Building surface crack detection method and system based on machine vision

Abstract

The invention discloses a building surface crack detection method and system based on machine vision, and relates to the technical field of machine vision, image acquisition equipment adopts a multi-angle and multi-azimuth acquisition mode to realize image acquisition operation on a building surface needing to be detected, and the acquired image is processed after being received. According to the building surface crack detection method and system based on machine vision, the collected image data is received and then screened, then the feature part on the building surface is marked, the feature part is judged and broken, then the feature part determined as the crack is estimated and restored, and the crack detection accuracy is improved. In this way, feature interference of similar cracks is effectively avoided, the specific situation of the detected crack features can be estimated and restored, follow-up personnel can make maintenance judgment based on the detection situation conveniently, the speed and accuracy of building surface crack detection are remarkably improved, and the construction efficiency is improved. And powerful technical support is provided for building structure safety assessment and maintenance decision making.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and in particular to a method and system for detecting cracks on a building surface based on machine vision. Background Art

[0002] In the construction field, surface cracks on buildings are an important indicator of the health of the structure. Surface cracks may appear on building structures as they age and are affected by environmental factors (such as temperature changes, earthquakes, wind erosion, etc.).

[0003] The reference patent name is: A method for detecting cracks on the surface of a building based on a deep learning network (patent publication number: CN112330593A, patent publication date: 2021-02-05), which includes the following steps: creating a network training data set; building a deep learning network model and training it; using the trained deep learning network model to detect images and output predicted label images; analyzing and detecting predicted label images containing cracks based on crack features; calculating and outputting crack parameters, including the coordinate information of each crack, its length and width. Compared with traditional crack detection methods, a deep learning network for crack detection is constructed by combining the ideas of full convolutional networks and residual networks. It can take color images of any size as input and realize end-to-end detection. The model has high detection accuracy and strong generalization ability. In addition, the cracks are connected based on the features, and then their length and width are calculated, which provides a powerful means for quantitative evaluation of cracks.

[0004] Based on the description in the above-mentioned document, during the existing detection process of cracks on the building surface, although it is judged to have characteristics, it will be interfered with by objects with similar crack characteristics attached to the surface of the building, resulting in low detection accuracy. In addition, the crack detection operation fails to estimate the specific depth of the crack, so that the specific state and impact of the crack cannot be determined. For this reason, the present invention provides a method and system for detecting cracks on the building surface based on machine vision. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a method and system for detecting cracks on a building surface based on machine vision, which solves the problem that during the existing detection process of cracks on a building surface, although the cracks are judged to have characteristics, they may be interfered with by objects with similar crack characteristics attached to the surface of the building, resulting in low detection accuracy. In addition, the crack detection operation fails to estimate the specific depth of the crack, making it impossible to determine the specific state and impact of the crack.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for detecting cracks on a building surface based on machine vision, specifically comprising the following steps: A1. Use image acquisition equipment to acquire images of the building surface to be inspected in a multi-angle and multi-directional manner; A2. After receiving the collected image, it is processed to determine and extract the characteristics of the cracks on the building surface. The specific operation includes the following steps: A21, performing screening processing operations on the collected image data after receiving it, and retaining the required image data for analysis; A22, processing the image data, marking the characteristic parts on the building surface, evaluating the characteristic parts, and estimating and restoring the characteristic parts determined as cracks; A23. Based on the test results, the characteristics of the cracks are collected and located, and then the location and direction of the cracks on the building surface are determined, and the result data is transmitted; A3. Output the obtained results, display the data results through the display interface, and provide maintenance operation suggestions.

[0007] Preferably, the operation of performing screening processing on the collected image data after receiving is: B1. Arrange the collected images in order, and realize the matching and positioning operation between the image data and the actual building position according to the collected positions; B2, then determining whether there are features in the image data, retaining the image data with the features, and removing the image data without the features; B3. All data are transferred to the storage database for storage, and the retained data images are classified as the initial image data set for use.

[0008] Preferably, the step of marking the characteristic part on the building surface in A22 is: C1. Setting judgment points based on each image data item in the initial image data set, and setting judgment points equidistantly from bottom to top and from left to right starting from the lower left corner of the image; C2, then grayscale processing is performed on the image data in the initial image data set, the grayscale value at each judgment point is determined, the judgment point where the grayscale value changes is marked, and the boundary line of the feature part is determined according to the changed judgment point; C3. Marking operation of feature parts is implemented based on boundary lines.

[0009] Preferably, the boundary line operation of determining the characteristic part in C2 is: c21, select two adjacent judgment points whose grayscale values ​​change, set three determination points based on the equidistant distance between the two judgment points, and then determine the boundary points of the boundary line using the grayscale values ​​of the three determination points; c22. The result of determining the boundary point is: Result 1: If the grayscale values ​​of the three determined points are the same as those of the feature external judgment point, the midpoint between the determined point close to the feature internal judgment point and the feature internal judgment point is taken as the boundary point; Result 2: If the grayscale values ​​of the three determined points are the same as those of the feature internal judgment point, the midpoint between the determined point close to the feature external judgment point and the feature external judgment point is taken as the boundary point; Result 3: If the determined point close to the feature outside judgment point has the same grayscale value as the feature outside judgment point, and the other two determined points have the same grayscale value as the feature inside judgment point, then the midpoint between the determined point close to the feature outside judgment point and the determined point at the center is taken as the boundary point; Result 4: If the determined point close to the feature inner judgment point has the same grayscale value as the feature inner judgment point, and the other two determined points have the same grayscale value as the feature outer judgment point, then the midpoint between the determined point close to the feature inner judgment point and the determined point at the center is taken as the boundary point; c23. Connect adjacent and determined boundary points according to a smooth curve to form a closed-loop characteristic boundary line.

[0010] Preferably, the operation of evaluating the characteristic part in A22 is: D1. Establishing a three-dimensional coordinate model based on the image data with the marking feature, and using the three-dimensional coordinate model to determine whether the marking feature is a crack feature; D2, taking the lower left corner point of the front view image data with the marking feature as the model coordinate center point, and establishing the X-axis and Z-axis along the direction of the image data boundary with the center point, and establishing the Y-axis in the direction perpendicular to both the X-axis and the Z-axis from the center point, then placing the left view image data corresponding to the position of the marking feature image data in the intersection plane of the Y-axis and the Z-axis and the top view image data in the intersection plane of the Y-axis and the X-axis, aligning the center points and restoring to form a local building three-dimensional coordinate model; D3. Based on the internal marking identification point P (x, z) of the marking feature of the front view image, and based on the left view image data, determine whether there is a raised feature outside the wall at a height z. If so, the marking feature is a non-crack feature. Conversely, determine whether there is a raised feature outside the wall through the top view image data. If neither is present, the marking feature is a crack feature.

[0011] Preferably, the operation of performing the estimation and reduction processing in A22 is: E1. Select any image data with crack features, and find the deepest feature point of the crack based on the gray value change of the crack features; E2. Extracting image data at other angles and orientations corresponding to the image data, and selecting the collected image data directly facing the inner wall of the crack feature during the collection process; E3. Determine the three-dimensional shape of the crack based on the image data with crack characteristics and the corresponding collected image data directly facing the inner wall of the crack characteristics.

[0012] Preferably, the step of finding the deepest characteristic point of the crack in E1 is: e11. Multiple transverse points are set at equal intervals in the transverse middle of the crack characteristics, and the grayscale values ​​at the transverse points are extracted, and the grayscale values ​​of the outer transverse points gradually decrease toward the inner transverse points, and the transverse point with the lowest grayscale value is extracted; e12, and based on the horizontal points, multiple vertical points are set at equal intervals based on the vertical position of the crack characteristics, and the grayscale values ​​at the vertical points are extracted, and the grayscale values ​​of the outer horizontal points gradually decrease toward the inner horizontal points, and the vertical point with the lowest grayscale value is extracted; e13. The vertical point finally extracted is the deepest feature point of the crack feature and is marked as M.

[0013] Preferably, the operation of selecting the collected image data directly facing the characteristic inner wall of the crack during the collection process in E2 is: e21. By analyzing image data from other angles and orientations; e22. During the acquisition process, light is irradiated to the inner wall of the crack feature to determine whether there is a shadow part from the deepest feature point M to the boundary in the formed image data; e23. The image data without shadow parts in the image data are retained, and the retained image data are the collected image data facing the inner wall of the crack feature.

[0014] Preferably, the operation of determining the three-dimensional shape of the crack according to the image data having the crack feature and the corresponding collected image data directly facing the inner wall of the crack feature in E3 is: e31. Based on the three-dimensional coordinate model of the building, the collected image data of the inner wall facing the crack feature and the corresponding crack feature image are introduced; e32, the coordinates of the deepest feature point M in the crack feature image are (x1, z1), and the coordinates of the set point N at the boundary of the crack feature image are (x2, z2), and the distance from the feature point M to the set point N is L; e33, the coordinates of the deepest feature point M in the collected image data facing the inner wall of the crack feature are still (x1, z1), while the coordinates of the set point N' in the collected image data facing the inner wall of the crack feature and corresponding to the set point N at the boundary are (x3, z3), and the distance from the feature point M to the set point N' is H; e34, and calculate the depth of the corresponding crack feature according to the coordinates in the operations e32 and e33, and repeat the operation to obtain the overall inner wall condition of the crack feature and form three-dimensional crack data in the three-dimensional coordinate model; The calculation formula for the depth of crack characteristics is: ; F is the distance from the deepest feature point M to the crack cornice; and ; and .

[0015] The present invention also discloses a building surface crack detection system based on machine vision, comprising: The data acquisition unit realizes multi-angle and multi-directional image acquisition operations of the detection area; The crack detection system selects the required image data through the image screening unit, then marks the characteristic parts through the characteristic marking unit, and finally completes the recognition of the characteristics and the conversion of the three-dimensional state through the characteristic recognition unit; The result transmission module is responsible for transmitting the test results to the display and storage module in a suitable data format; The display and storage module provides data storage and display operations, making it convenient for users to analyze and view the test results.

[0016] The present invention provides a method and system for detecting cracks on building surfaces based on machine vision. Compared with the prior art, it has the following beneficial effects: (1) The machine vision-based building surface crack detection method and system performs screening and processing operations on the collected image data, and then marks the characteristic parts on the building surface to achieve the evaluation of the characteristic parts, and then estimates and restores the characteristic parts determined to be cracks. This not only effectively avoids the feature interference of similar cracks, but also can estimate and restore the specific situation of the detected crack characteristics, which is convenient for subsequent personnel to make maintenance judgments based on the detection situation, and significantly improves the speed and accuracy of building surface crack detection, providing strong technical support for building structure safety assessment and maintenance decision-making.

[0017] (2) The machine vision-based building surface crack detection method and system selects two adjacent judgment points with changed grayscale values, sets three determination points based on the equidistant interval between the two judgment points, and then uses the grayscale values ​​of the three determination points to determine the boundary points of the boundary line, and selects the boundary points based on the determination results. In this way, not only can the features be effectively identified, but also the boundaries of the features can be determined more accurately, thereby achieving accurate marking operations on the feature parts and improving the efficiency of subsequent evaluation of the crack features.

[0018] (3) The machine vision-based building surface crack detection method and system selects any image data with crack features, finds the deepest feature point of the crack based on the gray value change of the crack features, extracts image data at other angles and orientations corresponding to the image data, and selects the collected image data directly facing the inner wall of the crack features during the collection process. The three-dimensional shape of the crack is determined based on the image data with crack features and the corresponding collected image data directly facing the inner wall of the crack features. After determining the crack features, a more accurate judgment is made on the specific situation of the crack features, and the specific shape of the crack is estimated and restored, which is convenient for subsequent prediction of the crack situation and timely maintenance and processing by personnel. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is an operation flow chart of the crack detection method of the present invention; Figure 2 This is a principle block diagram of the crack detection system of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1-Figure 2 , the present invention provides three technical solutions: Embodiment 1: A method and system for detecting cracks on a building surface based on machine vision, specifically comprising the following steps: A1. Use image acquisition equipment to acquire images of the building surface to be inspected in a multi-angle and multi-directional manner; A2. After receiving the collected image, it is processed to determine and extract the characteristics of the cracks on the building surface. The specific operation includes the following steps: A21, performing screening processing operations on the collected image data after receiving it, and retaining the required image data for analysis; A22, processing the image data, marking the characteristic parts on the building surface, evaluating the characteristic parts, and estimating and restoring the characteristic parts determined as cracks; A23. Based on the test results, the characteristics of the cracks are collected and located, and then the location and direction of the cracks on the building surface are determined, and the result data is transmitted; A3. Output the obtained results, display the data results through the display interface, and provide maintenance operation suggestions.

[0022] By screening and processing the collected image data, marking the characteristic parts on the building surface, judging the characteristic parts, and then estimating and restoring the characteristic parts determined to be cracks, it not only effectively avoids the interference of similar crack features, but also estimates and restores the specific characteristics of the detected cracks, making it easier for subsequent personnel to make maintenance judgments based on the detection situation. It also significantly improves the speed and accuracy of building surface crack detection, providing strong technical support for building structure safety assessment and maintenance decisions.

[0023] In the embodiment of the present invention, the operation of screening the collected image data after receiving it is as follows: B1. Arrange the collected images in order, and realize the matching and positioning operation between the image data and the actual building position according to the collected positions; B2, then determining whether there are features in the image data, retaining the image data with the features, and removing the image data without the features; B3. All data are transferred to the storage database for storage, and the retained data images are classified as the initial image data set for use.

[0024] In the embodiment of the present invention, the steps of marking the characteristic parts on the building surface in A22 are: C1. Setting judgment points based on each image data item in the initial image data set, and setting judgment points equidistantly from bottom to top and from left to right starting from the lower left corner of the image; C2, then grayscale processing is performed on the image data in the initial image data set, the grayscale value at each judgment point is determined, the judgment point where the grayscale value changes is marked, and the boundary line of the feature part is determined according to the changed judgment point; C3. Marking operation of feature parts is implemented based on boundary lines.

[0025] The formula for grayscale processing is: ; Among them, R, G, and B are all color values, and the 2.2th power and the 2.2th root are used to convert the grayscale value into physical light power.

[0026] In the embodiment of the present invention, the boundary line operation of determining the characteristic part in C2 is: c21, select two adjacent judgment points whose grayscale values ​​change, set three determination points based on the equidistant distance between the two judgment points, and then determine the boundary points of the boundary line using the grayscale values ​​of the three determination points; c22. The result of determining the boundary point is: Result 1: If the grayscale values ​​of the three determined points are the same as those of the feature external judgment point, the midpoint between the determined point close to the feature internal judgment point and the feature internal judgment point is taken as the boundary point; Result 2: If the grayscale values ​​of the three determined points are the same as those of the feature internal judgment point, the midpoint between the determined point close to the feature external judgment point and the feature external judgment point is taken as the boundary point; Result 3: If the determined point close to the feature outside judgment point has the same grayscale value as the feature outside judgment point, and the other two determined points have the same grayscale value as the feature inside judgment point, then the midpoint between the determined point close to the feature outside judgment point and the determined point at the center is taken as the boundary point; Result 4: If the determined point close to the feature inner judgment point has the same grayscale value as the feature inner judgment point, and the other two determined points have the same grayscale value as the feature outer judgment point, then the midpoint between the determined point close to the feature inner judgment point and the determined point at the center is taken as the boundary point; c23. Connect adjacent and determined boundary points according to a smooth curve to form a closed-loop characteristic boundary line.

[0027] By selecting two adjacent judgment points whose grayscale values ​​change, three determination points are set based on the equidistant distance between the two judgment points, and then the boundary points of the boundary line are determined using the grayscale values ​​of the three determination points, and the boundary points are selected based on the determination results. This not only effectively identifies the features, but also can more accurately determine the boundaries of the features, achieve accurate marking of the feature parts, and improve the efficiency of subsequent evaluation of crack features.

[0028] In the embodiment of the present invention, the evaluation operation of the feature part in A22 is: D1. Establishing a three-dimensional coordinate model based on the image data with the marking feature, and using the three-dimensional coordinate model to determine whether the marking feature is a crack feature; D2, taking the lower left corner point of the front view image data with the marking feature as the model coordinate center point, and establishing the X-axis and Z-axis along the direction of the image data boundary with the center point, and establishing the Y-axis in the direction perpendicular to both the X-axis and the Z-axis from the center point, then placing the left view image data corresponding to the position of the marking feature image data in the intersection plane of the Y-axis and the Z-axis and the top view image data in the intersection plane of the Y-axis and the X-axis, aligning the center points and restoring to form a local building three-dimensional coordinate model; D3. Based on the internal marking identification point P (x, z) of the marking feature of the front view image, and based on the left view image data, determine whether there is a raised feature outside the wall at a height z. If so, the marking feature is a non-crack feature. Conversely, determine whether there is a raised feature outside the wall through the top view image data. If neither is present, the marking feature is a crack feature.

[0029] In the embodiment of the present invention, the operation of performing the estimation and restoration process in A22 is: E1. Select any image data with crack features, and find the deepest feature point of the crack based on the gray value change of the crack features; E2. Extracting image data at other angles and orientations corresponding to the image data, and selecting the collected image data directly facing the inner wall of the crack feature during the collection process; E3. Determine the three-dimensional shape of the crack based on the image data with crack characteristics and the corresponding collected image data directly facing the inner wall of the crack characteristics.

[0030] In the embodiment of the present invention, the step of finding the deepest characteristic point of the crack in E1 is: e11. Multiple transverse points are set at equal intervals in the transverse middle of the crack characteristics, and the grayscale values ​​at the transverse points are extracted, and the grayscale values ​​of the outer transverse points gradually decrease toward the inner transverse points, and the transverse point with the lowest grayscale value is extracted; e12, and based on the horizontal points, multiple vertical points are set at equal intervals based on the vertical position of the crack characteristics, and the grayscale values ​​at the vertical points are extracted, and the grayscale values ​​of the outer horizontal points gradually decrease toward the inner horizontal points, and the vertical point with the lowest grayscale value is extracted; e13. The vertical point finally extracted is the deepest feature point of the crack feature and is marked as M.

[0031] In the embodiment of the present invention, the operation of selecting the collected image data directly facing the inner wall of the crack feature during the collection process in E2 is: e21. By analyzing image data from other angles and orientations; e22. During the acquisition process, light is irradiated to the inner wall of the crack feature to determine whether there is a shadow part from the deepest feature point M to the boundary in the formed image data; e23. The image data without shadow parts in the image data are retained, and the retained image data are the collected image data facing the inner wall of the crack feature.

[0032] In the embodiment of the present invention, the operation of determining the three-dimensional shape of the crack in E3 according to the image data having the crack feature and the corresponding collected image data directly facing the inner wall of the crack feature is: e31. Based on the three-dimensional coordinate model of the building, the collected image data of the inner wall facing the crack feature and the corresponding crack feature image are introduced; e32, the coordinates of the deepest feature point M in the crack feature image are (x1, z1), and the coordinates of the set point N at the boundary of the crack feature image are (x2, z2), and the distance from the feature point M to the set point N is L; e33, the coordinates of the deepest feature point M in the collected image data facing the inner wall of the crack feature are still (x1, z1), while the coordinates of the set point N' in the collected image data facing the inner wall of the crack feature and corresponding to the set point N at the boundary are (x3, z3), and the distance from the feature point M to the set point N' is H; e34, and calculate the depth of the corresponding crack feature according to the coordinates in the operations e32 and e33, and repeat the operation to obtain the overall inner wall condition of the crack feature and form three-dimensional crack data in the three-dimensional coordinate model; The calculation formula for the depth of crack characteristics is: ; F is the distance from the deepest feature point M to the crack cornice; and ; and .

[0033] By selecting any image data with crack characteristics, finding the deepest feature point of the crack based on the gray value change of the crack characteristics, extracting the image data of other angles and orientations at the corresponding image data, and selecting the acquired image data that is directly opposite to the inner wall of the crack characteristics during the acquisition process, the three-dimensional shape of the crack is determined according to the image data with crack characteristics and the corresponding acquired image data that is directly opposite to the inner wall of the crack characteristics. After determining the crack characteristics, a more accurate judgment can be made on the specific situation of the crack characteristics, and the specific shape of the crack can be estimated and restored, which is convenient for the subsequent prediction of the crack situation and timely maintenance and processing by personnel.

[0034] Embodiment 2 is different from Embodiment 1 in that: the present invention further discloses a building surface crack detection system based on machine vision, including: The data acquisition unit realizes multi-angle and multi-directional image acquisition operations of the detection area; The crack detection system selects the required image data through the image screening unit, then marks the characteristic parts through the characteristic marking unit, and finally completes the recognition of the characteristics and the conversion of the three-dimensional state through the characteristic recognition unit; The result transmission module is responsible for transmitting the test results to the display and storage module in a suitable data format; The display and storage module provides data storage and display operations, making it convenient for users to analyze and view the test results.

[0035] The data acquisition unit includes a camera acquisition unit to realize image acquisition operations, an auxiliary lighting unit to realize the illumination operation of the inner wall of the crack during the acquisition process, an auxiliary adjustment unit to realize the position adjustment or angle change of the data acquisition unit, and the auxiliary lighting unit changes synchronously with the camera acquisition unit.

[0036] Embodiment 3 is different from Embodiment 1 and Embodiment 2 in that the existing building surface crack detection method and the building surface crack detection method based on machine vision of the present invention are applied to multiple building surfaces with known specific conditions for detection operations, and the efficiency of the detection completion and the final detection results are recorded. The specific results are shown in Table 1: Table 1 Result generation table

[0037] To sum up, by adopting the building surface crack detection method based on machine vision of the present invention to perform building surface detection operations, the time to complete the detection is shorter, and the accuracy of the specific results after detection is higher, so it can be better applied to the detection operation of building surface cracks.

[0038] Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0039] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0040] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting cracks on building surfaces based on machine vision, characterized in that: The specific steps include: A1. Use image acquisition equipment to acquire images of the building surface to be inspected in a multi-angle and multi-directional manner; A2. After receiving the collected image, it is processed to determine and extract the characteristics of the cracks on the building surface. The specific operation includes the following steps: A21, performing screening processing operations on the collected image data after receiving it, and retaining the required image data for analysis; A22, processing the image data, marking the characteristic parts on the building surface, evaluating the characteristic parts, and estimating and restoring the characteristic parts determined as cracks; A23. Based on the test results, the characteristics of the cracks are collected and located, and then the location and direction of the cracks on the building surface are determined, and the result data is transmitted; A3. Output the obtained results, display the data results through the display interface, and provide maintenance operation suggestions.

2. The method for detecting cracks on building surfaces based on machine vision according to claim 1, characterized in that: The operation of screening the collected image data after receiving it is as follows: B1. Arrange the collected images in order, and realize the matching and positioning operation between the image data and the actual building position according to the collected positions; B2, then determining whether there are features in the image data, retaining the image data with the features, and removing the image data without the features; B3. All data are transferred to the storage database for storage, and the retained data images are classified as the initial image data set for use.

3. The method for detecting cracks on building surfaces based on machine vision according to claim 2, characterized in that: The steps of marking the characteristic parts on the building surface in A22 are: C1. Setting judgment points based on each image data item in the initial image data set, and setting judgment points equidistantly from bottom to top and from left to right starting from the lower left corner of the image; C2, then grayscale processing is performed on the image data in the initial image data set, the grayscale value at each judgment point is determined, the judgment point where the grayscale value changes is marked, and the boundary line of the feature part is determined according to the changed judgment point; C3. Marking operation of feature parts is implemented based on boundary lines.

4. The method for detecting cracks on building surfaces based on machine vision according to claim 3, characterized in that: The boundary line operation of determining the feature part in C2 is: c21, select two adjacent judgment points whose grayscale values ​​change, set three determination points based on the equidistant distance between the two judgment points, and then determine the boundary points of the boundary line using the grayscale values ​​of the three determination points; c22. The result of determining the boundary point is: Result 1: If the grayscale values ​​of the three determined points are the same as those of the feature external judgment point, the midpoint between the determined point close to the feature internal judgment point and the feature internal judgment point is taken as the boundary point; Result 2: If the grayscale values ​​of the three determined points are the same as those of the feature internal judgment point, the midpoint between the determined point close to the feature external judgment point and the feature external judgment point is taken as the boundary point; Result 3: If the determined point close to the feature outside judgment point has the same grayscale value as the feature outside judgment point, and the other two determined points have the same grayscale value as the feature inside judgment point, then the midpoint between the determined point close to the feature outside judgment point and the determined point at the center is taken as the boundary point; Result 4: If the determined point close to the feature inner judgment point has the same grayscale value as the feature inner judgment point, and the other two determined points have the same grayscale value as the feature outer judgment point, then the midpoint between the determined point close to the feature inner judgment point and the determined point at the center is taken as the boundary point; c23. Connect adjacent and determined boundary points according to a smooth curve to form a closed-loop characteristic boundary line.

5. The method for detecting cracks on building surfaces based on machine vision according to claim 1, characterized in that: The evaluation operation of the feature part in A22 is as follows: D1. Establishing a three-dimensional coordinate model based on the image data with the marking feature, and using the three-dimensional coordinate model to determine whether the marking feature is a crack feature; D2, taking the lower left corner point of the front view image data with the marking feature as the model coordinate center point, and establishing the X-axis and Z-axis along the direction of the image data boundary with the center point, and establishing the Y-axis in the direction perpendicular to both the X-axis and the Z-axis from the center point, then placing the left view image data corresponding to the position of the marking feature image data in the intersection plane of the Y-axis and the Z-axis and the top view image data in the intersection plane of the Y-axis and the X-axis, aligning the center points and restoring to form a local building three-dimensional coordinate model; D3. Based on the internal marking identification point P (x, z) of the marking feature of the front view image, and based on the left view image data, determine whether there is a raised feature outside the wall at a height z. If so, the marking feature is a non-crack feature. Conversely, determine whether there is a raised feature outside the wall through the top view image data. If neither is present, the marking feature is a crack feature.

6. The method for detecting cracks on building surfaces based on machine vision according to claim 5, characterized in that: The operations of the estimation and reduction process in A22 are: E1. Select any image data with crack features, and find the deepest feature point of the crack based on the gray value change of the crack features; E2. Extracting image data at other angles and orientations corresponding to the image data, and selecting the collected image data directly facing the inner wall of the crack feature during the collection process; E3. Determine the three-dimensional shape of the crack based on the image data with crack characteristics and the corresponding collected image data directly facing the inner wall of the crack characteristics.

7. The machine vision-based building surface crack detection method according to claim 6, characterized in that: The steps for finding the deepest feature point of the crack in E1 are: e11. Multiple transverse points are set at equal intervals in the transverse middle of the crack characteristics, and the grayscale values ​​at the transverse points are extracted, and the grayscale values ​​of the outer transverse points gradually decrease toward the inner transverse points, and the transverse point with the lowest grayscale value is extracted; e12, and based on the horizontal points, multiple vertical points are set at equal intervals based on the vertical position of the crack characteristics, and the grayscale values ​​at the vertical points are extracted, and the grayscale values ​​of the outer horizontal points gradually decrease toward the inner horizontal points, and the vertical point with the lowest grayscale value is extracted; e13. The vertical point finally extracted is the deepest feature point of the crack feature and is marked as M.

8. The machine vision-based building surface crack detection method according to claim 7, characterized in that: The operation of selecting the collected image data directly facing the inner wall of the crack feature during the collection process in E2 is: e21. By analyzing image data from other angles and orientations; e22. During the acquisition process, light is irradiated to the inner wall of the crack feature to determine whether there is a shadow part from the deepest feature point M to the boundary in the formed image data; e23. The image data without shadow parts in the image data are retained, and the retained image data are the collected image data facing the inner wall of the crack feature.

9. The method for detecting cracks on building surfaces based on machine vision according to claim 7, characterized in that: The operation of determining the three-dimensional shape of the crack in E3 according to the image data with crack characteristics and the corresponding collected image data directly facing the inner wall of the crack characteristics is: e31. Based on the three-dimensional coordinate model of the building, the collected image data of the inner wall facing the crack feature and the corresponding crack feature image are introduced; e32, the coordinates of the deepest feature point M in the crack feature image are (x1, z1), and the coordinates of the set point N at the boundary of the crack feature image are (x2, z2), and the distance from the feature point M to the set point N is L; e33, the coordinates of the deepest feature point M in the collected image data facing the inner wall of the crack feature are still (x1, z1), while the coordinates of the set point N' in the collected image data facing the inner wall of the crack feature and corresponding to the set point N at the boundary are (x3, z3), and the distance from the feature point M to the set point N' is H; e34, and calculate the depth of the corresponding crack feature according to the coordinates in the operations e32 and e33, and repeat the operation to obtain the overall inner wall condition of the crack feature and form three-dimensional crack data in the three-dimensional coordinate model; The calculation formula for the depth of crack characteristics is: ； F is the distance from the deepest feature point M to the crack cornice; and ; and .

10. A machine vision-based building surface crack detection system, using a machine vision-based building surface crack detection method according to any one of claims 1 to 9, characterized in that: include: The data acquisition unit realizes multi-angle and multi-directional image acquisition operations of the detection area; The crack detection system selects the required image data through the image screening unit, then marks the characteristic parts through the characteristic marking unit, and finally completes the recognition of the characteristics and the conversion of the three-dimensional state through the characteristic recognition unit; The result transmission module is responsible for transmitting the test results to the display and storage module in a suitable data format; The display and storage module provides data storage and display operations, making it convenient for users to analyze and view the test results.

Images