Bridge concrete quality monitoring method and system based on edge detection

Abstract

The invention relates to the technical field of bridge quality monitoring, in particular to a bridge concrete quality monitoring method and system based on edge detection, and the method comprises the steps: S1, image collection and dynamic standardization; s2, concrete surface pretreatment; s3, concrete crack edge detection; s4, quantifying the geometrical characteristics of the crack; and S5, performing quality grading and disposal. According to the scheme, the unmanned aerial vehicle carries a multispectral camera to realize automatic acquisition of bridge concrete surface images, flies in a grid path, and is matched with a silicon photodiode sensor to obtain environmental illumination in real time, illumination correction is performed on the images through a dynamic compensation coefficient, and overexposure / underexposure influences are eliminated. In the preprocessing stage, median filtering and a CLAHE algorithm are adopted, and noise interference is reduced while edge features are effectively reserved. Compared with traditional manual inspection or fixed camera shooting, the process has the advantages that data acquisition uniformity and image quality are greatly improved, and subjective errors of manual operation are avoided.

Description

Technical Field

[0001] This invention relates to the field of bridge quality monitoring technology, and in particular to a method and system for monitoring the quality of bridge concrete based on edge detection. Background Technology

[0002] As a critical component of transportation infrastructure, bridge concrete quality monitoring is essential, as cracks and defects in the concrete structure directly impact load-bearing capacity and durability. The formation and propagation of concrete cracks is a gradual process; if early, minute cracks are not monitored in time, they may develop into through cracks, leading to steel corrosion, reduced structural stiffness, and even safety accidents.

[0003] Traditional bridge inspection relies on manual inspections or fixed-point sensor sampling, which has limited detection range, low efficiency, and large subjective errors. It is particularly difficult to capture early crack features under conditions of uneven lighting and complex surface texture interference, failing to meet the demands of modern bridge operation and maintenance for automated, high-precision monitoring. Therefore, we propose a bridge concrete quality monitoring method and system based on edge detection. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for monitoring the quality of bridge concrete based on edge detection, thereby solving the technical problems mentioned in the background section.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for monitoring the quality of bridge concrete based on edge detection includes the following steps: S1: Image acquisition and dynamic standardization. The drone is equipped with a multispectral camera to acquire images of the bridge concrete surface. The ambient illuminance is obtained through an integrated silicon photodiode sensor. The gray-scale mean and dynamic compensation coefficient of the image are calculated. The image is corrected pixel by pixel. The gray-scale variance of the corrected image and the proportion of overexposed and underexposed pixels are calculated. If the gray-scale variance or the proportion of overexposed / underexposed pixels exceeds the set threshold, the reshoot mechanism is triggered. S2: Concrete surface preprocessing, noise removal from the acquired image, followed by adaptive histogram equalization (CLAHE) to enhance contrast, calculation of the signal-to-noise ratio (SNR) and the mean gray-level jump value at the block boundary of the preprocessed image. If the SNR or the mean gray-level jump value at the block boundary does not meet the requirements, return to S1. S3: Concrete crack edge detection, construct gradient field to extract edge direction and intensity, generate dynamic threshold to adaptively distinguish edges, enhance edge continuity through fracture repair algorithm, calculate continuity index, if the continuity index or edge pixel ratio does not meet the requirements, return to S2; S4: Crack geometric feature quantization. The length of the crack edge after skeletonization is calculated by segment integration. The width is calculated by extracting the gray function along the crack normal. If the crack length or average width exceeds the set threshold, an alarm is triggered. S5: Quality Classification and Treatment. Based on crack length, average width, and distribution density, and in accordance with the requirements of bridge structural safety specifications, the quality status of concrete cracks is divided into five levels, and corresponding treatment measures are taken according to different levels.

[0006] In one possible implementation, during the S1 image acquisition and dynamic normalization step, the UAV flies along a 1.5m × 1.5m grid path at a flight altitude of... The ambient light sensor has a spectral response range of 400-760nm, with an overlap rate of 15% between adjacent grids and a range of 2.0±0.1m.

[0007] In one possible implementation, in the S2 concrete surface preprocessing step, median filtering uses a 5×5 window, the CLAHE algorithm divides the image into 8×8 sub-blocks, the contrast limit threshold is 40, and local texture features are enhanced.

[0008] In one possible implementation, the S3 concrete crack edge detection step uses a 3×3 weighted sliding window to calculate the horizontal and vertical gradients, with the gradient magnitude and direction... Discretized into four main directions: 0°, 45°, 90°, and 135°, the dynamic threshold is calculated by excluding flat regions with gradient values ​​less than 10.

[0009] In one possible implementation, in the S4 crack geometric feature quantization step, the crack length is calculated using the skeleton point piecewise integration method, and the crack width is calculated using the 90% energy width method of the normal grayscale function. If the crack length... >1000mm or average width If the diameter is greater than 1.0 mm, a level 3 alarm will be triggered.

[0010] In one possible implementation, the quality grading and handling steps of S5 divide the crack quality status into five levels: normal state is set as level I, minor defect state is set as level II, general defect state is set as level III, serious defect state is set as level IV and dangerous state is set as level V, and different handling measures are taken according to different levels. For defects of level III and above, a re-inspection process is automatically triggered.

[0011] In one possible implementation, the formula for calculating the crack length is: ,in, =0.12mm / pixel is the camera calibration factor. , The difference in coordinates between adjacent skeleton points. This represents the number of skeleton points.

[0012] In one possible implementation, the formula for calculating the crack width is: ,in, The number of sampling points. For the first The energy width of each sampling point.

[0013] In one possible implementation, the bridge concrete quality monitoring system based on edge detection includes: The image acquisition and dynamic standardization module uses a drone equipped with a multispectral camera to acquire images of the bridge concrete surface and performs dynamic standardization processing to ensure that the image quality meets the requirements of subsequent processing. This module includes an image acquisition unit, an illumination compensation unit, and a re-shot triggering unit. The concrete surface preprocessing module performs noise removal and contrast enhancement on the acquired images to highlight crack features. This module includes a noise filtering unit and a contrast enhancement unit. The concrete crack edge detection module extracts crack edges through gradient field analysis and dynamic threshold segmentation, and enhances edge continuity to improve the accuracy of crack detection. This module includes a gradient calculation unit, a threshold generation unit, and an edge repair unit. The crack geometry feature quantification module performs quantitative analysis on the geometric features of cracks, calculates the length and width of cracks, and is used to assess the severity of cracks. This module includes a length calculation unit and a width calculation unit. The quality grading and treatment module classifies the quality status of cracks based on their geometric characteristics and takes corresponding treatment measures to achieve automated management and maintenance of cracks. This module includes a grading judgment unit and a treatment execution unit.

[0014] Beneficial effects compared to existing technologies: 1. In this solution, a UAV equipped with a multispectral camera is used to automatically acquire images of the bridge concrete surface. Flying along a grid path, it utilizes a silicon photodiode sensor to obtain real-time ambient illumination. Dynamic compensation coefficients are used to correct the illumination of the images, eliminating overexposure / underexposure. The preprocessing stage employs median filtering and the CLAHE algorithm to effectively preserve edge features while reducing noise interference. Compared to traditional manual inspections or fixed-position shooting, this process significantly improves the uniformity of data acquisition and image quality, while avoiding subjective errors inherent in manual operation. 2. In this scheme, a weighted gradient field calculation and a dual threshold system are combined with a fracture repair algorithm to effectively extract minute cracks. Length quantization uses piecewise integration of skeleton points, and width is calculated using energy width. Based on the quantized length and width, an alarm is automatically triggered. Compared with traditional edge detection algorithms, this invention significantly improves the crack recognition rate against complex texture backgrounds. 3. This solution establishes a five-level grading standard based on crack length, width, and distribution density, and automatically triggers a re-inspection process for defects of level III and above. After treatment, data comparison generates a closed-loop acceptance report. This system quantitatively integrates crack development trends, impact on structural bearing capacity, and the urgency of repair, ensuring that treatment measures match the severity of defects. This significantly improves the efficiency of bridge maintenance decisions and, through distributed storage and real-time alarm mechanisms, achieves full-process traceability management from detection to acceptance. Attached Figure Description

[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0016] Figure 1 This is a schematic diagram of the steps of the quality monitoring method of the present invention; Figure 2 This is a schematic diagram of the quality monitoring system framework of the present invention. Detailed Implementation

[0017] Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the present invention can also be implemented in various different forms, and therefore the present invention is not limited to the embodiments described below. In addition, for the purpose of more clearly describing the present invention, parts not connected to the invention will be omitted from the drawings. The technical solutions in this application are designed to address the problems described in the background, and are generally as follows: Example 1: This embodiment introduces a bridge concrete quality monitoring method based on edge detection, including the following steps: S1 Image Acquisition and Dynamic Standardization Images of the bridge's concrete surface were captured by a drone equipped with a multispectral camera, which operated at a constant altitude. =2.0 Flying along the grid path at a distance of 0.1m, with a grid lateral spacing. =1.5m, longitudinal spacing =1.5m. This path is generated by the UAV flight control system according to preset grid coordinates, ensuring 100% coverage of the bridge surface and controlling the overlap rate of adjacent grids to 15% to avoid splicing errors. Multispectral cameras simultaneously acquire raw images. Ambient illuminance is obtained through an integrated silicon photodiode sensor. The unit is The sensor has a response spectrum range of 400-760nm.

[0018] Calculate the average gray level of the image The formula is: ,in, =4096 2160 = 8847360, which represents the total number of pixels. For pixels The grayscale value ranges from 0 to 255. Based on and Calculate the dynamic compensation coefficient : Where 10000 is the illumination normalization constant, and 128 is the median of the 8-bit grayscale image. The pixel-by-pixel correction formula is: Real-time correction is achieved through GPU parallel computing.

[0019] Calculate the grayscale variance of the corrected image The formula is Meanwhile, overexposure statistics were also compiled. >240 and underexposure <10 pixels in proportion. If If the number of overexposed / underexposed pixels exceeds 150 or the percentage of overexposed / underexposed pixels exceeds 5%, a reshoot mechanism is triggered: the drone returns along the original path, adjusts the camera's ISO value (range 100-800) or shutter speed (1 / 1000-1 / 30s), and re-captures the image, up to a maximum of 3 reshoots. If the image is still unsatisfactory after 3 reshoots, the current coordinates are marked. Skip that area and continue with the detection of adjacent areas. Skipped areas should be marked in red in the final report.

[0020] S2 concrete surface pretreatment right First, a 5×5 median filter is used to remove salt-and-pepper noise. This algorithm traverses the image through a sliding window, sorts the grayscale values ​​of the 25 pixels within the window, and replaces the center pixel with the median value. Specifically, if a window contains a noise point of 255 (overexposed), the median value after sorting is 120, effectively eliminating noise while preserving edges. The filtering process is optimized using SIMD instructions.

[0021] Contrast-limited adaptive histogram equalization (CLAHE) was then used to enhance contrast. Specifically, the image was divided into 8×8 sub-blocks, and the contrast threshold within each sub-block was limited to 40 during histogram equalization. Contrast exceeding this threshold was evenly distributed to other grayscale ranges, and bilinear interpolation was used to process the sub-block boundaries. Taking a crack region as an example, the original grayscale difference was 20, which was increased to 45 after enhancement, significantly improving the distinction between the crack and the background. The local contrast enhancement coefficient after CLAHE processing is shown in the figure. Defined as: ,in, , These represent the standard deviations of the grayscale values ​​of the sub-blocks before and after processing.

[0022] Calculate the signal-to-noise ratio of the preprocessed image The formula is: ,in, This is the average grayscale value after preprocessing. This represents the grayscale variance. Simultaneously, the mean grayscale jump at the block boundaries is calculated: the image is divided into 16×16 sub-blocks, and the average absolute value of the grayscale difference is calculated for a 3-pixel wide region representing the common boundary between adjacent sub-blocks. If... If the value is less than 35dB or mean_jump is greater than 15, return to S1 to retake the shot. During the retake, adjust the flight altitude to... =1.8 0.1m, change the imaging scale to reduce the impact of noise.

[0023] S3 Concrete Crack Edge Detection Before detecting concrete cracks, the edge direction and intensity are extracted by constructing a gradient field. Then, a dynamic threshold is generated to adaptively distinguish the salience of the edges. Next, a fracture repair algorithm is used to enhance the edge continuity. Finally, the detection effect is verified by a judgment index, as follows: Constructing the gradient field: Calculating the horizontal gradient using a 3×3 weighted sliding window. and vertical gradient The formula for calculating the horizontal gradient is: ,in, The values ​​represent pixel positions within the window, and the coefficients are designed to be centrally symmetric to enhance the response at horizontal edges. The vertical gradient calculation formula is as follows: gradient magnitude ,direction Discretized into four main directions: 0°, 45°, 90°, and 135°, with each direction corresponding to a gradient histogram channel.

[0024] Generate dynamic thresholds: exclude gradient values Calculate the effective average gradient in a flat region of 10. : , >10, of which The number of valid blocks, This represents the average gradient value across all blocks. High threshold. ,in To maximize the gradient after excluding the top 0.1% outliers, 0.6 and 0.4 are weighting coefficients. Low threshold. This forms a dual-threshold system, specifically, when =200、 When =50, =140, =28.

[0025] Fracture repair algorithm: For weak edge points, i.e. < < Along the direction Check the points within ±3 pixels of the gradient extension line. If strong edge points exist, i.e. and gradient difference Then, weak edge points are upgraded. Specifically, weak edge points... =80, =140, =28, a neighborhood exists. If a strong edge point with a gradient of 120 has a gradient difference of 40 < 42, then upgrade that point.

[0026] For fracture gap A 3-pixel edge segment, if the directions at both ends differ... At a 15° angle, the fill extends along the direction to fill the middle pixels. Specifically, the gap... When the pixel value is 2 pixels and the directional difference is 10°, continuous edges are generated by linear interpolation.

[0027] Calculate continuity index ,like If the median filter value is less than 0.85 or the edge pixel percentage is greater than 30%, return to S2 to adjust the preprocessing parameters: change the median filter window to 7×7 or adjust the CLAHE contrast limit to 30, reprocess, and then execute S3. If the results are still unsatisfactory after two consecutive adjustments, trigger manual intervention and check whether there is moisture or stains on the concrete surface.

[0028] S4 Crack Geometric Feature Quantization Length calculation: Segmental integration over the skeletonized crack edge: ,in, =0.12mm / pixel is the camera calibration coefficient, calibrated on-site using a 200mm×200mm calibration board. , The difference in coordinates between adjacent skeleton points. This refers to the number of skeleton points. Specifically, for a crack containing 100 skeleton points, the average coordinate difference between adjacent points is... pixels, total length mm.

[0029] Width calculation: Extracting grayscale function along crack normal 90% energy width satisfy = = ,but Average width: ,in, For the first The energy width of each sampling point The number of sampling points is 10% of the crack length; for example, 10 points are taken for a 100mm crack. This represents the maximum grayscale value of the normal profile. (A sampling point) =200, =-2 pixels, =2 pixels, then =4 pixels, corresponding to an actual width of 0.48mm.

[0030] like >1000mm or If the measurement exceeds 1.0mm, a Level 3 alarm is triggered: an audible and visual alarm is activated, an engineer is notified via SMS, the system automatically generates a work order (including coordinates, length, and width), and the detection is paused. If the measurement does not exceed the limit, the detection continues, and the progress is automatically saved to the distributed file system every 100 grid areas to ensure that no data is lost in the event of a power outage.

[0031] S5 Quality Classification and Handling Crack length obtained based on S4 step quantization Average width and distribution density Based on the requirements of bridge structural safety specifications, the quality status of concrete cracks is divided into five levels. This classification system comprehensively considers the degree of impact of cracks on the structural bearing capacity, their development trend, and the urgency of repair, ensuring that the treatment measures are matched to the severity of the defects.

[0032] (1) Normal state (Level I) Length of a single crack 100mm, average width 0.1 mm, distribution density 0.5m / ,in, , The total length of the regional cracks, The area to be tested.

[0033] These types of cracks are mostly shrinkage microcracks generated during the hardening process of concrete. They do not penetrate the concrete cover and their impact on structural stiffness and durability is negligible. The proportion of edge pixels in the grayscale image is less than 5%, indicating that the crack characteristics are extremely indistinct.

[0034] For this type of crack: A crack development database is established every quarter using 3D laser scanning, and crack propagation trends are monitored by comparing historical data. Only routine maintenance (such as surface dust removal and cleaning of debris) is required, and no engineering intervention is needed. Because the cracks do not form a through channel, water vapor and corrosive media cannot easily penetrate the interior.

[0035] (2) Minor defect status (Level II) Single crack length 100mm 300mm or average width 0.1mm 0.3 mm, distribution density 0.5 m / 1.5m / .

[0036] These cracks have extended to the middle of the protective layer but have not reached the reinforcing steel. If left untreated, moisture may seep through the cracks to the surface of the steel, potentially causing corrosion over time. (Preprocessed image signal-to-noise ratio) 35dB indicates that the crack features are clear but have not formed a network distribution.

[0037] For this type of crack: Low-viscosity epoxy resin sealant is used, and its viscosity... 500 Surface sealing is performed, and the construction temperature is controlled at 15-25℃ (to ensure the fluidity of the colloid) and the humidity is <85% (to prevent moisture from affecting the bonding). Vehicles are prohibited from loading the sealant within 72 hours after treatment, allowing it to fully cure. A follow-up inspection should then be conducted to determine the bond strength between the sealant and the concrete. 2.5MPa, blocking the water vapor permeation path by sealing the cracks.

[0038] (3) General defect status (Level III) Single crack length 300mm 500mm or average width 0.3mm 0.5mm, distribution density 1.5m / 3.0m / .

[0039] These cracks have penetrated the concrete cover, exposing the reinforcing steel to the external environment and significantly increasing the risk of corrosion. The weak edges account for 15%-25% of the gradient field, indicating that the cracks have some continuity but are not fully connected.

[0040] For this type of crack: Pressure grouting is used, with the pressure set at 0.3-0.5 MPa. Modified polyurethane grouting material is injected, and the spacing between grouting holes is 200-300 mm. The specific spacing is adjusted according to the crack density. Before construction, a trial grouting should be performed in a typical 1m² area to record the optimal grouting pressure and time (based on grout overflow from the outlet without air bubbles) to ensure that the grout fills the full depth of the cracks. Core sampling is required to test the grout density. 95% of the concrete is restored through grouting, which also isolates the reinforcing steel from the outside environment.

[0041] (4) Critical defect status (Level IV) Single crack length 500mm 1000mm or average width 0.5mm 1.0 mm, distribution density 3.0 m / s 5.0m / .

[0042] These cracks have penetrated most of the cross-section; steel reinforcement corrosion has caused volume expansion, further exacerbating concrete cracking and reducing structural stiffness. Edge detection continuity index: 0.85 A value of 0.9 indicates that the cracks have formed a continuous channel but have not yet completely destroyed the structural bearing capacity.

[0043] For this type of crack: Reinforcement with bonded carbon fiber cloth; standard value of tensile strength of the cloth. 3400MPa, elastic modulus 2.4× MPa, reinforcement range extending 100mm beyond the crack edge, overlap length of each layer of carbon fiber cloth 150mm; The matching resin adhesive must meet the requirements for positive tensile bond strength. The stress was 2.5 MPa, and the failure was cohesive within the concrete. The high strength of the carbon fiber fabric was used to distribute the load. Finite element analysis required that the stress intensity factor at the crack tip be reduced after reinforcement. 40%, preventing the cracks from expanding further.

[0044] (5) Dangerous state (Level V) Length of a single crack 1000mm or average width 1.0 mm, distribution density 5.0m / .

[0045] Such cracks cause a decrease in the effective load-bearing capacity of the structural cross-section by more than 30%, and may result in visible deformation or abnormal vibration. An overexposed / underexposed pixel ratio greater than 10% in the corrected image indicates that the cracks have caused significant changes in surface morphology, posing a sudden safety risk.

[0046] For this type of crack: Traffic control measures will be implemented immediately, and load limits will be imposed. 5t, equipped with audible and visual warning devices; External prestressing reinforcement or additional supports are adopted, and a real-time monitoring system is deployed during construction. The treatment plan must be reviewed by the original design unit and approved by experts. After reinforcement, a static load test must be conducted, requiring the load-bearing capacity to be restored to more than 95% of the design value to ensure structural safety.

[0047] If S5 is determined to be Level III or above, the system automatically triggers the S1-S4 re-inspection process, reducing the grid spacing to 1.0m and adjusting the flight altitude to 1.5m. 0.1m, and simultaneously adjust the S3 dual threshold to , This enhances the ability to identify minute cracks.

[0048] After the treatment is completed, S1-S4 will be repeated. If the re-inspection results are downgraded to Level II or below, the crack length, width, and distribution density data before and after the treatment must be compared to form a closed-loop acceptance report. If it is still maintained at Level III or above, the treatment is deemed a failure, and a third-party testing agency must be brought in to re-evaluate and adjust the plan.

[0049] Specifically, inspection of the prestressed concrete box girder bridge revealed that the main span web had... =800mm =0.6mm crack, distribution density =3.2m / m², judged as a Class IV defect. After reinforcement with carbon fiber cloth, a re-inspection showed that the crack width had been reduced to 0.2mm and the length to 450mm, and the level was downgraded to Class III. Subsequent treatment was completed through pressure grouting. At the final acceptance, all indicators met the Class II standard, and the machine was restored to normal use.

[0050] Example 2: This embodiment introduces a bridge concrete quality monitoring system based on edge detection, including an image acquisition and standardization module, an image preprocessing module, a crack edge detection module, a crack quantification analysis module, and a quality grading and treatment module. These modules work closely together to ensure the accuracy and efficiency of bridge concrete crack monitoring.

[0051] I. Standardized Image Acquisition Module This module utilizes an UAV-borne multispectral imaging system to automatically acquire images of the bridge's concrete surface and perform illumination compensation, ensuring spatial coverage and photometric consistency of the original data. The UAV flies along a 1.5m × 1.5m grid path, maintaining an altitude of 2.0 ± 0.1m with a 15% overlap between adjacent grids. It simultaneously acquires ambient illuminance using a silicon photodiode sensor with a spectral response of 400-760nm, providing a basis for subsequent dynamic correction. This module includes an image acquisition unit and an illumination compensation unit.

[0052] 1. Image acquisition unit The flight control system generates a grid coordinate path, monitors flight altitude and positioning in real time, and controls a multispectral camera to acquire raw images at 1.5m horizontal / vertical intervals. After acquisition, the total number of pixels (8,847,360) and the average grayscale value are calculated. The average grayscale value is calculated using the grayscale average value formula in Example 1 to establish an initial image database.

[0053] If the flight altitude deviation exceeds ±0.15m or the grid overlap rate is less than 12%, the intra-cell path replanning mechanism is triggered, and flight parameters are adjusted before data acquisition is repeated. If three consecutive acquisitions still fail to meet the positioning accuracy (i.e., grid spacing error > 0.2m), the current area coordinates are marked and skipped, and the data flows into the anomaly processing queue for manual re-measurement. If the acquisition parameters meet the standards, i.e., the altitude deviation is satisfied... With a resolution of 0.1m and an overlap rate of 15%±2%, the original image and illuminance data are synchronously transmitted to the illumination compensation unit.

[0054] 2. Illumination compensation unit Based on the ambient illuminance value and the average gray value, the dynamic compensation coefficient is calculated using the dynamic compensation coefficient formula as described in Example 1. Pixel-by-pixel correction of the original image is achieved through GPU parallel computing. After correction, the grayscale variance is calculated using the grayscale variance formula in Example 1, and the percentage of overexposed pixels (grayscale > 240) and underexposed pixels (grayscale < 10) is statistically analyzed.

[0055] If the grayscale variance > 150 or the overexposed / underexposed pixel ratio > 5%, the camera will automatically adjust its parameters: ISO will be adjusted between 100-800, or shutter speed will be switched between 1 / 1000-1 / 30s, with a maximum of 3 retakes. If the performance still does not meet the requirements after 3 attempts (variance...), the camera will automatically adjust its parameters. 150 and percentage If the correction is successful (5%), the area is marked as an "abnormal illumination area," skipped, and highlighted in red in the final report; if the correction is successful, the standardized image flows into the image preprocessing module.

[0056] II. Image Preprocessing Module This module improves image quality through median filtering and contrast enhancement algorithms, highlighting the grayscale difference between cracks and the background. First, a 5×5 window median filter is used to eliminate salt-and-pepper noise. Then, the CLAHE algorithm divides the image into 8×8 sub-blocks, limiting the contrast threshold to 40 to enhance local texture features. This module includes a noise removal unit and a contrast enhancement unit.

[0057] 1. Noise filtering unit A median filtering algorithm optimized using SIMD instructions is employed. The median value of each 25 pixels within a 5×5 sliding window is sorted and used to replace the center pixel. For example, the overexposed noise pixel 255 is replaced with the median value 120, preserving edge details while reducing noise interference. After filtering, edge blur and noise density are calculated.

[0058] If the noise density is greater than 5% or the edge blurring is greater than 3 pixels, automatically switch to a 7×7 filtering window for reprocessing. If large-area texture distortion still exists after two filtering steps, such as loss of concrete aggregate features, return to the image acquisition standardization module, adjust the flight altitude to 1.8±0.1m, and reduce the noise impact by decreasing the imaging scale. If the noise density after filtering... 3% and edge retention rate 90% of the image enters the contrast enhancement unit.

[0059] 2. Contrast enhancement unit The CLAHE algorithm is used to divide the image into 8×8 sub-blocks. Histogram equalization with a contrast limit of 40 is performed on each sub-block, and bilinear interpolation is used to process the sub-block boundaries. Taking the crack region as an example, the grayscale difference can be increased from 20 to 45 after processing. Simultaneously, the local contrast enhancement coefficient is calculated using the enhancement coefficient formula in Example 1. .

[0060] Through Example 1 Formula for calculating the signal-to-noise ratio of a preprocessed image The mean_jump of the grayscale jump at the block boundary. If If the value is <35dB or mean_jump>15, return to the image acquisition module, adjust the flight altitude to 1.8±0.1m and retake the image; if the target value is still not met after retaking the image, i.e. 35dB and mean_jump 15) Trigger the manual inspection process to check for dampness or stains on the concrete surface. If the preprocessing is satisfactory, the image flows into the edge detection module.

[0061] III. Crack Edge Detection Module This module extracts crack edges through gradient field analysis and dynamic threshold segmentation, and improves edge continuity by combining it with a fracture repair algorithm. First, it calculates the horizontal / vertical gradients within a 3×3 window, generating histograms of gradient magnitude and direction. Then, it distinguishes between strong and weak edges using a dual threshold system, and finally performs direction matching repair on the fracture edge segments. This module includes a gradient calculation unit, a threshold generation unit, and an edge repair unit.

[0062] 1. Gradient Calculation Unit Using a weighted sliding window, the horizontal gradient is calculated using the horizontal gradient formula from Example 1. The vertical gradient is calculated using the vertical gradient formula in Example 1. ,pass Obtain the gradient magnitude and then the direction. Discretize the data into four main directional channels: 0°, 45°, 90°, and 135°, and construct the gradient feature matrix.

[0063] If there are valid points in the gradient magnitude If the percentage of valid points is greater than 10% and less than 20%, it indicates that the image texture is too smooth, possibly lacking cracks or having insufficient resolution. Return to the image preprocessing module, reduce the CLAHE contrast limit from 40 to 30, and recalculate the gradient. 20% of the gradient data flows into the threshold generation unit.

[0064] 2. Threshold generation unit exclude For a flat region of 10, the effective average gradient is calculated using the effective average gradient formula as described in Example 1. Combined with the maximum gradient after removing the first 0.1% outliers Generate a dual threshold: high threshold low threshold .

[0065] like >200 or A value <10 indicates abnormal lighting or severe noise interference; the system will return to the lighting compensation unit to recalculate the dynamic compensation coefficient. If the threshold remains abnormal after correction, i.e. If the threshold is greater than 200, the drone re-shooting mechanism will be triggered. If both thresholds are within a reasonable range, i.e. 200 and 10. Threshold parameters are passed to the edge repair unit.

[0066] 3. Edge Repair Unit For weak edge points, i.e. < < Along the direction examine Strong edge points within a 3-pixel neighborhood If a weak edge point exists and the gradient difference is less than 42, then upgrade the weak edge point; for fracture gaps 3 pixels and directional difference The 15° edge segment was filled using linear interpolation. After repair, the continuity index was calculated using the continuity index formula from Example 1. .

[0067] like If the median filter value is less than 0.85 or the edge pixel percentage is greater than 30%, return to the image preprocessing module, change the median filter window to 7×7 or adjust the CLAHE contrast limit to 30, reprocess, and then perform edge detection; if the conditions are still not met after two consecutive adjustments... 0.85 and proportion 30% is marked as a "complex texture area," and the manual-assisted edge recognition process is initiated. If the repair is satisfactory, the edge detection results flow into the crack quantification analysis module.

[0068] IV. Crack Quantitative Analysis Module This module calculates the geometric parameters of the skeletonized crack edges, quantifying the length and width, and triggering graded alarms based on thresholds. The length calculation uses a piecewise integration method based on skeleton points, while the width is determined by the 90% energy width of the normal grayscale function, converted to the actual physical size using a camera calibration coefficient (0.12mm / pixel). This module includes a length calculation unit and a width calculation unit.

[0069] 1. Length Calculation Unit Piecewise integration was performed on the skeletonized crack edge using the length calculation formula in Example 1, and the coordinate difference between adjacent skeleton points was used. Calculate the pixel distance and multiply it by a calibration factor to obtain the actual length. An example is given with 100 skeleton points and the average coordinate difference. The crack in the pixel has a total length of approximately 26.9 mm.

[0070] If the length of a single crack If the length is greater than 1000mm, immediately trigger a Level 3 alarm: audible and visual alarm, SMS notification to the engineer, generation of a work order with coordinates, and system pause monitoring; if the length is less ... 1000mm, data flows into the width calculation unit. If the length calculation results in a broken skeleton point, for example, the distance between adjacent points is greater than 5 pixels, it returns to the edge repair unit for reprocessing.

[0071] 2. Width Calculation Unit Extract the grayscale function along the crack normal and calculate the result. = = The average width is calculated by taking 10% of the sample points representing 90% of the crack length, based on the energy width. .

[0072] like >1000mm or If the error exceeds 1.0 mm, a three-level alarm process is initiated. If the error does not exceed the limit, data for every 100 grid areas is automatically saved to the distributed file system to ensure no data loss during power outages. If abnormal skeleton point coordinates are found during saving, the process returns to the edge detection module to re-extract the edges. If the quantization data is complete, the geometric parameters flow into the quality grading and processing module.

[0073] V. Quality Classification and Handling Module This module classifies concrete quality into five levels (Levels I-V) based on crack length, width, and distribution density, in accordance with bridge specifications. It automatically matches treatment plans and triggers a closed-loop management process. For Level III and above defects, a re-inspection mechanism is initiated, and an acceptance report is generated by comparing data after treatment. This module includes a grading determination unit and a treatment execution unit.

[0074] 1. Grading and Determination Unit This unit is based on the length of a single crack. Average width Distribution density Classification: Level I: 100mm and 0.1mm and 0.5m / Shrink microcracks, with edge pixels accounting for less than 5%; Grade II: 100mm 300mm or 0.1mm 0.3mm or 0.5m / 1.5m / Crack in the middle of the protective layer; Level III: 300mm 500mm or 0.3mm 0.5mm or 1.5m / 3.0m / It penetrated the protective layer crack; Level IV: 500mm 1000mm or 0.5mm 1.0mm or 3.0m / 5.0m / A crack that penetrates the cross-section; Level V: 1000mm or 1.0mm or 5.0m / The structural bearing capacity decreased by more than 30%.

[0075] If the classification is III-V, the re-inspection process will be automatically initiated: the grid spacing will be reduced to 1.0m, and the flight altitude will be adjusted to 1.5m. 0.1m, edge detection double threshold changed to , If it is determined to be Level I or II, it will be directly transferred to the disposal execution unit.

[0076] 2. Execution Unit This unit implements corresponding measures based on the classification results, and then re-inspects the entire process after the treatment. If the re-inspection result is downgraded to Level II or below, a closed-loop acceptance report is generated by comparing the data before and after the treatment; if it is still Level III or above, the treatment is deemed a failure, and a third-party testing agency is brought in for reassessment.

[0077] If the crack indicators worsen during re-inspection, for example, the length increases by 20% or the width increases by 0.1mm, the system will automatically upgrade the treatment level and submit it to the original design unit for review; if the indicators meet the standards after treatment, for example, level IV is downgraded to level III, subsequent treatment will continue until level II standards are met; if multiple treatments are ineffective, the expert review process will be initiated.

[0078] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for monitoring the quality of bridge concrete based on edge detection, characterized in that, Includes the following steps: S1: Image acquisition and dynamic standardization. The drone is equipped with a multispectral camera to acquire images of the bridge concrete surface. The ambient illuminance is obtained through an integrated silicon photodiode sensor. The gray-scale mean and dynamic compensation coefficient of the image are calculated. The image is corrected pixel by pixel. The gray-scale variance of the corrected image and the proportion of overexposed and underexposed pixels are calculated. If the gray-scale variance or the proportion of overexposed / underexposed pixels exceeds the set threshold, the reshoot mechanism is triggered. S2: Concrete surface preprocessing, noise removal from the acquired image, followed by adaptive histogram equalization (CLAHE) to enhance contrast, calculation of the signal-to-noise ratio (SNR) and the mean gray-level jump value at the block boundary of the preprocessed image. If the SNR or the mean gray-level jump value at the block boundary does not meet the requirements, return to S1. S3: Concrete crack edge detection, construct gradient field to extract edge direction and intensity, generate dynamic threshold to adaptively distinguish edges, enhance edge continuity through fracture repair algorithm, calculate continuity index, if the continuity index or edge pixel ratio does not meet the requirements, return to S2; S4: Crack geometric feature quantization. The length of the crack edge after skeletonization is calculated by segment integration. The width is calculated by extracting the gray function along the crack normal. If the crack length or average width exceeds the set threshold, an alarm is triggered. S5: Quality Classification and Treatment. Based on crack length, average width, and distribution density, and in accordance with the requirements of bridge structural safety specifications, the quality status of concrete cracks is divided into five levels, and corresponding treatment measures are taken according to different levels.

2. The bridge concrete quality monitoring method based on edge detection as described in claim 1, characterized in that, In the S1 image acquisition and dynamic standardization step, the UAV flies along a 1.5m × 1.5m grid path at a flight altitude of [missing information]. The ambient light sensor has a spectral response range of 400-760nm, with an overlap rate of 15% between adjacent grids and a range of 2.0±0.1m.

3. The bridge concrete quality monitoring method based on edge detection as described in claim 1, characterized in that, In the S2 concrete surface preprocessing step, median filtering uses a 5×5 window, the CLAHE algorithm divides the image into 8×8 sub-blocks, the contrast limit threshold is 40, and local texture features are enhanced.

4. The bridge concrete quality monitoring method based on edge detection as described in claim 1, characterized in that, In the S3 concrete crack edge detection step, a 3×3 weighted sliding window is used to calculate the horizontal and vertical gradients, with gradient magnitude and direction... Discretized into four main directions: 0°, 45°, 90°, and 135°, the dynamic threshold is calculated by excluding flat regions with gradient values ​​less than 10.

5. The bridge concrete quality monitoring method based on edge detection as described in claim 1, characterized in that, In the S4 crack geometric feature quantification step, the crack length is calculated using the skeleton point piecewise integration method, and the crack width is calculated using the 90% energy width method of the normal gray-level function. If the crack length... >1000mm or average width If the diameter is greater than 1.0 mm, a level 3 alarm will be triggered.

6. The bridge concrete quality monitoring method based on edge detection as described in claim 1, characterized in that, In the S5 quality grading and handling steps, the crack quality status is divided into five levels: normal status is set as level I, minor defect status is set as level II, general defect status is set as level III, serious defect status is set as level IV, and dangerous status is set as level V. Different handling measures are taken according to different levels, and the re-inspection process is automatically triggered for defects of level III and above.

7. The bridge concrete quality monitoring method based on edge detection as described in claim 5, characterized in that, The formula for calculating the crack length is: ,in, =0.12mm / pixel is the camera calibration factor. , The difference in coordinates between adjacent skeleton points. This represents the number of skeleton points.

8. The bridge concrete quality monitoring method based on edge detection as described in claim 5, characterized in that, The formula for calculating the crack width is: ,in, The number of sampling points. For the first The energy width of each sampling point.

9. A bridge concrete quality monitoring system based on edge detection, performing the method as described in any one of claims 1 to 8, characterized in that, include: The image acquisition and dynamic standardization module uses a drone equipped with a multispectral camera to acquire images of the bridge concrete surface and performs dynamic standardization processing to ensure that the image quality meets the requirements of subsequent processing. This module includes an image acquisition unit, an illumination compensation unit, and a re-shot triggering unit. The concrete surface preprocessing module performs noise removal and contrast enhancement on the acquired images to highlight crack features. This module includes a noise filtering unit and a contrast enhancement unit. The concrete crack edge detection module extracts crack edges through gradient field analysis and dynamic threshold segmentation, and enhances edge continuity to improve the accuracy of crack detection. This module includes a gradient calculation unit, a threshold generation unit, and an edge repair unit. The crack geometry feature quantification module performs quantitative analysis on the geometric features of cracks, calculates the length and width of cracks, and is used to assess the severity of cracks. This module includes a length calculation unit and a width calculation unit. The quality grading and treatment module classifies the quality status of cracks based on their geometric characteristics and takes corresponding treatment measures to achieve automated management and maintenance of cracks. This module includes a grading judgment unit and a treatment execution unit.

Images