An intelligent detection method for damage and full-life health monitoring of cable-membrane structures

Abstract

The present invention discloses an intelligent detection method for damage and full-life health monitoring of cable-membrane structures, which includes two parts: the initial detection of membrane surface damage of the cable-membrane structure to be measured and the periodic full-life monitoring during the service period of the cable-membrane structure to be measured. The former matches the real-time captured image of the cable-membrane structure with the cable-membrane template to identify the membrane surface damage of the cable-membrane structure to be measured; the latter establishes a damage detection matrix set based on the previous detection results and conducts periodic comparison detection with the fixed-point module of the damage detection matrix set to complete the periodic monitoring of the cable-membrane structure product. The present invention realizes actively acquiring the image of the cable-membrane structure by relying on a single camera sensor device, can stably, accurately and quickly judge whether there is damage or dirt on the cable-membrane structure to be measured, and periodically saves the damage comparison results, which has certain reference value for relevant research in the field of cable-membrane structure life detection, can obtain the periodic damage degree during the service period of the cable-membrane structure, and is of great significance for the implementation of the cable-membrane structure damage identification system.

Description

Technical Field

[0001] The present invention relates to the technical field of full-life monitoring of membrane structure building materials, and particularly relates to a method for intelligent detection of cable-membrane structure damage and full-life health monitoring. Background Art

[0002] In the 1990s of the 20th century, cable-membrane structure buildings began to be applied in China. Representative projects include Shanghai 80,000-seat Stadium, Shanghai Hongkou Football Stadium, Qingdao Yizhong Stadium, Wuhu Olympic Sports Center, etc. Although the application of membrane structures in China is nearly 50 years later than that in foreign countries, in the past 20 years, the application and development speed of membrane structures in China is higher than that in any region of the world, and they are widely used in various large-scale stadiums, exhibition halls, aviation, railway, culture and entertainment and other public buildings. However, in contrast to its rapid development speed, the development of damage detection and health monitoring of cable-membrane structures is relatively slow. Under the combined action of factors such as environmental erosion, material aging, long-term fatigue effect of loads, and sudden accident overload effect during the service period of large-span cable-membrane structures, which are landmark buildings in various places, damage accumulation and resistance attenuation in the membrane structure system will inevitably occur, thus reducing the ability of the structure to resist natural disasters and even normal service loads. Under extreme load conditions such as strong winds, typhoons and heavy snow, catastrophic accidents are extremely likely to occur, causing heavy casualties and economic losses.

[0003] At present, the method of manual regular inspection is adopted for the damage detection of cable-membrane structures. However, the inspection by staff will be affected by various subjective and objective factors such as mood, weather, knowledge level and carelessness, which greatly reduces the accuracy of damage detection of cable-membrane structure products. On the other hand, at present, cable-membrane structure buildings are mainly some landmark buildings or building features such as carports. Quite a number of relevant buildings in service belong to public facilities and are in a state of being unattended for a long time, which needs to be taken seriously. Using the method of manual regular inspection will increase public and labor costs.

[0004] Therefore, considering multiple aspects such as building safety, inspection accuracy and cost, it has high engineering significance to develop relevant unmanned and intelligent methods for intelligent detection of cable-membrane structure damage and full-life health monitoring. Summary of the Invention

[0005] Technical problems to be solved: In view of the above technical problems, the present invention provides a method for intelligent detection of cable-membrane structure damage and full-life health monitoring, which can effectively solve the deficiencies that the above-mentioned method of manual regular inspection is affected by subjective factors such as the mood, knowledge level and carefulness of staff and weather conditions, and there is no one to monitor the full life cycle under the service state.

[0006] Technical solution: A method for intelligent detection of cable-membrane structure damage and full-life health monitoring includes the following steps:

[0007] S1. Collect a healthy cable-membrane structure as a cable-membrane template, take a new image of the cable-membrane structure to be measured and compare it with the cable-membrane template to detect damage. Use the rows and columns of the new image of the cable-membrane structure to be measured as the database row-column coordinates, and save the results of the damage detection as the initial damage detection matrix set V1 according to the database row-column coordinates, thus completing the initial detection of the membrane surface damage of the cable-membrane structure to be measured.

[0008] S2. Take 1 week as a monitoring period, take the i-th image of the cable-membrane structure to be measured according to the database row-column coordinates described in step S1, compare the i-th image with the (i - 1)-th damage detection matrix set V i-1 to detect damage, and save the damage detection matrix set V i according to the monitoring times to obtain periodic damage results, thus completing the periodic monitoring of the membrane surface damage of the cable-membrane structure to be measured, where i is the current monitoring times and i is an integer greater than 0. When i = 1, the (i - 1) = 0-th damage detection matrix set V0 is the cable-membrane template.

[0009] Preferably, step S1 specifically includes the following steps:

[0010] S11. Use a monocular camera to take pictures of the healthy part of the cable-membrane structure, store it as a cable-membrane template, and perform a 32*32 scaling process, and store it as template data.

[0011] S12. Conduct a full-coverage scan and take pictures of the cable-membrane structure to be measured.

[0012] S13. Perform image preprocessing on the new image obtained from the full-coverage shooting in step S12 to obtain the preprocessed new image.

[0013] S14. Use a similarity detection algorithm to traverse and match the preprocessed new image obtained in step S13 with the cable-membrane template and template data obtained in step S11.

[0014] S15. Save the images and their coordinate data that are not successfully matched in step S14, and store them as the initial damage detection matrix set V1.

[0015] Furthermore, the cable-membrane template described in step S11 is an image of the healthy part of the cable-membrane structure collected in the past or to be measured.

[0016] Furthermore, the full-coverage scan and shooting described in step S12 include the following steps:

[0017] S121. Use a camera to take pictures of the left and right edges and the upper and lower edges of the cable-membrane structure to be measured respectively, and record the camera lens angles.

[0018] S122. Starting from the upper left corner of the cable-membrane structure to be measured, use the camera to horizontally take pictures to the right edge of the cable-membrane structure product with a rotation step of 5 degrees.

[0019] S123. The camera is adjusted downward by 5 degrees and horizontally shoots from right to left with an angular step of 5 degrees until the left edge of the cable-membrane structure to be measured is reached.

[0020] S124. Repeat steps S122 - S123 until reciprocating full-coverage shooting is completed, and record the number of columns and rows of the shooting and save them in the newly imported image matrix set U. U is an M×N matrix, where M and N represent the row number and column number respectively.

[0021] Preferably, step S2 specifically includes the following steps:

[0022] S21. Determine the shooting position of the monocular camera according to the initial detection row and column coordinates, and shoot the i-th image at the corresponding coordinate position of the cable-membrane structure to be measured.

[0023] S22. Perform image preprocessing on the i-th image to obtain the preprocessed i-th image.

[0024] S23. Use the similarity detection algorithm to calculate the i-th image preprocessed in step S22 and the (i - 1)-th image at the corresponding coordinate position to obtain the hash value and Hamming distance α of the i-th image.

[0025] S24. Update the damage detection matrix set V i , record the damage condition of the cable-membrane structure to be measured at different monitoring time periods, and complete the full-life health monitoring during the service period of the cable-membrane structure to be measured.

[0026] Furthermore, the damage detection matrix set V i is when the Hamming distance of the i-th image is detected to be non-zero, record the coordinate position of the i-th image in the newly imported image matrix set U and its Hamming distance α, and V i is shown as follows:

[0027]

[0028] In the formula, HASH mn represents the hash value of the i-th newly imported shooting image in the m-th row and n-th column and the (i - 1)-th comparison image, where m = 1, 2, …, M and n = 1, 2, …, N; ρ mn = α i,mn -α i-1,mn represents the change value of the Hamming distance of the newly imported shooting image in the m-th row and n-th column at the i-th shooting compared with the previous record. When i = 1, ρ mn = 0.

[0029] Furthermore, the specific operation of the image preprocessing is:

[0030] Step 1) Perform adaptive style transfer on the image to be preprocessed according to the characteristics of the cable-membrane template: Extract the data of the illumination, hue, color saturation, and brightness of the cable-membrane template as the cable-membrane template characteristics, and use Equation ① as the image loss function:

[0031]

[0032] In the formula: μ(X T ) is the style loss degree, representing the style similarity between the image to be preprocessed and the cable-membrane template. X T represents the image data to be preprocessed. H, S, and I represent hue, color saturation, and brightness respectively. represents the hue similarity between the cable-membrane template and the image to be preprocessed. represents the color saturation similarity between the cable-membrane template and the image to be preprocessed. represents the brightness similarity between the cable-membrane template and the image to be preprocessed. Use the logarithmic function to integrate and calculate the hue similarity, color saturation similarity, and brightness similarity to obtain the style loss degree, and perform style transfer on the image to be preprocessed when the spatial misalignment occurs according to the style loss degree;

[0033] Step 2) Perform scaling processing on the image to be preprocessed after the style transfer in Step 1): Both the cable-membrane template and the preprocessed image are color RGB images. Use Equation ② to calculate the three channels of the color RGB image to obtain a grayscale image.

[0034] Gray = (30×R + 59×G + 11×B + 50) / 100 ②

[0035] In the formula, Gray is the grayscale value, and R, G, and B represent the values of the red, green, and blue channels of the color image respectively. Then, scale the image according to the method of pixel averaging to a ratio of 32*32, obtaining 1024 pixel points.

[0036] Preferably, the specific process of Step S14 is as follows:

[0037] S141, Calculate the average value of the 1024 pixel point values;

[0038] S142, Perform 0-1 matrix processing on the new incoming image matrix set U according to the average value obtained in S141. Pixel points greater than or equal to the average value are set to 1, and vice versa;

[0039] S143, Readjust the number of rows and columns of the matrix, and record the matrix values in the order of "from left to right, from top to bottom" to obtain a hash value composed of 0s and 1s in a row;

[0040] S144, Use the exclusive OR operation to calculate the hash values of the cable-membrane template and the new incoming image, and sum to obtain the Hamming distance α between the two sets of data.

[0041] Beneficial effects: (1) The present invention proposes an intelligent detection method for cable-membrane structure damage and a full-life health monitoring method, which can intelligently detect the damaged area on the membrane surface during the initial damage detection of the cable-membrane structure, reduce the deficiencies of manual detection, and improve the accuracy of membrane surface detection;

[0042] (2) The present invention proposes a periodic damage monitoring method for cable-membrane structures, which can perform periodic detection using an intelligent detection method for cable-membrane structure damage based on the Hamming distance of hashing, and reasonably save damage data, realizing the full-life health monitoring of cable-membrane service structures, and contributing to the implementation of a cable-membrane structure damage identification system;

[0043] (3) The present invention combines image processing technology with the field of full-life health monitoring of cable-membrane structures, provides a reference for the health detection of building materials such as cable-membrane structures, and has engineering significance. Description of the Drawings

[0044] Figure 1 is a flowchart of an intelligent detection method for cable-membrane structure damage and a full-life health monitoring method according to an embodiment of the present invention;

[0045] Figure 2 is a schematic diagram of an intelligent detection method for cable-membrane structure damage and a full-life health monitoring method according to an embodiment of the present invention;

[0046] Figure 3 is a schematic diagram of a similarity detection algorithm. Detailed Embodiments

[0047] The present invention will be described in detail below with reference to the drawings and specific embodiments:

[0048] Embodiment 1

[0049] As Figures 1-3 , an intelligent detection method for cable-membrane structure damage and a full-life health monitoring method includes the following steps:

[0050] S1. Collect a healthy cable-membrane structure as a cable-membrane sample, compare and detect damage by taking a new image of the cable-membrane structure to be measured with the cable-membrane sample, use the rows and columns of the new image of the cable-membrane structure to be measured as the database row-column coordinates, and save the results of the damage detection as the initial damage detection matrix set V1 according to the database row-column coordinates to complete the initial detection of the membrane surface damage of the cable-membrane structure to be measured. The specific steps include:

[0051] S11. Use a monocular camera to take pictures of the healthy part of the cable-membrane structure, store it as a cable-membrane sample, and perform a 32*32 scaling process, and store it as sample data;

[0052] S12. Perform full-coverage scanning and shooting on the cable-membrane structure to be measured. The full-coverage scanning and shooting includes the following steps:

[0053] S121. Use a camera to separately capture the left and right edges and the upper and lower edges of the cable-membrane structure to be measured, and record the camera lens angles.

[0054] S122. Starting from the upper left corner of the cable-membrane structure to be measured, use a camera to horizontally capture to the right edge of the cable-membrane structure product with a rotation step of 5 degrees.

[0055] S123. Adjust the camera downward by 5 degrees and horizontally capture from right to left to the left edge of the cable-membrane structure to be measured with a rotation step of 5 degrees.

[0056] S124. Repeat steps S122 - S123 until reciprocating full coverage shooting is completed, record the number of columns and rows of the shooting, and save them in the newly imported image matrix set U. U is a matrix of M×N, where M and N represent the row number and column number respectively.

[0057] S13. Perform image preprocessing on the newly imported images obtained from the full coverage shooting in step S12 to obtain the preprocessed newly imported images. The specific operations of the image preprocessing are as follows:

[0058] Step 1) Perform adaptive style transfer on the image to be preprocessed according to the cable-membrane template features: Extract the data of the illumination, hue, color saturation, and brightness of the cable-membrane template as the cable-membrane template features, and use Equation ① as the image loss function:

[0059]

[0060] In the formula: μ(X T ) is the style loss degree, indicating the style similarity between the image to be preprocessed and the cable-membrane template. X T represents the image data to be preprocessed. H, S, and I represent hue, color saturation, and brightness respectively. represents the hue similarity between the cable-membrane template and the image to be preprocessed. represents the color saturation similarity between the cable-membrane template and the image to be preprocessed. represents the brightness similarity between the cable-membrane template and the image to be preprocessed. Use the logarithmic function to integrate and calculate the hue similarity, color saturation similarity, and brightness similarity to obtain the style loss degree, and perform style transfer on the image to be preprocessed when the spatial misalignment occurs according to the style loss degree.

[0061] Step 2) Perform scaling processing on the image to be preprocessed after the style transfer in step 1): Both the cable-membrane template and the preprocessed image are color RGB images. Use Equation ② to calculate the three channels of the color RGB image to obtain a grayscale image.

[0062] Gray = (30×R + 59×G + 11×B + 50) / 100 ②

[0063] In the formula, Gray is the grayscale value, and R, G, and B respectively represent the values of the red, green, and blue channels of the color image. Then, the image is scaled according to the method of pixel averaging to a ratio of 32 * 32, obtaining 1024 pixel points.

[0064] S14. Use the similarity detection algorithm to traverse and match the preprocessed new incoming image obtained in step S13 with the cable-membrane sample and template data obtained in step S11. The specific process is as follows:

[0065] S141. Calculate the average value of the 1024 pixel point values;

[0066] S142. Perform 0-1 matrix processing on the new incoming image matrix set U according to the average value obtained in S141. Pixel points greater than or equal to the average value are set to 1, and vice versa to 0;

[0067] S143. Readjust the number of rows and columns of the matrix, and record the matrix values in the order of "from left to right, from top to bottom" to obtain a hash value composed of 0s and 1s in a row;

[0068] S144. Use exclusive OR operation to calculate the hash values of the cable-membrane template and the new incoming image, and sum to obtain the Hamming distance α between the two sets of data.

[0069] S15. Save the images and their coordinate data that are not successfully matched in step S14, and store them as the initial damage detection matrix set V1.

[0070] S2. Using 1 week as a monitoring period, take the i-th image of the cable-membrane structure to be measured according to the database row and column coordinates described in step S1. Compare the i-th image with the (i - 1)-th damage detection matrix set V i-1 for damage detection, and save the damage detection matrix set V according to the monitoring times i to obtain periodic damage results and complete the periodic monitoring of the membrane surface damage of the cable-membrane structure to be measured. Here, i is the current monitoring times, and i is an integer greater than 0. When i = 1, the (i - 1) = 0-th damage detection matrix set V0 is the cable-membrane template. The specific steps include the following:

[0071] S21. Determine the shooting position of the monocular camera according to the initial detection row and column coordinates, and take the i-th image of the corresponding coordinate position of the cable-membrane structure to be measured;

[0072] S22. Perform image preprocessing on the i-th image to obtain the preprocessed i-th image;

[0073] S23. Use the similarity detection algorithm to calculate the i-th image preprocessed in step S22 with the (i - 1)-th image at the corresponding coordinate position to obtain the hash value and Hamming distance α of the i-th image;

[0074] S24. Update the set V of damage detection matrices i , record the damage conditions of the cable-membrane structure to be measured during different monitoring periods, and complete the full-life health monitoring during the service period of the cable-membrane structure to be measured. The set V of damage detection matrices i is when the Hamming distance of the i-th image is not 0, record the coordinate position of the i-th image in the new image matrix set U and its Hamming distance α, V i is shown as the following formula:

[0075]

[0076] In the formula, HASH mn represents the hash value of the i-th newly taken image in the m-th row and n-th column compared with the (i - 1)-th comparison image, where m = 1, 2,..., M and n = 1, 2,..., N; ρ mn = α i,mn - α i-1,mn represents the change value of the Hamming distance of the newly taken image in the m-th row and n-th column at the i-th shooting compared with the previously recorded one. When i = 1, ρ mn = 0.

[0077] In summary, the present invention proposes a method for intelligent detection of cable-membrane structure damage and full-life health monitoring, which can intelligently detect the damaged area of the membrane surface in the initial damage detection of the cable-membrane structure, reduce the deficiencies of manual detection, and improve the accuracy of membrane surface detection; the proposed periodic damage monitoring method for cable-membrane structures can perform periodic detection using the intelligent damage detection method for cable-membrane structures based on the Hamming distance of hashing, and reasonably save damage data, realizing the full-life health monitoring of cable-membrane service structures, which is helpful for the implementation of the cable-membrane structure damage identification system; combining image processing technology with the field of full-life health monitoring of cable-membrane structures provides a reference for the health detection of building materials such as cable-membrane structures and has engineering significance.

[0078] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent detection method for damage and full-life health monitoring of cable-membrane structures, characterized in that: It includes the following steps: S1. Collect a healthy cable-membrane structure as a cable-membrane template, take a new image of the cable-membrane structure to be measured and compare it with the cable-membrane template to detect damage. Use the rows and columns of the new image of the cable-membrane structure to be measured as the database row-column coordinates, and save the results of the damage detection as the initial damage detection matrix set V1 according to the database row-column coordinates, completing the initial detection of the membrane surface damage of the cable-membrane structure to be measured. The specific steps are as follows: S11. Use a monocular camera to take pictures of the healthy part of the cable-membrane structure, store it as a cable-membrane template, and perform a 32*32 scaling process, and store it as template data; S12. Conduct a full-coverage scan and shooting of the cable-membrane structure to be measured; S13. Perform image preprocessing on the new image obtained from the full-coverage shooting in step S12 to obtain the preprocessed new image; S14. Use a similarity detection algorithm to traverse and match the preprocessed new image obtained in step S13 with the cable-membrane template and template data obtained in step S11; S15. Save the images and their coordinate data that fail to match in step S14, and store them as the initial damage detection matrix set V1; S2. Taking one week as a monitoring period, photograph the i-th image of the cable-membrane structure to be measured according to the row and column coordinates of the database described in step S1, compare the i-th image with the (i - 1)-th damage detection matrix set V i-1 to detect damage, and save the damage detection matrix set V according to the monitoring times i to obtain periodic damage results and complete the periodic monitoring of the membrane surface damage of the cable-membrane structure to be measured. Where i is the current monitoring times and i is an integer greater than 0. When i = 1, the (i - 1)-th damage detection matrix set V0 is the cable-membrane template, which specifically includes the following steps: S21. Determine the shooting position of the monocular camera according to the initial detection row-column coordinates, and take the i-th image of the corresponding coordinate position of the cable-membrane structure to be measured; S22. Perform image preprocessing on the i-th image to obtain the preprocessed i-th image; S23. Using a similarity detection algorithm, calculate the \(i\)-th image preprocessed in step S22 and the \((i - 1)\)-th image at the corresponding coordinate position to obtain the hash value and Hamming distance \(\alpha\) of the \(i\)-th image, and the damage detection matrix set \(V\). i When the Hamming distance of the \(i\)-th image is detected to be non-zero, record the coordinate position of the \(i\)-th image in the new incoming image matrix set \(U\) and its Hamming distance \(\alpha\), \(V\). i As shown in the following formula: In the formula, HASH mn represents the hash value of the newly captured image in the m-th row, n-th column, and i-th time compared with the (i - 1)-th comparison image, where m = 1, 2, …, M and n = 1, 2, …, N; ρ mn = α i,mn - α i-1,mn represents the change value of the Hamming distance of the newly captured image in the m-th row, n-th column at the i-th capture compared with the previous record. When i = 1, ρ mn = 0; S24, update the set V of damage detection matrices i , record the damage conditions of the cable-membrane structure to be measured during different monitoring periods, and complete the full-life health monitoring of the cable-membrane structure to be measured during its service period; The specific operation of the image preprocessing is as follows: Step 1) Perform adaptive style transfer on the image to be preprocessed according to the cable-membrane template features: Extract the data of the illumination, hue, color saturation, and brightness of the cable-membrane template as the cable-membrane template features, and use Equation ① as the image loss function: Where: μ(X T ) is the style loss degree, representing the style similarity between the image to be preprocessed and the cable-membrane template. X T represents the image data to be preprocessed, and H, S, and I represent hue, saturation, and brightness respectively. represents the hue similarity between the cable-membrane template and the image to be preprocessed. represents the saturation similarity between the cable-membrane template and the image to be preprocessed. represents the brightness similarity between the cable-membrane template and the image to be preprocessed. The logarithmic function is used to integrate and calculate the hue similarity, saturation similarity, and brightness similarity to obtain the style loss degree. The style transfer of the image to be preprocessed is performed according to the style loss degree in the case of spatial misalignment. Step 2) Perform a scaling process on the image to be preprocessed after the style transfer in step 1): Both the cable-membrane template and the preprocessed image are color RGB images. Use Equation ② to calculate the three channels of the color RGB image to obtain a grayscale image, Gray=(30×R+59×G+11×B+50) / 100 ② In the formula, Gray is the grayscale value, and R, G, and B respectively represent the values of the red, green, and blue 3 channels of the color image. Then, scale the image according to the method of pixel averaging to a ratio of 32*32, obtaining 1024 pixel points.

2. The intelligent detection method for damage and full-life health monitoring of cable-membrane structures according to claim 1, wherein: The cable-membrane template described in step S11 is an image collected from the healthy part of the previous or cable-membrane structure to be measured.

3. A method for intelligent detection of cable-membrane structure damage and full-life health monitoring according to claim 1, characterized in that The full-coverage scan and shooting described in step S12 include the following steps: S121. Use a camera to take pictures of the left and right edges and the upper and lower edges of the cable-membrane structure to be measured respectively, and record the camera lens angle; S122. Starting from the upper left corner of the cable-membrane structure to be measured, use the camera to horizontally shoot to the right edge of the cable-membrane structure product with a rotation step of 5 degrees; S123. Adjust the camera downward by 5 degrees, and horizontally shoot from right to left with a rotation step of 5 degrees to the left edge of the cable-membrane structure to be measured; S124. Repeat steps S122 - S123 until the reciprocating full-coverage shooting is completed, and record the number of columns and rows of the shooting and save them in the new image matrix set U. U is an M×N matrix, and M and N respectively represent the row number and column number.

4. The intelligent detection method for damage and full-life health monitoring of cable-membrane structures according to claim 1, characterized in that, The specific process of step S14 is as follows: S141. Calculate the average value of 1024 pixel values; S142. Perform 0-1 matrix processing on the newly entered image matrix set U according to the average value obtained in S141. Pixel points greater than or equal to the average value are set to 1, and vice versa to 0; S143. Readjust the number of rows and columns of the matrix, and record the matrix values in the order of "from left to right, from top to bottom" to obtain a hash value composed of 0 and 1 in a row; S144. Use exclusive OR operation to calculate the hash values of the cable membrane template and the newly entered image, and sum to obtain the Hamming distance α between the two sets of data.

Images