Drainage pipeline defect evaluation method based on double-source input and Hough transform

Abstract

The invention discloses a drainage pipeline defect assessment method based on double-source input and Hough transform. The method comprises the following steps: S1, synchronously acquiring an original pipeline image and a defect mask image generated by a semantic segmentation network, and preprocessing; s2, selecting a reference image based on HSV color space analysis; s3, generating a candidate circle set, combining the position weight and the confidence coefficient weight to obtain a candidate circle selection weight, performing optimization screening, and generating a reference circle; s4, performing expansion adjustment on the radius of the reference circle according to the geometrical characteristics of the defect area to obtain an adjusted reference circle; s5, calculating a defect geometric parameter based on the adjusted reference circle, and calculating a defect area proportion or a deviation distance parameter; and S6, performing quantitative evaluation on the defect level. According to the method, the circle fitting precision in a pipeline image scene is effectively improved, the accuracy of functional defect evaluation is remarkably improved, the problem of parameter distortion caused by noise interference in a traditional method is solved, and the method has high engineering application value.

Description

Technical Field

[0001] This invention belongs to the field of pipeline defect detection technology, specifically relating to a method for evaluating defects in drainage pipelines based on dual-source input and Hough transform. Background Technology

[0002] As a crucial component of urban infrastructure, the health of drainage pipelines directly impacts the safe operation and flood control capabilities of urban drainage systems. With accelerating urbanization, the number and complexity of drainage pipelines are constantly increasing, making traditional manual inspection methods insufficient for efficient and accurate detection. In recent years, automatic defect detection technology for drainage pipelines based on image processing and artificial intelligence has gradually become a research hotspot. For example, semantic segmentation networks (such as U-Net) are used to extract defect regions, combined with deep learning algorithms to achieve defect classification. However, existing methods still have the following shortcomings in defect quantitative assessment: 1. The calculation of mask pixel ratio has significant limitations. Traditional methods calculate area loss by statistically analyzing the defect mask pixel ratio, but this is easily affected by noise points, edge blurring, and image resolution, leading to parameter distortion; 2. Insufficient accuracy in fitting the reference circle. Defect assessment requires calculating geometric parameters (such as cross-sectional area changes) based on the reference circle of the pipeline cross-section, but existing fitting methods often rely on a single image source (such as the original image), making it difficult to consider both the geometric features of the defect region and mask information, resulting in fitting bias.

[0003] According to the "Technical Specification for Inspection and Evaluation of Urban Drainage Pipelines" (CJJ 181-2012) and other requirements, defects in drainage pipelines need to be classified into structural defects (such as ruptures, deformations, and misalignments) and functional defects (such as sedimentation and blockages), and graded by preset thresholds. However, the quantitative analysis of defects such as misalignments and deformations in existing evaluation methods still relies on manual experience and lacks a unified mathematical model to support it.

[0004] In summary, existing drainage pipeline defect detection technologies still have shortcomings in terms of automation and accuracy. There is an urgent need for a defect assessment method that combines multi-source data input and optimized geometric modeling to solve the problems of noise interference, edge blurring and parameter distortion, thereby improving the reliability and engineering applicability of pipeline defect level assessment. Summary of the Invention

[0005] The purpose of this invention is to provide a method for evaluating defects in drainage pipes based on dual-source input and Hough transform, which aims to solve the problem of defect quantification deviation caused by image noise, blurred defect edges, and insufficient accuracy of reference circle fitting in the prior art.

[0006] To achieve the above objectives, the method proposed in this invention includes the following steps: A method for evaluating defects in drainage pipes based on dual-source input and Hough transform, characterized by the following steps: S1, synchronously acquire the original pipeline image and the defect mask image generated by the semantic segmentation network, and perform preprocessing; S2. Based on HSV color space analysis, a reference image is selected. The defect mask image is used first for reference circle positioning. If no circle is detected in the defect mask image, the original image is used for reference circle positioning. S3. A candidate circle set is generated by the Hough circle transform algorithm. The candidate circle selection weight is obtained by combining the position weight and confidence weight and then optimized and filtered to generate the baseline circle. S4. The radius of the reference circle is expanded and adjusted according to the geometric characteristics of the defect area to obtain the adjusted reference circle; S5, calculate the defect geometric parameters based on the adjusted reference circle, and calculate the defect area ratio or deviation distance parameter; S6. Based on the results of step 5, the defect level is quantitatively evaluated using a preset threshold.

[0007] Compared with the prior art, the present invention has the following beneficial effects: (1) By introducing dual-source input and weighted screening strategy, the fitting accuracy of the pipe cross-section circle is significantly improved, overcoming the parameter distortion problem caused by noise and edge blurring in traditional methods; (2) By introducing a dynamic radius expansion mechanism, the algorithm adapts to the boundary changes of different morphological defects, thereby improving its adaptability to complex defect scenarios; (3) Establish a defect quantification model based on geometric parameters using existing data and experimental data, and output the defect level in combination with the grading standard to provide a scientific basis for engineering maintenance decisions; (4) The systematic design enables the entire evaluation process to be free from human intervention, greatly improving detection efficiency and consistency, and meeting the needs of intelligent operation and maintenance of urban drainage pipe networks.

[0008] In summary, this invention achieves high-precision cross-sectional circle fitting by fusing the original image with a defect mask image generated from semantic segmentation and combining it with an improved Hough transform algorithm. Furthermore, it calculates the defect area ratio or length deviation based on geometric parameters, effectively improving the accuracy of circle fitting in pipeline image scenarios and significantly enhancing the accuracy of functional defect assessment. It solves the parameter distortion problem caused by noise interference in traditional methods and improves the level of automation, demonstrating high engineering application value. It can be widely applied in fields such as municipal drainage network maintenance, intelligent detection of water conservancy projects, and health status assessment of underground pipelines. Attached Figure Description

[0009] Figure 1 This is a flowchart of the drainage pipeline defect assessment method based on dual-source input and Hough transform according to the present invention.

[0010] Figure 2The image shown is S1 in the embodiment; where the left side is the original image and the right side is the mask image; Figure 3 This refers to the set of candidate circles generated in step S3 of the embodiment. Figure 4 The reference circle generated in step S3 of the embodiment; Figure 5 The parameter results obtained in S4 of the example are shown. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0012] Reference Figure 1 The drainage pipe defect assessment method based on dual-source input and Hough transform provided in this invention specifically includes the following steps: S1: Simultaneously acquire the original pipeline image and the defect mask image generated by the semantic segmentation network, and perform preprocessing.

[0013] The preprocessing specifically involves: using... The median filter removes impulse noise. Gaussian smoothing ( To further suppress high-frequency interference, morphological opening operations are performed using elliptic kernels to effectively remove small connected regions, thereby obtaining smoother edge information.

[0014] S2. Based on HSV color space analysis, a reference image is selected. The defect mask image is used first for reference circle positioning. If no circle is detected in the defect mask image, the original image is used for reference circle positioning.

[0015] This step guides the selection of detection strategies in subsequent processing. Since the mask image has been processed by a pre-trained model, its noise is low and the target area is clearly defined, making it more suitable for initial localization. Therefore, the defect mask image is preferred for baseline circle localization. If a defect area exists in the defect mask image, the defect mask image is selected as the detection baseline; otherwise, the original image is used. If no circle is detected in the defect mask image, the original image is used to detect the cross-sectional circle.

[0016] S3 generates a set of candidate circles using the Hough circle transform algorithm, combines position weights and confidence weights to obtain the candidate circle selection weights, and optimizes and filters them to generate the baseline circle.

[0017] Specifically, it includes the following sub-steps: S31, parameter configuration is performed based on Hough gradient transform to generate a candidate circle set. The Euclidean distance between the center of the candidate circle and the center of the original image is calculated. A Gaussian decay function is applied to generate position weights to enhance the priority of candidate circles in the image center region. The expression is as follows:

[0018] in, The position weights of the candidate circles, The distance from the center of the candidate circle to the center of the original image. is the standard deviation of the Gaussian distribution.

[0019] This weighting mechanism creates a negative correlation between the center offset and the confidence score of the candidate circle; the farther the circle's center is from the image center, the lower its score. This reduces the probability of candidate circles that are off-center from the image center being selected.

[0020] S32 assigns exponentially decaying weights based on the original voting ranking using the Hough transform. This operation aims to reduce interference from low-ranked candidate circles and improve overall fitting stability.

[0021] Specifically, the expression for the exponentially decaying weight of the candidate circle is:

[0022] in, The exponential decay weights of the candidate circles are used as the confidence weights for the candidate circles. The original sorting index for candidate circles. This represents the total number of candidate circles. The circle with the most votes corresponds to... Its confidence weight is the largest.

[0023] S33, Multiply the position weight of the candidate circle by the confidence weight to obtain the selection weight of the candidate circle. The candidate circle with the highest score is selected from the candidate circle selection weight S as the reference circle for calculating the cross-sectional area.

[0024] Specifically, the selection weight of the reference circle The calculation method is as follows:

[0025] in, The adjustment parameter is used to control the influence of position weights on the selection weights of candidate circles.

[0026] S4. The radius of the reference circle is expanded and adjusted according to the geometric characteristics of the defect area to obtain the adjusted reference circle; After initial reference circle selection in S3, the radius of the reference circle is further optimized for the located defect areas to obtain a fitted circle, thereby improving the fitting quality and ensuring that the defect areas are completely enclosed. Specifically, the maximum Euclidean distance between the pixel coordinates of each defect area and the center of the fitted circle is calculated, and a margin coefficient λ is introduced to expand the radius of the reference circle based on this distance. This process can be expressed as:

[0027] in, The radius of the fitted circle, i.e., the radius of the reference circle; These are the pixel coordinates of the defect area. Using the coordinates of the center of the reference circle, The preset margin coefficient has a value range of [1.05, 1.15].

[0028] S5, calculate the defect geometric parameters based on the adjusted reference circle, and calculate the defect area ratio or deviation distance parameter; This invention focuses on the assessment methods for three types of defects: obstructions, sediments, and misalignments. Obstructions are quantified by the percentage loss of the pipe's cross-sectional area and are classified as area-related defects. Sediments, spread flat at the bottom of the pipe, have a greater impact on water flow obstruction; the assessment criterion is the ratio of sediment thickness to pipe diameter. Misalignments mainly manifest as flow blockage and leakage risks caused by misaligned joints; the assessment criterion is the joint deviation distance, classifying them as length-related defects.

[0029] For area-related defect assessment, this invention uses a pixel-area ratio-based calculation method. It calculates the defect's impact on the water flow cross-section by using the ratio of the projected area of ​​the defect region in the image to the area of ​​the pipe's reference cross-section, i.e., the reference circle. The formula is:

[0030] in, This is the area ratio. The area of ​​the defective pixel.

[0031] For length-related defect assessment, this invention applies the Canny edge detection algorithm to the mask image to extract edge features, calculates the center point and radius of the maximum inscribed circle of the defect, and draws the maximum inscribed circle; then, it calculates the ratio of the diameter of the maximum inscribed circle to the diameter of the reference circle, using the following formula:

[0032] Among them, in the misalignment defect This is the ratio of pipe diameter to pipe wall thickness, among other defects. , denoted by , where is the diameter of the largest inscribed circle of the defect, and r is the radius of the reference cross-section circle.

[0033] S6, quantitatively evaluate the defect level based on a preset threshold.

[0034] (1) In the regulations, the classification of obstacles is based on the percentage loss of the pipe cross-section, and a classification model is established: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four.

[0035] (2) In the regulations, sediment classification is based on the ratio of sediment thickness to pipe diameter, and a classification model is established: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four.

[0036] (3) In the regulations, the classification of misalignment is based on the ratio of misalignment deviation to pipe wall thickness. A classification model is established based on the ratio of pipe diameter to wall thickness: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four.

[0037] Example: S1, synchronously acquire the original image of the pipeline and the defect mask image generated by the semantic segmentation network, and perform preprocessing. The image contains defects and misalignments. S2, Select a reference image based on HSV color space analysis. Since no circle is detected in the mask image, the original image is used to locate the reference circle. S3, generate a candidate circle set using the Hough circle transform algorithm (see...) Figure 3 The candidate circle selection weights are obtained by combining position weights and confidence weights, and then optimized and filtered to generate the baseline circle (see [link]). Figure 4 ); S4, Adjust the radius of the reference circle by expanding it according to the geometric characteristics of the defect area. Calculated ;(See Figure 5 ) S5, calculate the defect geometric parameters based on the optimized reference circle, and calculate the defect area ratio or deviation distance parameter. Substitute:

[0038] have to ; S6, Quantitatively evaluate the defect level based on a preset threshold. It was determined to be a level four misalignment.

Claims

1. A method for evaluating defects in drainage pipes based on dual-source input and Hough transform, characterized in that, Specifically, the following steps are included: S1, synchronously acquire the original pipeline image and the defect mask image generated by the semantic segmentation network, and perform preprocessing; S2. Based on HSV color space analysis, a reference image is selected. The defect mask image is used first for reference circle positioning. If no circle is detected in the defect mask image, the original image is used for reference circle positioning. S3: A candidate circle set is generated using the Hough circle transform algorithm. The candidate circle selection weights are obtained by combining position weights and confidence weights, and then optimized and filtered to generate the baseline circle. S4. The radius of the reference circle is expanded and adjusted according to the geometric characteristics of the defect area to obtain the adjusted reference circle; S5, calculate the defect geometric parameters based on the adjusted reference circle, and calculate the defect area ratio or deviation distance parameter; S6. Based on the results of step 5, the defect level is quantitatively evaluated using a preset threshold.

2. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 1, characterized in that, In S1, the preprocessing specifically involves: using... The median filter removes impulse noise, and then... Gaussian smoothing is applied, and morphological opening is performed using an elliptic kernel.

3. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 1, characterized in that, S3 specifically includes the following sub-steps: S31, Parameters are configured and a candidate circle set is generated based on the Hough gradient transform. The Euclidean distance between the center of the candidate circle and the center of the original image is calculated. The Gaussian decay function is applied to generate position weights, the expression of which is: in, The position weights of the candidate circles, The distance from the center of the candidate circle to the center of the original image. The standard deviation is the Gaussian distribution. S32, assign exponential decay weights according to the original voting sorting based on the Hough transform; Specifically, the expression for the exponentially decaying weight of the candidate circle is: in, The exponential decay weights of the candidate circles are used as the confidence weights for the candidate circles. The original sorting index for candidate circles. The total number of candidate circles; the circle with the most votes corresponds to Its confidence weight is the largest; S33, Multiply the position weight of the candidate circle by the confidence weight to obtain the selection weight of the candidate circle. The candidate circle with the highest score is selected from the candidate circle selection weight S as the baseline circle; Specifically, the selection weight of the reference circle The calculation method is as follows: in, The adjustment parameter is used to control the influence of position weights on the selection weights of candidate circles.

4. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 3, characterized in that, S4 specifically involves calculating the maximum Euclidean distance between the pixel coordinates of each defect region and the center of the fitted circle, and then introducing a margin coefficient λ to expand the radius of the reference circle. This process is expressed as: in, The radius of the fitted circle, i.e., the radius of the reference circle; These are the pixel coordinates of the defect area. Using the coordinates of the center of the reference circle, This is the preset margin coefficient.

5. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 4, characterized in that, The value range is [1.05, 1.15].

6. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 5, characterized in that, The specific operation of S5 is as follows: For area-related defect assessment, the calculation formula is as follows: in, This is the area ratio. The area of ​​the defective pixel; For length-related defect assessment, the calculation formula is as follows: Among them, in the misalignment defect This is the ratio of pipe diameter to pipe wall thickness, among other defects. , denoted by , where is the diameter of the largest inscribed circle of the defect, and r is the radius of the reference cross-section circle.

7. The drainage pipe defect assessment method based on dual-source input and Hough transform as described in claim 6, characterized in that, The specific operation of S6 is as follows: (1) In the regulations, the classification of obstacles is based on the percentage loss of the pipe cross-section, and a classification model is established: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four at the time; (2) In the regulations, sediment classification is based on the ratio of sediment thickness to pipe diameter, and a classification model is established: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four at the time; (3) In the regulations, the classification of misalignment is based on the ratio of misalignment deviation to pipe wall thickness. A classification model is established based on the ratio of pipe diameter to wall thickness: when It was determined to be level one at that time. when It was determined to be level two at that time. when It was determined to be level three at the time. when It was determined to be level four.

Images