. the damage starting position. As can be seen from the figure, the damage starting position is 237.22-808.657, the damage length is 571.435, the actual damage length is 500 mm, and the error is 14.287%.

[0129] Specifically, as Figure 13The longitudinal positioning graph of the bilateral damage in the embodiment of the glass curtain wall damage positioning method based on binoculars and wavelet analysis of the application is shown. Under the condition of bilateral structural adhesive damage, the measured longitudinal edge damage positioning graph is shown. According to the SSD curve fitted by Gauss amp, eight SSD values are fitted, wherein the points on the left and right sides are the starting points, that is, the positions of 0 mm and 1000 mm, and the left upper side is the position of the structural adhesive of the glass curtain wall boundary. The horizontal horizontal line represents the damage threshold 0.001. When the point on the fitted curve is greater than 0.001, it is defined as the damage starting position from the position. X represents the intersection of the fitted curve and the damage threshold 0.001, that is, the damage starting position. As can be seen from the graph, the damage starting position is 239.858-808.640, the damage length is 568.782 mm, the actual damage length is 500 mm, and the error is 13.756%.

[0130] Further, in order to verify the effectiveness of the method and steps proposed in the application, a series of experiments are carried out, which are aimed at showing the performance of the application in practical application and its improvement over the prior art.

[0131] As Figure 14 shown in the verification of the embodiment of the glass curtain wall damage positioning method based on binoculars and wavelet analysis of the application, the hidden frame glass curtain wall damage working condition situation graph is shown, wherein (a) shows the hidden frame glass curtain wall without structural adhesive damage, (b) shows the hidden frame glass curtain wall without structural adhesive single edge damage 50%, (c) shows the hidden frame glass curtain wall without structural adhesive double edge damage 50%, and the black part is the structural adhesive delamination area.

[0132] As Figure 15 shown in the arrangement diagram of the verification experiment system of the embodiment of the glass curtain wall damage positioning method based on binoculars and wavelet analysis of the application, the binocular vision detection system and the surface vibration displacement test system of the hidden frame glass curtain wall 9 are built: the hidden frame glass curtain wall 9 test piece is fixed on the wall 14, and the first acceleration sensor 15, the second acceleration sensor 16, the third acceleration sensor 17, the fourth acceleration sensor 18, the fifth acceleration sensor 19, the sixth acceleration sensor 20, the first laser displacement meter 13 and the second laser displacement meter 12 are arranged on the back of the curtain wall; according to Figure 2 the binocular camera acquisition system is installed.

[0133] After the test system is built, the hidden frame glass curtain wall center is knocked on the back of the curtain wall with a force hammer, and data acquisition is carried out.

[0134] After Figure 1The comparison diagram of the three-dimensional displacement reconstruction displacement map obtained by the process of the present application and the displacement map collected by the displacement meter is as follows Figure 6 As shown in FIG. 6, the single-side damage positioning and Figure 11 the SSD damage index cloud diagram of the present application.The following conclusions can be drawn from the diagram: Figure 12 Figure 13 1) Non-contact micro-vibration measurement is realized, and It can be seen that the displacement measurement accuracy is 0.14 mm (RMSE)

[0135] Figure 6 2) The single-side damage positioning error is only 0.32%, and the maximum double-side damage positioning error is 14.287%.

[0136] As shown in FIG. 6, the single-side damage positioning and

[0137] The system block diagram of the glass curtain wall damage positioning system based on binoculars and wavelet analysis of the present application is shown in FIG. 6, the present application provides a glass curtain wall damage positioning system based on binoculars and wavelet analysis, which is applied to a glass curtain wall damage positioning method based on binoculars and wavelet analysis, the system comprises: Figure 16 A binocular vision acquisition module 1 is used to deploy a first camera 102 and a second camera 101 on both sides of a glass curtain wall 9, adjust the cameras so that the optical axes of the cameras are parallel to each other and perpendicular to the plane where the glass curtain wall is located, and collect vibration video sequences on the surface of the glass curtain wall 9 in a synchronous triggering mode;

[0138] A phase motion amplification module 2 is used to obtain enhanced vibration video sequences through phase motion amplification technology according to the vibration video sequences on the surface of the curtain wall 9;

[0139] A displacement tracking module 3 is used to select feature points according to the enhanced vibration video sequences, and obtain high-precision displacement measurement sequences of the feature points through an adaptive KLT algorithm;

[0140] A three-dimensional displacement reconstruction module 4 is used to obtain a three-dimensional displacement field on the surface of the curtain wall 9 through a three-dimensional displacement reconstruction algorithm according to the high-precision displacement measurement sequences of the feature points;

[0141] A wavelet packet analysis module 5 is used to extract energy sequences of each frequency band through wavelet packet energy decomposition according to the three-dimensional displacement field on the surface of the curtain wall 9, calculate an energy mean vector representing damage characteristics, and obtain a health reference model;

[0142] A damage positioning module 6 is used to obtain a damage index spatial distribution cloud diagram through a damage quantification algorithm according to the health reference model;

[0143]

[0144] ​The position determining module 7 is used for obtaining the position area of the structural adhesive damage of the glass curtain wall 9 by setting a damage threshold according to the damage index space distribution cloud diagram.

[0145] The application provides a glass curtain wall damage positioning method and system based on binoculars and wavelet analysis. The application first collects curtain wall vibration video sequences through binocular cameras, and enhances micro-vibration features by using phase motion amplification technology. Then, the adaptive KLT algorithm is used to realize sub-pixel level displacement tracking, and a three-dimensional displacement field is constructed in combination with stereo calibration parameters. Secondly, wavelet packet decomposition is performed on the displacement signal, energy features of multiple frequency bands are extracted, a mean energy vector is calculated, and a health benchmark model is generated. Thirdly, the SSD index of the to-be-tested signal is calculated based on the health benchmark model, and a space distribution cloud diagram of the SSD index is obtained. Finally, a damage threshold is set, and the position and area boundary of the structural adhesive damage of the curtain wall are obtained through damage gradient detection.

[0146] It can be understood that the application is described by the above embodiments, which should not be interpreted as a limitation on the embodiments and scope of the application. Those skilled in the art know that various changes or equivalent replacements can be made to the features and embodiments without departing from the spirit and scope of the application. In addition, the features and embodiments can be modified to adapt to specific conditions and materials under the guidance of the application without departing from the spirit and scope of the application. Therefore, the application is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of the application are within the scope of the application.

Claims

1. A method for locating damage in glass curtain walls based on binocular and wavelet analysis, characterized in that, The method includes: S1. Deploy two cameras on both sides of the glass curtain wall, adjust the cameras so that their optical axes are parallel to each other and perpendicular to the plane of the glass curtain wall, and use a synchronous triggering method to collect the vibration video sequence of the glass curtain wall surface. S2. Based on the vibration video sequence of the curtain wall surface, an enhanced vibration video sequence is obtained through phase motion amplification technology; S3. Based on the enhanced vibration video sequence, select feature points and obtain a high-precision displacement measurement sequence of the feature points through the adaptive KLT algorithm; S4. Based on the high-precision displacement measurement sequence of the feature points, a three-dimensional displacement field on the curtain wall surface is obtained through a three-dimensional displacement reconstruction algorithm. S5. Based on the three-dimensional displacement field of the curtain wall surface, the energy sequence of each frequency band is extracted by wavelet packet energy decomposition, and the mean energy vector representing the damage characteristics is calculated to obtain the health benchmark model. S6. Based on the health benchmark model, obtain the spatial distribution cloud map of damage indicators through the damage quantification algorithm; S7. Based on the spatial distribution cloud map of the damage index, the location area of ​​the damage to the structural adhesive of the glass curtain wall is obtained by setting a damage threshold. S5, based on the three-dimensional displacement field of the curtain wall surface, extracts the energy sequence of each frequency band through wavelet packet energy decomposition, calculates the mean energy vector representing damage characteristics, and obtains a health baseline model, including: S51. Based on the three-dimensional displacement field of the curtain wall surface, the coefficient sequence vectors of different frequency bands are obtained using the wavelet packet energy decomposition formula (4). in, (4) in, M represents the coefficient sequence vector of the j-th frequency band, where M represents the coefficient length and j represents the j-th frequency band. S52. Based on the coefficient sequence vectors of the different frequency bands, the energy sequence vector of each frequency band is calculated using formula (5). , (5) Where M represents the coefficient length, and j represents the j-th frequency band range. A sequence vector representing the i-th coefficient of the j-th frequency band. The energy value of the j-th frequency band is represented by N, and N represents the number of displacement signal groups of the three-dimensional displacement field on the surface of the curtain wall. S53. Based on the energy sequence vectors of each frequency band, calculate the average energy of each frequency band using formula (6) to obtain the average energy vector representing the damage characteristics. , （6） In the formula, Representing the The energy value of the j-th frequency band of the displacement signal of the three-dimensional displacement field described in the group. The average energy of the j-th frequency band; S54. Based on the energy mean vector representing the damage characteristics, obtain the following information in the healthy state of the glass curtain wall: The three-dimensional displacement field of the curtain wall surface of the group is used to calculate the mean energy vector representing the damage characteristics under healthy conditions by repeating S51-S53, so as to obtain the healthy baseline model.

2. The glass curtain wall damage localization method based on binocular and wavelet analysis according to claim 1, characterized in that, In step S2, an enhanced vibration video sequence is obtained based on the vibration video sequence of the curtain wall surface using phase motion amplification technology, including: S21. Based on the vibration video sequence of the curtain wall surface, a multi-layer sub-band image pyramid is obtained through Laplace pyramid decomposition. S22. Based on the multi-layer sub-band image pyramid, a target frequency band is set. Extract the target level sub-band image. in, It is the low-frequency cutoff frequency of the filter. It is the high-frequency cutoff frequency of the filter; S23. Based on the target hierarchical sub-band image, through the target frequency band The second-stage target-level sub-band image is obtained; S24. Based on the second-stage target layer sub-band image, the motion phase difference is obtained by comparing consecutive frames of the vibration video sequence on the curtain wall surface within the same layer. S25. Based on the motion phase difference, obtain the magnified image value using formula (1); （1） in, It is the image pixel value of the current frame in the consecutive frames. It is the image pixel value of the next frame in the consecutive frames. It is the x-axis. It is the ordinate. It is a time series. yes Time coordinates The motion phase difference corresponding to the spatial position change between the consecutive frames This refers to the magnification factor. S26. Based on the magnified image values, the third-stage target layer sub-band image is obtained by overlaying the corresponding second-stage target layer sub-band image. S27. Based on the target-level sub-band image of the third stage, an enhanced vibration video sequence is obtained through pyramid reconstruction.

3. The glass curtain wall damage localization method based on binocular and wavelet analysis according to claim 1, characterized in that, In step S3, feature points are selected based on the enhanced vibration video sequence, and a high-precision displacement measurement sequence of the feature points is obtained through the adaptive KLT algorithm, including: S31. Based on the enhanced vibration video sequence, select feature points to obtain a vibration video sequence with feature point information; S32. Based on the vibration video sequence with feature point information, the feature points are tracked using an adaptive KLT algorithm to obtain a high-precision displacement measurement sequence of the feature points.

4. The glass curtain wall damage localization method based on binocular and wavelet analysis according to claim 1, characterized in that, The three-dimensional displacement reconstruction algorithm includes: （2） （3） in, Baseline distance The focal length is the parallax. It is the x-coordinate of the feature point on the left side of the camera. It is the x-coordinate of the feature point on the right-hand camera. It is the disparity of the feature point. It is the ordinate of the feature point on the left side of the camera. It is the horizontal position of the left camera principal point in the image coordinate system. Z is the vertical position of the left camera principal point in the image coordinate system, and Z is the vibration displacement of the feature point. It is the lateral coordinate of the feature point in the camera coordinate system. It is the ordinate of the feature point in the camera coordinate system.

5. The glass curtain wall damage localization method based on binocular and wavelet analysis according to claim 1, characterized in that, S6, based on the health benchmark model, obtains a spatial distribution cloud map of damage indicators through a damage quantification algorithm, including: S61. Based on the aforementioned health benchmark model, the quantitative damage index is obtained through formula (7). （7） In the formula, SSD represents the quantization impairment index, n represents the total number of frequency bands after wavelet decomposition, and j represents the j-th frequency band. This represents the energy value of the j-th frequency band collected. This represents the average energy of the j-th frequency band of the glass curtain wall under healthy conditions; S62. Based on the quantified damage index, a spatial distribution cloud map of the damage index is obtained through spatial calculation.

6. A glass curtain wall damage localization system based on binocular and wavelet analysis, used to implement the glass curtain wall damage localization method based on binocular and wavelet analysis as described in any one of claims 1-5, characterized in that, The system includes: The binocular vision acquisition module is used to deploy two cameras on both sides of the glass curtain wall, adjust the cameras so that their optical axes are parallel to each other and perpendicular to the plane of the glass curtain wall, and acquire the vibration video sequence of the glass curtain wall surface using a synchronous triggering method. The phase motion amplification module is used to obtain an enhanced vibration video sequence based on the vibration video sequence of the curtain wall surface using phase motion amplification technology. The displacement tracking module is used to select feature points based on the enhanced vibration video sequence and obtain a high-precision displacement measurement sequence of the feature points through the adaptive KLT algorithm. The three-dimensional displacement reconstruction module is used to obtain the three-dimensional displacement field of the curtain wall surface based on the high-precision displacement measurement sequence of the feature points and through a three-dimensional displacement reconstruction algorithm. The wavelet packet analysis module is used to extract the energy sequence of each frequency band based on the three-dimensional displacement field of the curtain wall surface through wavelet packet energy decomposition, calculate the energy mean vector representing the damage characteristics, and obtain the health benchmark model. The damage localization module is used to obtain a spatial distribution cloud map of damage indicators based on the health benchmark model and through a damage quantification algorithm. The location determination module is used to determine the location area of ​​damage to the structural adhesive of the glass curtain wall by setting a damage threshold based on the spatial distribution cloud map of the damage index.

Images