An intelligent visual detection and early warning system for water conservancy engineering slope cracks
Patent Information
- Application Number
- CN202611228378.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-13
- Publication Date
- 2026-10-30
AI Technical Summary
在实际检测过程中，有些裂缝个体可能较长较宽，但形态已经稳定，不会进一步的变化，相对稳定，风险程度较低；而有些裂缝个体虽然较小，但具有较快的延伸扩张速度，故仅依靠单独裂缝的静态特征进行检测具有局限性，影响边坡隐患的检测准确性
[0028]本发明突破传统通过连通域外接矩形长宽特征评估静态风险的局限，结合裂缝面积、强损伤节点与平均宽度的差异特征计算静态基础风险程度，更准确地表征裂缝当前的损伤程度与后续扩张的失稳可能性。后续结合多个时刻的裂缝图像计算强损伤节点的动态变化强度，基于强损伤节点分布特征、动态变化强度、不同时刻的骨架增长特征，完整构建出裂缝扩张轨迹与裂缝扩张方向，可精准捕捉裂缝尖端延伸、局部损伤点扩张等动态变化行为，突破仅依靠静态特征判断风险的局限，进一步提高裂缝风险判断的准确性。通过聚类划分相互影响的裂缝区域，根据裂缝共向度量化区域内多道裂缝共同向同一位置集聚扩张的趋势，解决了边坡高应力下裂缝集群场景中，单条裂缝独立评估准确度大幅下降的缺陷。最后融合区域裂缝数量、单条裂缝静态基础风险、裂缝共向度得到综合扩张演化风险值并划分多级预警等级，兼顾当前静态损伤状态与未来动态扩张演化趋势，从而大幅提升水利边坡裂缝检测预警的整体准确性，提前规避边坡局部滑塌、整体失稳的风险，保障水利工程结构运行安全。
Smart Images
Figure CN123016608A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of visual inspection technology, specifically to an intelligent visual inspection and early warning system for slope cracks in water conservancy projects. Background Technology
[0002] Slopes in water conservancy projects are subjected to the combined effects of rainfall erosion, water level fluctuations, groundwater seepage, temperature changes, and geological stress over long periods. This can lead to cracks and crack expansion, potentially causing localized landslides or even overall instability, threatening the structural safety and stability of the water conservancy project. Therefore, regular inspection of cracks in water-cooled engineering slopes is necessary to detect cracks and their potential expansion in a timely manner. Currently, crack risk is typically assessed based on static characteristics such as location, length, and width. However, due to the high bearing stress of slopes, cracks often cluster during their development, which reduces the accuracy of individual crack assessments. In actual inspections, some individual cracks may be long and wide but have a stable morphology and are unlikely to change further, indicating relatively low risk. Conversely, some smaller cracks may have a rapid expansion rate, making detection based solely on the static characteristics of individual cracks limited and affecting the accuracy of slope hazard detection. Summary of the Invention
[0003] To address the aforementioned technical problems, the present invention aims to provide an intelligent visual detection and early warning system for slope cracks in hydraulic engineering projects. The specific technical solution adopted is as follows:
[0004] The image acquisition module is used to acquire crack images on the slope at different times; and to extract the crack connected components and skeleton from the crack images.
[0005] The static analysis module is used to obtain the average width feature value based on the area characteristics of the crack connected domain and the length characteristics of the skeleton; to obtain the strong damage nodes in the crack based on the average width feature value and the local width of the crack; and to obtain the static basic risk level of the crack based on the area characteristics of the crack connected domain, the difference between the width at the strong damage node and the average width feature value.
[0006] The change analysis module is used to obtain the dynamic change intensity based on the width change characteristics of the strong damage nodes in the crack images at subsequent time steps; to obtain the crack propagation trajectory and crack propagation direction based on the distribution characteristics of different strong damage nodes, the dynamic change intensity, and the skeleton growth characteristics at different time steps; to obtain different crack regions based on the distribution characteristics of the crack propagation trajectory; to obtain the intersection points and intersection center points based on the concentration characteristics of different crack propagation directions in the crack regions; and to obtain the crack co-directionality based on the number of intersection points and the distance between the intersection points and the intersection center point.
[0007] The risk warning module is used to obtain a comprehensive expansion and evolution risk value based on the number of cracks in the crack region, the static basic risk level, and the crack co-directionality; and to issue a warning based on the comprehensive expansion and evolution risk value of the crack region.
[0008] Furthermore, the step of obtaining the average width feature value based on the area characteristics of the crack connected domain and the length characteristics of the skeleton includes:
[0009] The average width feature value is obtained by calculating the ratio of the number of pixels between the crack connected region and the corresponding skeleton.
[0010] Further, the step of obtaining the strong damage nodes in the crack based on the average width feature value and the local width of the crack includes:
[0011] The vertical distance from the skeleton pixel to the boundary of the two crack connected regions is taken as the width of the skeleton pixel; the skeleton pixel with a width greater than the average width feature value is taken as a strong damage node.
[0012] Furthermore, the step of obtaining the static basic risk level of the crack based on the area characteristics of the crack's connected domain and the difference between the width and average width characteristics at the strongly damaged node includes:
[0013] In the formula, A represents the static foundation risk level of the crack. Let S represent the number of pixels in the crack connected region, and N represent the number of strongly damaged nodes in the crack connected region. This represents the width at the nth strongly damaged node. This represents the average width feature value.
[0014] Furthermore, the step of obtaining the dynamic change intensity based on the width change characteristics of the crack image at subsequent time points of the strongly damaged node includes:
[0015] In the formula, D represents the dynamic change intensity of the strongly damaged node. The normalization is indicated by H, which represents the number of width changes of the strongly damaged node in the crack image at subsequent time steps. This represents the width difference between the h-th crack image (where the width varies) and the previous crack image at the strongly damaged node.
[0016] Furthermore, the step of obtaining the crack propagation trajectory and crack propagation direction based on the distribution characteristics of different strong damage nodes and the dynamic change intensity and skeleton growth characteristics at different times includes:
[0017] The skeleton endpoints in the first crack image are determined. If the skeleton endpoints in the crack images at subsequent time steps are extended or changed, the adjacent newly added endpoints are connected in sequence to obtain the crack propagation trajectory. The direction from the skeleton endpoints in the first crack image directly to the last skeleton endpoint of the crack propagation trajectory is taken as the crack propagation direction.
[0018] Starting from the strongest damage node with the greatest dynamic change intensity in the crack connected domain, connect the other adjacent strong damage nodes in sequence to obtain the crack propagation trajectory; take the direction from the end of the strong damage node in the crack propagation trajectory to the first strong damage node as the crack propagation direction; starting from the order of dynamic change intensity from large to small, traverse the remaining strong damage nodes that do not appear in the crack propagation trajectory in sequence to obtain all crack propagation trajectories and crack propagation directions.
[0019] Furthermore, the step of obtaining different crack regions based on the distribution characteristics of the crack propagation trajectory includes:
[0020] The sum of the Euclidean distances between the strong damage nodes corresponding to any two crack propagation trajectories is calculated to obtain the clustering distance between the two crack propagation trajectories. The crack propagation trajectories are then clustered using the K-means clustering algorithm and the clustering distance to obtain different crack regions.
[0021] Furthermore, the step of obtaining the intersection point and the intersection center point based on the concentration characteristics of different crack propagation directions in the crack region includes:
[0022] In the crack region, extend rays along the crack propagation direction for all crack propagation trajectories, and take the points where the rays intersect as the intersection points; take the point that is closest to all intersection points as the intersection center point.
[0023] Furthermore, the step of obtaining the crack co-directionality based on the quantity characteristics of the intersection points and the distance characteristics between the intersection points and the intersection center point includes:
[0024] Calculate the reciprocal of the sum of the Euclidean distances from the intersection center point to all intersection points to obtain the concentration characteristic value; calculate the product of the number of intersection points and the concentration characteristic value to obtain the crack co-directionality.
[0025] Furthermore, the step of obtaining the comprehensive expansion and evolution risk value based on the crack quantity characteristics in the crack region, the static foundation risk level, and the crack co-directionality includes:
[0026] In the formula, Q represents the comprehensive risk value of the expansion and evolution of the crack region. This indicates normalization, and M represents the number of connected fracture regions in the fracture region. V represents the static basic risk level of the m-th crack connected domain, and V represents the crack co-orientation of the crack region.
[0027] The present invention has the following beneficial effects:
[0028] This invention overcomes the limitations of traditional methods that assess static risk based solely on the length and width of the bounding rectangle of a connected domain. It combines crack area, differences between strongly damaged nodes and average width to calculate the degree of static basic risk, more accurately characterizing the current damage level of the crack and the likelihood of subsequent instability. Subsequently, it calculates the dynamic intensity of strongly damaged nodes by combining crack images from multiple time points. Based on the distribution characteristics of strongly damaged nodes, their dynamic intensity, and the skeleton growth characteristics at different times, it comprehensively constructs the crack expansion trajectory and direction, accurately capturing dynamic changes such as crack tip extension and local damage point expansion. This overcomes the limitations of relying solely on static features to judge risk, further improving the accuracy of crack risk assessment. By clustering and dividing mutually influential crack regions, and based on the trend of multiple cracks within a co-directional quantitative region clustering and expanding towards the same location, it addresses the significant decrease in accuracy of independent assessment of a single crack in high-stress slope crack cluster scenarios. Finally, by integrating the number of cracks in the region, the static foundation risk of a single crack, and the co-directionality of cracks, a comprehensive expansion and evolution risk value is obtained and a multi-level early warning system is established. This system takes into account both the current static damage state and the future dynamic expansion and evolution trend, thereby significantly improving the overall accuracy of crack detection and early warning for water conservancy slopes, avoiding the risks of local landslides and overall instability of slopes in advance, and ensuring the safe operation of water conservancy engineering structures. Attached Figure Description
[0029] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0030] Figure 1 This is a block diagram of an intelligent visual detection and early warning system for slope cracks in water conservancy projects, provided as an embodiment of the present invention. Detailed Implementation
[0031] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent visual detection and early warning system for slope cracks in hydraulic engineering projects proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0033] The following description, in conjunction with the accompanying drawings, details the specific scheme of the intelligent visual detection and early warning system for slope cracks in water conservancy projects provided by this invention.
[0034] Please see Figure 1 The diagram illustrates a block diagram of an intelligent visual detection and early warning system for slope cracks in water conservancy projects, according to an embodiment of the present invention. The system includes the following modules:
[0035] The image acquisition module S1 is used to acquire crack images on the slope at different times; and to extract the crack connected components and skeleton from the crack images.
[0036] A fixed industrial camera is used, mounted on a fixed bracket opposite the slope. It can be installed at fixed locations such as monitoring towers, corridors, dams, and bridges. The camera maintains a fixed viewing angle and acquires visible light images of the slope. One image is taken every hour until multiple images from various time points are collected. These images are then analyzed by combining them. The specific number of time points can be 8 to 24, determined by the implementer based on the implementation scenario. For each slope image, a multi-scale Retinex algorithm is first used for illumination correction to eliminate grayscale differences on the slope surface caused by shadows, localized strong light, and uneven illumination. A color restoration factor is then used to enhance image contrast. Next, the Sauvola adaptive threshold segmentation algorithm is used to extract crack areas. The local window size is typically set to 31×31 pixels, with parameter k=0.2 and a dynamic parameter range R=128 to adapt to the complex texture of the slope surface and changes in illumination. After obtaining the binary image, morphological opening operations are used to remove isolated noise points, and then closing operations are used to connect local fractures and cracks to improve the integrity of the crack region. Subsequently, an eight-neighbor connected component labeling algorithm is used to divide the binary image into connected components and label each independent crack region. To more accurately evaluate the static characteristics of the cracks, the skeleton of the crack connected components is extracted using the Skeleton algorithm. Simultaneously, manually detected slope crack image samples with and without expansion risk are acquired for subsequent auxiliary risk identification. When a crack connected component is detected in a slope image, the slope image is referred to as a crack image. In this embodiment of the invention, 24 crack images of the same location from the next 24 hours are analyzed together to comprehensively determine the expansion risk of the slope cracks.
[0037] The static analysis module S2 is used to obtain the average width feature value based on the area characteristics of the crack connected domain and the length characteristics of the skeleton; to obtain the strong damage nodes in the crack based on the average width feature value and the local width of the crack; and to obtain the static basic risk level of the crack based on the area characteristics of the crack connected domain, the difference between the width at the strong damage node and the average width feature value.
[0038] First, the width characteristic of the crack is determined. The average width characteristic value is obtained based on the area characteristic of the crack's connected region and the length characteristic of the skeleton. Preferably, in this embodiment of the invention, the step of obtaining the average width characteristic value includes: calculating the ratio of the number of pixels between the crack's connected region and the corresponding skeleton to obtain the average width characteristic value. The number of pixels in the crack's connected region can characterize the area of the connected region, and the number of pixels in the skeleton can characterize the length of the skeleton. By approximating the crack as a long rectangular region, the average width of the crack is obtained through the ratio of area to length. The crack width can characterize the current degree of damage to the crack.
[0039] Furthermore, since the stress on different parts of the crack may vary, the crack width may differ in different segments of the same crack. Therefore, strong damage nodes in the crack are obtained based on the average width feature value and the local crack width. Preferably, in this embodiment of the invention, the step of obtaining strong damage nodes includes: taking the vertical distance from the skeleton pixel to the boundary of the crack's connected domain on both sides as the width of the skeleton pixel; and taking skeleton pixels with a width greater than the average width feature value as strong damage nodes. Because wider crack nodes increase the value of the average width of the crack's connected domain, normally uniform cracks are generally smaller than the average width, while the width value of relatively wide cracks is greater than the average width. Therefore, the difference between the width at each skeleton pixel and the average width feature value is compared. Strong damage nodes indicate that the crack width at that location is large, usually indicating that a large cumulative damage or deformation has occurred in that local area. If the crack expands further, this area may become more sensitive.
[0040] The larger the overall area of the current crack connected domain, the stronger the damage to the slope caused by the crack, and the higher the risk of the crack area. On the other hand, fewer strong damage nodes in the crack, or smaller differences between the width of most strong damage nodes and the average width, indicate a more uniform overall crack width. Cracks with more uniform width are more likely to represent stable and uniformly expanding cracks. The more strong damage nodes and the wider the crack, the more nodes with concentrated local damage, non-uniform structure, or abnormal stress transmission there are in the current crack, and the higher the uncertainty and potential risk of the crack as a whole. Therefore, the static foundation risk level of the crack is obtained based on the area characteristics of the crack connected domain and the difference between the width of strong damage nodes and the average width. Preferably, in this embodiment of the invention, the step of obtaining the static foundation risk level includes:
[0041]
[0042] In the formula, A represents the static foundation risk level of the crack. Normalization is performed based on manually labeled crack samples, using maximum and minimum values. S represents the number of pixels in the crack's connected region; a higher number of pixels indicates a larger connected region area and a higher risk level for the crack. N represents the number of strongly damaged nodes in the crack's connected region. This represents the width at the nth strongly damaged node. This represents the average width feature value. When there are more strongly damaged nodes, and their widths are greater than the average width feature value, then... The larger the value, the more nodes with concentrated local damage, uneven structure, or abnormal stress transmission exist within the crack, indicating a greater static foundation risk. Calculating the static foundation risk of a crack is more accurate than existing techniques that directly assess static risk based on the length and width characteristics of the bounding rectangle of the connected domain.
[0043] The change analysis module S3 is used to obtain the dynamic change intensity based on the width change characteristics of strong damage nodes in the crack images at subsequent time steps; to obtain the crack propagation trajectory and crack propagation direction based on the distribution characteristics and dynamic change intensity of different strong damage nodes and the skeleton growth characteristics at different time steps; to obtain different crack regions based on the distribution characteristics of crack propagation trajectories; to obtain the intersection points and intersection center points based on the concentration characteristics of different crack propagation directions in the crack regions; and to obtain the crack co-directionality based on the number of intersection points and the distance characteristics between intersection points and intersection center points.
[0044] Static features can only characterize the current damage level of cracks and the possibility of instability during re-expansion, but cannot further characterize the existence of expansion risk. Some cracks with larger widths may simply be the result of the slope stress being released, and in reality, the crack part has already stabilized and will not expand further. Therefore, in order to improve the accuracy of crack risk assessment, it is necessary to combine the dynamic change characteristics of cracks represented by crack images at multiple time points for analysis. Based on all the strong damage nodes in the crack image at the first time point within an analysis period, the width of the crack in the crack images at subsequent time points is calculated to determine whether there is a width change at the location. If the width is greater than the width at the strong damage node at the previous time point, it indicates that the crack has widened, there is still expansion stress at that location, and there is a dynamic stress risk. Therefore, the dynamic change intensity is obtained based on the width change characteristics of strong damage nodes in the crack images at subsequent time points. Preferably, in this embodiment of the invention, the step of obtaining the dynamic change intensity includes:
[0045]
[0046] In the formula, D represents the dynamic change intensity of the strongly damaged node. Normalization is indicated. H represents the number of width changes of the strong damage node in the crack image at subsequent time steps. When the width of the strong damage node in any crack image is greater than the width of the crack image at the previous time step, it is recorded as one width change. The more width changes there are, the more frequent the crack expansion is, and the greater the intensity of dynamic change at that point. This represents the width difference between the h-th crack image (where width changes) and the previous crack image at the point of strong damage. A larger width difference indicates a greater width expansion at that location, higher risk, and stronger dynamic change.
[0047] Furthermore, for mechanical analysis, the most dangerous location of a crack, besides the existing wide crack damage area, is the crack tip region. If the crack still has stress and is at risk of expansion, the crack tip is likely to be one of the stress release gaps. Therefore, stress concentration usually occurs at the crack tip, inducing the crack to continue expanding. Further analysis of the crack expansion trajectory and direction is then conducted. Based on the distribution characteristics and dynamic changes of different strong damage nodes, and the skeleton growth characteristics at different times, the crack expansion trajectory and direction are obtained. Preferably, in this embodiment of the invention, the steps for obtaining the crack expansion trajectory and direction include: determining the skeleton endpoints in the first crack image, where the skeleton endpoints represent the end positions of the crack; if the skeleton endpoints extend in subsequent crack images, it indicates that the crack is expanding, the skeleton lengthens, and new skeleton endpoints appear. Then, adjacent newly added endpoints are connected sequentially to obtain the crack expansion trajectory; this crack expansion trajectory reflects the extension state of the crack tip. Because the stress at the crack tip is released along the tip direction, the direction directly from the skeleton endpoint in the first crack image to the last skeleton endpoint of the crack expansion trajectory is taken as the crack expansion direction; the crack expansion direction reflects the approximate crack extension direction. If the skeleton endpoint does not change in subsequent crack images, no expansion trajectory will be constructed.
[0048] After analyzing the expansion trajectory at the crack tip, the expansion within the crack is handled as follows: Starting from the strongest damage node with the highest dynamic change intensity in the crack's connected domain, adjacent strong damage nodes are connected sequentially to obtain the crack expansion trajectory; this process continues until an adjacent pixel is no longer a strong damage node. Because the expansion trajectory construction begins from the strongest damage node with the highest dynamic change intensity, this location experiences frequent dynamic changes and has the highest expansion degree. Therefore, the direction from the strongest damage node at the end of the crack expansion trajectory to the first strongest damage node is taken as the crack expansion direction. A strong damage node with a large dynamic change intensity may correspond to multiple crack expansion trajectories. Starting from the strongest dynamic change intensity in descending order, the remaining strong damage nodes not appearing in the crack expansion trajectory are traversed sequentially, and their corresponding expansion trajectories and directions are constructed to obtain all crack expansion trajectories and directions. Expansion trajectories are not constructed for isolated strong damage nodes. This step, by constructing the crack expansion trajectories and directions at the crack tip and strong damage points within the crack, provides a foundation for subsequent dynamic expansion risk analysis of the crack.
[0049] Furthermore, since clustered cracks may interact during crack expansion, increasing the risk of slope crack expansion, crack expansion trajectories that are close together are divided into the same region that can influence each other, based on the distribution characteristics of crack expansion trajectories. Therefore, different crack regions are obtained based on the distribution characteristics of crack expansion trajectories. Preferably, in this embodiment, the step of obtaining crack regions includes: calculating the sum of Euclidean distances between strong damage nodes corresponding to any two crack expansion trajectories to obtain the clustering distance between the two crack expansion trajectories; clustering the crack expansion trajectories using the K-means clustering algorithm and the clustering distance to obtain different crack regions. When the lengths of two crack expansion trajectories are the same, the sum of Euclidean distances between corresponding pixels in the trajectory is calculated as the clustering distance; if the trajectory lengths are different, the same number of pixels as the short trajectory are marked at equal intervals on the longer trajectory, and the sum of Euclidean distances between corresponding pixels is calculated as the clustering distance. It should be noted that the calculation order is the expansion direction corresponding to the expansion trajectory. After clustering, close crack expansion trajectories can be divided into the same crack region for analysis.
[0050] Since the direction of crack propagation reflects the main propagation direction of cracks with dynamic expansion characteristics, when such cracks tend to extend and cluster together, it indicates that these cracks have a common evolutionary trend. The center position of the propagation direction has a greater risk of dynamic expansion, the stress concentration in the entire area is stronger, and the overall risk of the area is higher. Therefore, the intersection point and the intersection center point are obtained based on the concentration characteristics of different crack propagation directions in the crack region. In this embodiment of the invention, the steps of obtaining the intersection point and the intersection center point include: extending rays along the crack propagation direction of all crack propagation trajectories in the crack region, where the rays can characterize the approximate future expansion direction of the crack propagation trajectory; taking the points where the rays intersect as intersection points; and taking the point with the shortest distance from all intersection points as the intersection center point. When the rays do not intersect, it indicates that the expansion directions are significantly different. The intersection point reflects the convergence position of different cracks' common extension and evolution trends. The more intersection points there are and the closer the intersection center point is to each intersection point, the stronger the trend of crack propagation trajectories extending to the same position in the crack region, the stronger the common evolutionary trend of the cracks, the greater the risk of expansion and deterioration at the center position, and the higher the overall risk of the area. Furthermore, crack co-directionality is obtained based on the number of intersection points and the distance between intersection points and the intersection center point. Preferably, in this embodiment of the invention, the step of obtaining crack co-directionality includes: calculating the reciprocal of the sum of the Euclidean distances from the intersection center point to all intersection points to obtain a concentration characteristic value; the smaller the sum of the Euclidean distances from the intersection center point to all intersection points, the closer the distance to each intersection point, the more obvious the risk of cracks expanding towards the same location, and the stronger the co-directional evolution trend. The product of the number of intersection points and the concentration characteristic value is calculated to obtain crack co-directionality. The more intersection points and the larger the concentration characteristic value, the greater the crack co-directionality, and the higher the risk of crack co-directional expansion in that crack region.
[0051] The risk warning module S4 is used to obtain a comprehensive expansion and evolution risk value based on the characteristics of the number of cracks in the crack area, the degree of static foundation risk, and the co-directionality of cracks; and to issue a warning based on the comprehensive expansion and evolution risk value of the crack area.
[0052] The greater the number of cracks in the same crack region, the higher the corresponding static foundation risk level, and the greater the crack co-orientation, the stronger the overall expansion and evolution risk of the current crack region, and the more attention is needed to this crack region; therefore, the overall expansion and evolution risk value is obtained based on the crack number characteristics, static foundation risk level, and crack co-orientation in the crack region; preferably, in this embodiment of the invention, the step of obtaining the overall expansion and evolution risk value includes:
[0053]
[0054] In the formula, Q represents the comprehensive risk value of the expansion and evolution of the crack region. The normalization is represented by M, which represents the number of fracture connected domains in the fracture region. The fracture region is obtained by clustering the fracture propagation trajectory. The more different fracture connected domains corresponding to the fracture propagation trajectory, the denser the fractures are, and the higher the risk of subsequent expansion and evolution of the fracture region. The static basic risk level represents the m-th crack's connected domain. A higher static basic risk level indicates that the crack already has a certain degree of risk. V represents the crack co-directionality of the crack region; a higher crack co-directionality indicates a more obvious trend of cracks in the crack region expanding and evolving towards the same location. Therefore, a higher comprehensive expansion and evolution risk value means a greater degree of damage from cracks in the crack region, requiring more attention. Finally, an early warning can be issued based on the comprehensive expansion and evolution risk value of the crack region. The comprehensive expansion and evolution risk value ranges from 0 to 2. In this embodiment, three assessment levels are defined: when the comprehensive expansion and evolution risk value of the crack region is less than 0.66, the crack's risk is considered low, and continued observation is sufficient; when it is not less than 0.66 and not greater than 1.33, a concentrated expansion risk is considered, requiring marking and close observation; when it is greater than 1.33, the current crack's risk is considered high, requiring immediate warning and relevant personnel to carry out inspection and maintenance. Implementers can determine the risk assessment method according to the implementation scenario; no limitations are imposed here. Thus, by combining static risk assessment of cracks with dynamic expansion trends to obtain a comprehensive expansion and evolution risk value, the accuracy of slope crack status detection and early warning has been improved.
[0055] In summary, this invention provides an intelligent visual detection and early warning system for slope cracks in hydraulic engineering projects. It obtains average width feature values and highly damaged nodes based on the crack connectivity and framework; it obtains the static foundation risk level based on the width of the crack connectivity, highly damaged nodes, and average width feature values; it obtains the dynamic change intensity based on the width variation characteristics of highly damaged nodes; it obtains the crack expansion trajectory and crack expansion direction based on the distribution of highly damaged nodes, dynamic change intensity, and framework growth characteristics; it obtains the intersection points and intersection center points based on the concentration characteristics of different crack expansion directions in the crack area; and it obtains crack co-directionality based on the intersection points and intersection center points. This invention obtains a comprehensive expansion evolution risk value based on the number of cracks in the crack area, the static foundation risk level, and crack co-directionality; and it provides early warning based on this comprehensive expansion evolution risk value, thus improving the accuracy of crack detection and early warning.
[0056] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0057] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.
Claims
1. An intelligent visual detection and early warning system for slope cracks in hydraulic engineering projects, characterized in that, The system includes the following modules: The image acquisition module is used to acquire crack images on the slope at different times; and to extract the crack connected components and skeleton from the crack images. The static analysis module is used to obtain the average width feature value based on the area characteristics of the crack connected domain and the length characteristics of the skeleton; to obtain the strong damage nodes in the crack based on the average width feature value and the local width of the crack; and to obtain the static basic risk level of the crack based on the area characteristics of the crack connected domain, the difference between the width at the strong damage node and the average width feature value. The change analysis module is used to obtain the dynamic change intensity based on the width change characteristics of the strong damage nodes in the crack images at subsequent time steps; to obtain the crack propagation trajectory and crack propagation direction based on the distribution characteristics of different strong damage nodes, the dynamic change intensity, and the skeleton growth characteristics at different time steps; to obtain different crack regions based on the distribution characteristics of the crack propagation trajectory; to obtain the intersection points and intersection center points based on the concentration characteristics of different crack propagation directions in the crack regions; and to obtain the crack co-directionality based on the number of intersection points and the distance between the intersection points and the intersection center point. The risk warning module is used to obtain a comprehensive expansion and evolution risk value based on the crack quantity characteristics in the crack region, the static basic risk level, and the crack co-directionality. Early warning is issued based on the comprehensive risk value of the expansion and evolution of the crack area.
2. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the average width feature value based on the area characteristics of the crack connected domain and the length characteristics of the skeleton includes: The average width feature value is obtained by calculating the ratio of the number of pixels between the crack connected region and the corresponding skeleton.
3. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the strong damage nodes in the crack based on the average width characteristic value and the local width of the crack includes: The vertical distance from the skeleton pixel to the boundary of the two crack connected domains is taken as the width of the skeleton pixel; the skeleton pixel with a width greater than the average width feature value is taken as a strong damage node.
4. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the static basic risk level of the crack based on the area characteristics of the crack's connected domain and the difference between the width at the strongly damaged node and the average width characteristic value includes: In the formula, A represents the static foundation risk level of the crack. Let S represent the normalization, S represent the number of pixels in the crack connected region, and N represent the number of strongly damaged nodes in the crack connected region. This represents the width at the nth strongly damaged node. This represents the average width feature value.
5. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the dynamic change intensity based on the width change characteristics of the crack image at subsequent time steps of the strongly damaged node includes: In the formula, D represents the dynamic change intensity of the strongly damaged node. The normalization is indicated by H, which represents the number of width changes of the strongly damaged node in the crack image at subsequent time steps. This represents the width difference between the h-th crack image (where the width varies) and the previous crack image at the strongly damaged node.
6. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The steps of obtaining the crack propagation trajectory and crack propagation direction based on the distribution characteristics of different strong damage nodes, the dynamic change intensity, and the skeleton growth characteristics at different times include: The skeleton endpoints in the first crack image are determined. If the skeleton endpoints in the crack images at subsequent time steps are extended or changed, the adjacent newly added endpoints are connected in sequence to obtain the crack propagation trajectory. The direction from the skeleton endpoints in the first crack image directly to the last skeleton endpoint of the crack propagation trajectory is taken as the crack propagation direction. Starting from the strongest damage node with the greatest dynamic change intensity in the crack connected domain, connect the other adjacent strong damage nodes in sequence to obtain the crack propagation trajectory; take the direction from the end of the strong damage node in the crack propagation trajectory to the first strong damage node as the crack propagation direction; starting from the order of dynamic change intensity from large to small, traverse the remaining strong damage nodes that do not appear in the crack propagation trajectory in sequence to obtain all crack propagation trajectories and crack propagation directions.
7. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining different crack regions based on the distribution characteristics of the crack propagation trajectory includes: The sum of the Euclidean distances between the strong damage nodes corresponding to any two crack propagation trajectories is calculated to obtain the clustering distance between the two crack propagation trajectories. The crack propagation trajectories are then clustered using the K-means clustering algorithm and the clustering distance to obtain different crack regions.
8. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the intersection point and the intersection center point based on the concentration characteristics of different crack propagation directions in the crack region includes: In the crack region, extend rays along the crack propagation direction for all crack propagation trajectories, and take the points where the rays intersect as the intersection points; take the point that is closest to all intersection points as the intersection center point.
9. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the crack co-axis based on the quantity characteristics of the intersection points and the distance characteristics between the intersection points and the intersection center point includes: Calculate the reciprocal of the sum of the Euclidean distances from the intersection center point to all intersection points to obtain the concentration characteristic value; calculate the product of the number of intersection points and the concentration characteristic value to obtain the crack co-directionality.
10. The intelligent visual detection and early warning system for slope cracks in water conservancy projects according to claim 1, characterized in that, The step of obtaining the comprehensive expansion and evolution risk value based on the crack quantity characteristics in the crack region, the static basic risk level, and the crack coaxiality includes: In the formula, Q represents the comprehensive risk value of the expansion and evolution of the crack region. This indicates normalization, and M represents the number of connected fracture regions in the fracture region. V represents the static basic risk level of the m-th fracture connected domain, and V represents the fracture co-orientation of the fracture region.