A municipal drainage pipeline image recognition and analysis system

Abstract

The application discloses a kind of municipal drainage pipeline image recognition and analysis system, it is related to image recognition technical field, the system includes: data acquisition module gathers pipeline inner wall image, water quality, flow and equipment speed data;Scene variable coupling module is constructed model by dynamic weight calculation and multi-peak membership analysis, and outputs comprehensive scene identification;Image enhancement collaborative module is sequentially self-adapting enhanced image according to filtering, standardization, correction, sharpening;Defect identification analysis module extracts feature, classifies defect and generates structured report;Adaptive iteration module compares identification rate threshold, and optimizes and updates parameter library;The application constructs multivariate coupling model, combines sewage noise adaptive filter kernel formula and dynamic distortion correction strategy, solves the problem that traditional system scene is not adapted, image processing is not good, and identification is not accurate, adapts diversified pipeline detection demand.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, specifically to an image recognition and analysis system for municipal drainage pipelines. Background Technology

[0002] Municipal drainage pipelines are a crucial component of urban infrastructure, and their operational status directly impacts urban drainage, flood control, and water quality. As pipelines age, defects such as cracks, siltation, and joint damage become increasingly common, necessitating regular inspections to promptly identify and address potential hazards. Current mainstream inspection methods largely employ image recognition technology, utilizing camera-equipped inspection devices to capture images of the pipeline's inner wall. Defects are then determined through image enhancement, feature extraction, and classification. The internal environment of pipelines is complex, presenting challenges such as low light levels, sewage noise, and equipment movement vibrations. Furthermore, variables such as sewage type, operating flow rate, and equipment speed interact, placing high demands on the scene adaptability and accuracy of image recognition systems. Traditional pipeline image recognition systems have significant technical shortcomings. Variable weight allocation often uses fixed values, failing to adapt to complex scenarios where various variables interact and influence each other, leading to substantial scene judgment errors. Image enhancement filtering and correction processes typically employ general algorithms without dynamically adjusting parameters based on sewage-specific noise and equipment speed, resulting in poor noise suppression and distortion correction. Furthermore, traditional systems lack targeted parameter iteration optimization mechanisms, making it difficult to adapt to the detection needs of different pipeline scenarios. This leads to issues such as missed or false defects, insufficient recognition accuracy and stability, and an inability to meet the practical needs of refined municipal drainage pipeline inspection. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a municipal drainage pipeline image recognition and analysis system. This system is based on data acquisition and uses a dynamic weight formula for scene variable interaction terms and multi-peak membership analysis to construct a multivariate coupled model to accurately match pipeline detection scenarios. At the same time, it combines a sewage noise adaptive filtering kernel formula and a distortion correction strategy with dynamic speed adjustment to specifically optimize image enhancement effects. With the addition of an improved feature extraction and iterative optimization mechanism, it solves the problems of inappropriate scene adaptation, poor image processing, and inaccurate recognition in traditional systems.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a municipal drainage pipeline image recognition and analysis system, the system comprising: The data acquisition module is used to acquire continuous images of the inner wall of the pipeline, the concentration of suspended solids and COD value of sewage, the real-time flow rate of the pipeline, and the speed of the detection equipment. The scene variable coupling module receives data collected by the data acquisition module, constructs a multivariate coupling model through dynamic weight calculation and multi-peak membership analysis, and outputs a comprehensive scene identifier. The image enhancement coordination module is used to receive comprehensive scene identifiers and perform image enhancement in a coordinated manner in the order of filtering, standardization, correction and sharpening, adapting to the wastewater type and equipment speed status; The defect identification and analysis module is used to receive the enhanced image, extract image features and classify defect types, determine defect information, and generate a structured report. The adaptive iteration module is used to receive defect identification results, compare them with preset identification rate thresholds, trigger parameter optimization, and update the technical parameter library.

[0005] Furthermore, the scene variable coupling module calculates the weight of each variable using a dynamic weight formula for scene variable interaction items, as follows: ; in, For the first i Dynamic weights of each variable, i =1 corresponds to the wastewater type, i =2 corresponds to the operating condition. i =3 corresponds to the device speed; For the basic weights of the variables, =0.4、 =0.3、 =0.3, this basic weight is based on the proportion of the influence of sewage type, operating conditions and equipment speed on the image enhancement effect in the municipal drainage pipeline scene, and is calibrated by statistical analysis of on-site measured data; β The interaction influence coefficient has a value range of 0.2-0.5. This coefficient is based on the actual linkage influence between various variables in the pipeline and is obtained through fitting and calibration of multiple sets of scenario experiments. For variables i With variables j The general interaction coefficient, j To and i Different variables have different specific values. 、 、 These correspond to the pairwise interaction coefficients between wastewater type and operating condition, wastewater type and equipment speed, and operating condition and equipment speed, respectively, with values ​​as follows: =0.6、 =0.3、 =0.4, the specific value of this group is obtained based on the actual correlation analysis between the variables of the pipeline and the on-site test calibration of the municipal drainage pipeline inspection. For the first j The normalized values ​​of each variable, ranging from 0 to 1, are obtained by linearly normalizing the actual collected values ​​of each variable. For variables kWith variables l Interaction coefficient For the first l The normalized values ​​of the variables, and the two... , These are traversal parameters of the same type, differing only in the traversal symbols of the variables in the numerator and denominator of the formula. Their meanings and value rules are the same as those of other traversal parameters. , Consistent; k is the variable traversed in the denominator. l To and k Different variables, k , l The values ​​range from 1 to 3, corresponding to wastewater type, operating condition, and equipment speed, respectively.

[0006] Furthermore, the filtering submodule of the image enhancement collaborative module constructs the filtering kernel using the sewage noise adaptive filtering kernel formula, as follows; ; ; in, pixel coordinates Adaptive filter kernel at the location; The normalization coefficient of the filter kernel is obtained by summing all pixel values ​​of the filter kernel and is used to ensure that the sum of the filter kernel weights is 1. The value is 5 for rainwater scene, 3 for domestic sewage scene, and 4 for industrial wastewater scene. This coefficient is based on the spatial noise distribution characteristics of different sewage types and was obtained through image denoising experiments and calibration. The gray-level similarity coefficient is obtained by adaptive matching of the gray-level variance features of the image to be processed. pixel coordinates The grayscale value at that location; The grayscale value of the center pixel of the filter kernel; The noise intensity adjustment coefficient is set to 0.8. This coefficient is based on the overall intensity characteristics of sewage noise from municipal drainage pipelines and was obtained through fitting and calibration of noise reduction tests on multiple types of sewage samples. This is a comprehensive wastewater noise index, used to quantify the noise intensity corresponding to different types of wastewater. This represents the actual suspended solids concentration in the wastewater. The concentration limit for suspended solids in sewage from municipal drainage pipelines is set at 1000 mg / L, determined in accordance with the water quality testing standards for the municipal drainage industry. This refers to the actual chemical oxygen demand (COD) of the wastewater. The limit for chemical oxygen demand (COD) of wastewater in municipal drainage pipelines is set at 2000 mg / L, determined according to the water quality testing standards for the municipal drainage industry. The 0.6 and 0.4 in the calculation formula are weighting coefficients, which are obtained by statistical calibration based on the contribution analysis of suspended solids concentration and chemical oxygen demand to sewage noise in municipal drainage pipelines. Among them, suspended solids concentration has a higher contribution to sewage noise, so it is assigned a weight of 0.6, and chemical oxygen demand is assigned a weight of 0.4.

[0007] Furthermore, the correction submodule of the image enhancement collaboration module dynamically adjusts the distortion correction coefficients according to the travel speed of the detection device. For high-speed conditions, Wiener filtering is used for motion blur compensation. The blur kernel size increases by 1×1 for every 0.1m / s increase in the device's travel speed, and the blur kernel size ranges from 3×3 to 9×9. The size change law is obtained by actual measurement and analysis of the degree of motion blur of the image under different travel speeds of the detection device. For low-speed conditions, third-order polynomial distortion correction is used. The distortion parameters are calibrated by acquiring standard checkerboard images and solving for the polynomial coefficients. The polynomial coefficients are then used to map and correct the image pixel coordinates.

[0008] Furthermore, the data acquisition module includes an image data acquisition unit, a wastewater type acquisition unit, an operating condition acquisition unit, and an equipment status acquisition unit; The image data acquisition unit uses a high-definition low-light industrial camera with a resolution of no less than 1920×1080, a frame rate of no less than 25fps, and a minimum adaptable light intensity of no more than 5 lux. The parameter indicators are calibrated based on the requirements of low light and pipeline defect identification in municipal drainage pipelines for image clarity. The wastewater type acquisition unit uses a water quality sensor to detect the concentration of suspended solids and COD value in wastewater; The operating condition acquisition unit uses a flow sensor to detect the real-time flow rate in the pipeline; The equipment status acquisition unit uses a speed sensor with an accuracy of ±0.05m / s to acquire and detect the moving speed of the equipment. The accuracy index is calibrated based on the impact of equipment speed on image correction effect.

[0009] Furthermore, the scenario variable coupling module uses a multi-peak membership function to calculate the fuzzy membership degree of each variable, and generates a comprehensive scenario identifier based on the combination of wastewater type, operating condition, and equipment speed levels. This identifier is a three-digit code, with the first digit corresponding to the wastewater type level, the second digit corresponding to the operating condition level, and the third digit corresponding to the equipment speed level. Each level of each variable is assigned a unique numerical code.

[0010] Furthermore, the standardization submodule of the image enhancement collaboration module determines the operating conditions by comparing the real-time flow rate of the pipeline with the designed flow rate of the pipeline. In rainy weather, the frame rate is set to 30fps and the resolution to 1280×720. In dry weather, the resolution is maintained at 1920×1080 and the image brightness is normalized to the range of 0-255. The frame rate and resolution parameters are calibrated based on the image transmission and processing efficiency and the requirements for the clarity of defect identification under different operating conditions. The sharpening submodule matches directional sharpening operators based on the type of target defect. Crack defects use the Sobel operator, siltation defects use the Laplacian operator, and interface damage defects use a combination of Harris corner detection and local sharpening. The operator matching rules are calibrated based on the image features of different pipeline defects and the measured analysis of sharpening effects.

[0011] Furthermore, the defect identification and analysis module includes a feature extraction unit, a defect classification unit, and a report generation unit; The feature extraction unit uses an improved ResNet-50 convolutional neural network to preprocess the enhanced image by mean filtering, and then performs feature mapping through a 3×3 convolutional kernel to extract the shape features, size features and gray-level distribution features of the image, and outputs a 256-dimensional feature vector. The feature vector dimension is calibrated based on the dimensional analysis of the pipeline defect features and the recognition accuracy requirements. The defect classification unit uses the Softmax classifier to classify the extracted feature vectors into three types of defects: cracks, siltation, and interface damage. The structured report generated by the report generation unit includes pipeline mileage location, defect type, defect size, defect percentage, and repair priority. This structured report supports export in PDF and Excel formats.

[0012] Furthermore, the adaptive iteration module presets a defect recognition rate threshold of 95%, which is calibrated based on the industry accuracy requirements for defect recognition in municipal drainage pipelines. When the defect recognition rate is lower than this threshold for 10 consecutive times in a certain comprehensive scenario, parameter iterative optimization is triggered. This number of iterations threshold is calibrated based on the convergence efficiency of the iterative optimization and the scenario stability requirements. Iterative optimization is performed using the gradient descent method, with a step size of 0.05. This step size is calibrated based on the optimization convergence speed and accuracy requirements of the image enhancement collaboration module parameters. The technical parameter weights of each sub-module in the image enhancement collaboration module are adjusted using this method, and the optimized technical parameters are stored in the scene parameter library, overwriting the original corresponding parameters in the library.

[0013] Compared with existing technologies, this municipal drainage pipeline image recognition and analysis system has the following advantages: I. This invention constructs a multivariate coupling model by using a dynamic weight formula for scene variable interaction terms and multimodal membership analysis. This solves the problem that traditional fixed weights cannot adapt to variable linkage, and realizes dynamic correlation weight calculation for sewage type, operating condition, and equipment speed. The interaction coefficient introduced by this formula, combined with calibration parameters based on on-site pipeline measurement data, accurately captures the influence of variable linkage on image enhancement. This enables the comprehensive scene identification to accurately match the actual pipeline detection scene, effectively avoiding scene deviations caused by single variable judgment. It provides parameter support that fits the actual situation for subsequent stages, achieving accurate adaptation in complex pipeline scenarios and making the entire recognition system more consistent with the variable characteristics of the actual detection environment.

[0014] II. This invention optimizes the image enhancement process by combining a wastewater noise adaptive filtering kernel formula with a distortion correction strategy based on dynamic speed adjustment. The filtering kernel formula incorporates suspended solids concentration and COD value to construct a comprehensive noise index, which can accurately suppress specific noises of different wastewater types. The correction strategy adjusts parameters according to equipment speed to solve image distortion problems under moving conditions, improves image quality in low-light and high-noise environments, and clearly presents defect features. Combined with feature extraction and iterative optimization mechanisms, it further ensures the stability of defect identification, providing high-quality image data for the identification process after core scene adaptation, and meeting the detection needs of diverse pipeline scenarios.

[0015] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0017] Figure 1 This is a flowchart illustrating the overall system structure of the present invention. Figure 2 This is a flowchart of the image enhancement collaborative processing of the present invention; Figure 3 This is a flowchart of the defect identification and adaptive iteration process of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0019] This embodiment discloses a specific implementation of an image recognition and analysis system for municipal drainage pipelines, which consists of a data acquisition module, a scene variable coupling module, an image enhancement and collaboration module, a defect recognition and analysis module, and an adaptive iteration module. Figure 1 As shown, during system operation, the data acquisition module first collects various types of data required for pipeline inspection. Then, the scene variable coupling module constructs a multivariate coupling model to accurately match the inspection scene. Subsequently, the image enhancement and collaboration module optimizes image quality in a targeted manner. Next, the defect identification and analysis module completes defect determination. Finally, the adaptive iteration module continuously optimizes system parameters. The entire process revolves around the actual needs of pipeline inspection, solving the problems of inappropriate scene adaptation, poor image processing effect, and inaccurate identification in traditional systems, making pipeline defect detection more closely aligned with actual on-site conditions.

[0020] The specific implementation process is as follows; The operation process of the data acquisition module: The data acquisition module is the core of the system for obtaining basic information. It includes an image data acquisition unit, a wastewater type acquisition unit, an operating condition acquisition unit, and an equipment status acquisition unit. All units work synchronously to ensure the comprehensiveness and real-time nature of data acquisition.

[0021] The image data acquisition unit uses a high-definition, low-light industrial camera. Given the dim lighting inside the pipes, this type of camera is well-suited to such environments, stably capturing continuous images of the pipe's inner wall, revealing texture details and potential defects. The image clarity must meet the requirements for subsequent processing and identification. The wastewater type acquisition unit employs a water quality sensor specifically designed to detect the concentration of suspended solids and chemical oxygen demand (COD) in wastewater. These two indicators directly reflect the characteristics of the wastewater, and their results are directly used for subsequent noise analysis, providing crucial information for image filtering.

[0022] The operational status acquisition unit, equipped with a flow sensor, monitors the sewage flow rate within the pipeline in real time. The flow data allows for the determination of the pipeline's current operating status, such as whether it's a high-flow condition during rain or a normal dry-flow condition, providing a reference for subsequent image parameter adjustments. The equipment status acquisition unit uses a high-precision speed sensor to record the movement speed of the detection equipment within the pipeline in real time. The equipment's movement speed affects image acquisition quality, potentially causing image distortion or blurring; therefore, this speed data is crucial for subsequent image correction.

[0023] Image data, water quality data, flow data, and velocity data collected by each acquisition unit will undergo a unified format conversion before being synchronously transmitted to the scene variable coupling module, providing complete data support for the module's analysis and calculation.

[0024] The operation process of the scene variable coupling module: After receiving all the data transmitted by the data acquisition module, the primary task of the scene variable coupling module is to construct a multivariate coupling model through dynamic weight calculation and multi-peak membership analysis, and finally generate a comprehensive scene identifier that can accurately reflect the current detection scene.

[0025] The dynamic weight calculation uses a dynamic weight formula for the interaction terms of scenario variables, specifically designed to solve for the dynamic weights of three core variables: wastewater type, operating condition, and equipment speed. The formula is as follows: ; in, Representing the i Dynamic weights of each variable, i Corresponding to wastewater type, operating condition, and equipment speed, the dynamic weights flexibly adjust the degree of influence of each variable on the image enhancement effect according to the actual situation on site, avoiding the problem that traditional fixed weights cannot adapt to the linkage changes of variables. These are the basic weights of variables. They are determined based on the actual proportion of each variable's impact on the image enhancement effect in the municipal drainage pipeline scenario, through statistical analysis of a large amount of field measurement data. They can reflect the basic importance of each variable in scenario adaptation. The interaction coefficient measures the strength of the interaction between different variables. Its value range is determined by analyzing the actual linkage patterns between the variables in the pipeline and fitting multiple sets of experiments in different scenarios to ensure that the linkage relationship between variables can be accurately captured. It is a variable i With variables j The general interaction coefficient, and and i For different variables, their specific values ​​correspond to the pairwise combinations of the three variables, respectively reflecting the correlation strength between sewage type and operating conditions, sewage type and equipment speed, and operating conditions and equipment speed. This set of values ​​is determined by analyzing the actual correlation between the variables in the pipeline and combining a large number of field tests of municipal drainage pipelines, which can accurately reflect the linkage characteristics of different variable combinations.

[0026] It is the first j The normalized values ​​of each variable are obtained by linearly normalizing the actual collected values ​​of each variable. This is done to eliminate the differences in units and numerical ranges of different variables, so that each variable can participate in the weight calculation on the same dimension. and , and These are all traversal parameters of the same type, differing only in the variable traversal symbols used in the numerator and denominator of the formula. Their meanings and value rules are completely consistent. k and l Similarly, these three core variables ensure the comprehensiveness and accuracy of the weight calculation.

[0027] After completing the dynamic weight calculation, the scene variable coupling module uses a multi-peak membership function to calculate the fuzzy membership degree of each variable, quantifying the degree to which each variable's current state belongs to different levels. Then, based on the level combination of wastewater type, operating condition, and equipment speed, it generates a comprehensive scene identifier composed of three-digit codes. Each digit corresponds to a variable level, and each level has a unique numerical code. This concise coding format allows the subsequent image enhancement collaboration module to quickly identify the current detection scene, providing a clear direction for parameter adjustments in each sub-module within the module.

[0028] The operation process of the image enhancement collaboration module: After receiving the comprehensive scene identifier output by the scene variable coupling module, the image enhancement collaboration module, such as Figure 2 As shown, image enhancement operations are carried out in the order of filtering, standardization, correction, and sharpening. The parameters of each step are adjusted according to the comprehensive scene identifier to ensure that the image enhancement effect is perfectly adapted to the current detection scene.

[0029] Running the filtering submodule: The core of the filtering submodule is to construct a filtering kernel using an adaptive filtering kernel formula for wastewater noise, specifically designed to filter noise specific to different types of wastewater. The formula is as follows: ; ; in, Pixel coordinates The adaptive filtering kernel at the pixel is used to filter noise around the pixel while preserving the feature information of the pipe defect, thus avoiding blurring of the defect during the filtering process. It is the normalization coefficient of the filter kernel, which is obtained by summing all pixel values ​​of the filter kernel. This is done to ensure that the weights of the filter kernel sum to 1, so as to prevent the gray values ​​of the filtered image from being distorted.

[0030] It is a spatial smoothing coefficient, the value of which is determined by the sewage type level in the comprehensive scene identifier. Different sewage types have different spatial noise distributions. This coefficient is determined by analyzing the noise distribution patterns of different sewage types and through a large number of image denoising experiments. It can accurately suppress spatial noise in different sewage scenes. It is a gray-level similarity coefficient, not a fixed value, but an adaptive matching based on the gray-level variance characteristics of the image to be processed, ensuring that the filtering process can adapt to the gray-level distribution characteristics of the image itself.

[0031] Pixel coordinates grayscale value at that location It is the gray value of the center pixel of the filter kernel. By calculating the gray value difference between the two, the pixel's own characteristics and noise interference can be distinguished, providing a basis for the filtering operation. It is a noise intensity adjustment coefficient, which is determined by fitting noise reduction tests of multiple types of sewage samples based on the overall intensity characteristics of sewage noise in municipal drainage pipelines. It is used to adjust the intensity of noise suppression and avoid the loss of defect features due to excessive filtering.

[0032] It is a comprehensive wastewater noise index, specifically used to quantify the noise intensity of the current wastewater type. Its calculation process incorporates the suspended solids concentration and chemical oxygen demand detected by the wastewater type acquisition unit. This refers to the actual suspended solids concentration in the wastewater. It is the limit value for the concentration of suspended solids in sewage in municipal drainage pipes, which is determined according to the water quality testing standards of the municipal drainage industry; This refers to the actual chemical oxygen demand (COD) of the wastewater. It is the limit value for chemical oxygen demand in sewage from municipal drainage pipes, which is also determined according to industry standards.

[0033] The 0.6 and 0.4 in the calculation formula are weighting coefficients. Based on the analysis of the contribution of suspended solids concentration and chemical oxygen demand to sewage noise in municipal drainage pipelines, and determined through statistical analysis of measured data, suspended solids concentration has a greater impact on sewage noise, so it is assigned a higher weight. By weighted summing of the two, the noise level of sewage can be comprehensively and accurately reflected, providing a core basis for the adaptive adjustment of the filter kernel.

[0034] Running the standardized submodule: The standardization submodule first compares the real-time flow rate and the design flow rate of the pipeline, and then, combined with the operating condition level in the comprehensive scenario identifier, clarifies the current operating condition of the pipeline. Under different operating conditions, the image acquisition environment and processing requirements vary, thus requiring targeted adjustments to image parameters: under high flow conditions, to ensure image transmission and processing efficiency while meeting the basic requirements of defect identification, the image frame rate and resolution are appropriately adjusted; under low flow conditions, a higher image resolution is maintained, while image brightness is normalized to a specific range to make the image brightness more uniform, creating favorable conditions for subsequent sharpening and feature extraction. These frame rate and resolution parameter settings are determined based on the image transmission efficiency, processing efficiency, and defect identification clarity requirements under different operating conditions, achieving a balance between efficiency and effectiveness.

[0035] Operation of the calibration submodule: The correction submodule dynamically adjusts the distortion correction coefficient based on the device speed level in the comprehensive scene identifier and the real-time travel speed collected by the device status acquisition unit, and adopts different correction strategies for different speed states.

[0036] When the detection equipment is moving at high speed, the captured images are prone to motion blur. At this time, Wiener filtering is used to compensate for motion blur. The size of the blur kernel will increase accordingly as the speed of the equipment increases. This size change law is determined by analyzing the degree of motion blur of the images at different speeds of the detection equipment through a large number of field tests. It can accurately match the size of the blur kernel with the degree of motion blur, effectively compensating for the image distortion caused by high-speed movement.

[0037] When the detection equipment is traveling at low speed, the image mainly suffers from optical distortion. At this time, a third-order polynomial distortion correction method is adopted. First, a standard checkerboard image is acquired to calibrate the distortion parameters. Then, the polynomial coefficients are calculated and solved. Finally, the image pixel coordinates are mapped and corrected using the polynomial coefficients to eliminate the distortion error caused by the optical system and ensure the accuracy of the image geometry.

[0038] Running the sharpening submodule: The sharpening submodule matches the corresponding targeted sharpening operator based on the type of target defect. Different pipe defects exhibit significant differences in image features: crack defects, which display linear characteristics, are sharpened using operators that enhance edges to highlight the linear contour of the crack; siltation defects, which display regional characteristics, are sharpened using operators that enhance regional grayscale differences to make the siltation area more distinct from the surrounding pipe wall; and interface damage defects, which have corner features, are sharpened using a combination of corner detection and local sharpening. First, the corner positions of the damaged interface are accurately located, and then the local area is sharpened to enhance the damage features. This operator matching rule was determined by analyzing the image features of different pipe defects and combining extensive experimental sharpening effects. This allows the sharpening operation to specifically enhance various defect features, making subsequent identification more accurate.

[0039] The operation process of the defect identification and analysis module: The defect identification and analysis module receives the enhanced image output by the image enhancement and collaboration module, such as... Figure 3 As shown, the accurate identification and structured presentation of pipeline defects are achieved through the collaborative work of the feature extraction unit, the defect classification unit, and the report generation unit.

[0040] The feature extraction unit employs an improved ResNet-50 convolutional neural network. First, the enhanced image undergoes mean filtering preprocessing to further eliminate residual minor noise and prevent noise interference with feature extraction. The preprocessed image is then input into the convolutional layer, where feature mapping is performed using convolutional kernels of a specific size. This kernel size ensures both the accuracy of feature extraction and reduces computational cost, thus improving processing efficiency.

[0041] After multiple rounds of convolution and pooling operations, the network extracts the shape features, size features, and grayscale distribution features of the image layer by layer. These features can comprehensively reflect the key information of pipeline defects and finally output a feature vector with fixed dimensions. The dimension of the feature vector is determined based on the dimensional analysis of pipeline defect features and the accuracy requirements of recognition. This not only preserves the core feature information but also ensures the efficiency of subsequent classification operations.

[0042] The defect classification unit employs a Softmax classifier, receiving feature vectors output from the feature extraction unit. By calculating the similarity between these feature vectors and standard features for various defect types, the feature vectors are categorized into their corresponding defect categories, ultimately classifying three common pipe defects: cracks, siltation, and interface damage. This classifier is trained on a large number of labeled pipe defect samples, ensuring classification accuracy and stability, effectively distinguishing different types of defects, and avoiding misclassification.

[0043] The report generation unit integrates key data such as defect type, defect size, and defect percentage based on the defect classification unit's judgment and pipeline mileage location information recorded by the data acquisition module. It then determines the repair priority based on the severity of the defects and ultimately generates a structured report. This report supports export in multiple common formats, facilitating viewing and archiving by inspection and management personnel, and providing convenience for subsequent repair plan development.

[0044] The operation process of the adaptive iterative module: The adaptive iteration module continuously receives defect identification results from the defect identification and analysis module, compares them with a preset defect identification rate threshold, and determines whether the current system parameters are suitable for the current detection scenario. This identification rate threshold is determined based on the industry accuracy requirements for defect identification in municipal drainage pipelines and is an important standard for measuring the system's identification performance.

[0045] When the defect recognition rate falls below a certain threshold multiple times in a given scenario, the system determines that the parameters of the current image enhancement co-module cannot meet the recognition requirements of that scenario, thus triggering iterative parameter optimization. The iterative optimization employs gradient descent, with the step size determined based on the convergence speed and accuracy requirements of the image enhancement co-module parameters. This step size is used to adjust the weights of the technical parameters of each sub-module within the image enhancement co-module, including the coefficients of the filtering sub-module, the correction parameters of the correction sub-module, and the image parameters of the standardization sub-module.

[0046] The optimized technical parameters will be stored in the scene parameter library, overwriting the original corresponding parameters in the library. In this way, when the same or similar scenes are encountered again, the system can directly call the optimized parameters, so that the recognition effect can be continuously improved and the system can better adapt to different pipeline detection scenarios.

[0047] This embodiment details the complete operation of a municipal drainage pipeline image recognition and analysis system. Based on data acquisition, the system accurately matches the detection scene through a scene variable coupling module, then optimizes image quality through an image enhancement and collaboration module, followed by defect identification and analysis by a defect detection module, and finally, continuous parameter optimization through an adaptive iteration module. The collaborative work of these modules solves the problems of inappropriate scene adaptation, poor image processing, and inaccurate recognition in traditional pipeline image recognition systems. The entire operation process aligns with the actual needs of municipal drainage pipeline inspection, enabling more accurate and efficient pipeline defect detection and adapting to diverse pipeline inspection scenarios.

[0048] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A municipal drainage pipeline image recognition and analysis system, characterized in that, The system includes: The data acquisition module is used to acquire continuous images of the inner wall of the pipeline, the concentration of suspended solids and COD value of sewage, the real-time flow rate of the pipeline, and the speed of the detection equipment. The scene variable coupling module receives data collected by the data acquisition module, constructs a multivariate coupling model through dynamic weight calculation and multi-peak membership analysis, and outputs a comprehensive scene identifier. The image enhancement coordination module is used to receive comprehensive scene identifiers and perform image enhancement in a coordinated manner in the order of filtering, standardization, correction and sharpening, adapting to the wastewater type and equipment speed status; The defect identification and analysis module is used to receive the enhanced image, extract image features and classify defect types, determine defect information, and generate a structured report. The adaptive iteration module is used to receive defect identification results, compare them with preset identification rate thresholds, trigger parameter optimization, and update the technical parameter library.

2. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The scene variable coupling module uses a dynamic weight formula for scene variable interaction items to calculate the weight of each variable, as follows: ； in, For the first i Dynamic weights of each variable, i =1 corresponds to the wastewater type, i =2 corresponds to the operating condition. i =3 corresponds to the device speed; For the basic weights of the variables; β This is the interaction coefficient; For variables i With variables j The general interaction coefficient; For the first j Normalized values ​​of each variable; For variables k With variables l Interaction coefficient, For the first l The normalized values ​​of the variables; k is the variable traversed in the denominator. l To and k Different variables.

3. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The filtering submodule of the image enhancement collaborative module uses the sewage noise adaptive filtering kernel formula to construct the filtering kernel, as follows; ； ； in, pixel coordinates Adaptive filter kernel at the location; These are the normalization coefficients of the filter kernel; It is the spatial smoothing coefficient; The gray-scale similarity coefficient; pixel coordinates The grayscale value at that location; The grayscale value of the center pixel of the filter kernel; This is the noise intensity adjustment coefficient; The comprehensive index of wastewater noise; This represents the actual suspended solids concentration in the wastewater. The limit for suspended solids concentration in municipal drainage pipes; This refers to the actual chemical oxygen demand (COD) of the wastewater. The limit value for chemical oxygen demand (COD) of sewage in municipal drainage pipelines; The 0.6 and 0.4 in the calculation formula are weighting coefficients.

4. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The correction submodule of the image enhancement collaboration module dynamically adjusts the distortion correction coefficients according to the travel speed of the detection equipment. For high-speed conditions, Wiener filtering is used for motion blur compensation. The blur kernel size increases by 1×1 for every 0.1m / s increase in the travel speed of the equipment. The blur kernel size ranges from 3×3 to 9×9. The size change law is obtained by actual measurement and analysis of the degree of motion blur of the image under different travel speeds of the detection equipment. For low-speed conditions, third-order polynomial distortion correction is used. The distortion parameters are calibrated by collecting standard checkerboard images and solving for the polynomial coefficients. The polynomial coefficients are then used to map and correct the image pixel coordinates.

5. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The data acquisition module includes an image data acquisition unit, a wastewater type acquisition unit, an operating condition acquisition unit, and an equipment status acquisition unit. The image data acquisition unit uses a high-definition low-light industrial camera with a resolution of no less than 1920×1080, a frame rate of no less than 25fps, and a minimum adaptable light intensity of no more than 5 lux. The wastewater type acquisition unit uses a water quality sensor to detect the concentration of suspended solids and COD value in wastewater; The operating condition acquisition unit uses a flow sensor to detect the real-time flow rate in the pipeline; The equipment status acquisition unit uses a speed sensor with an accuracy of ±0.05m / s to collect and detect the equipment's travel speed.

6. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The scenario variable coupling module uses a multi-peak membership function to calculate the fuzzy membership degree of each variable, and generates a comprehensive scenario identifier based on the combination of wastewater type, operating condition and equipment speed levels. This identifier is a three-digit code. The first digit corresponds to the wastewater type level, the second digit corresponds to the operating condition level, and the third digit corresponds to the equipment speed level. Each level of each variable is assigned a unique numerical code.

7. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The standardized submodule of the image enhancement collaboration module determines the operating conditions by comparing the real-time flow rate of the pipeline with the designed flow rate of the pipeline. In rainy weather, the frame rate is set to 30fps and the resolution is 1280×720. In dry weather, the resolution is maintained at 1920×1080 and the image brightness is normalized to the range of 0-255. The sharpening submodule matches directional sharpening operators based on the target defect type. Crack defects use the Sobel operator, siltation defects use the Laplacian operator, and interface damage defects use a combination of Harris corner detection and local sharpening.

8. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The defect identification and analysis module includes a feature extraction unit, a defect classification unit, and a report generation unit; The feature extraction unit uses an improved ResNet-50 convolutional neural network to preprocess the enhanced image by mean filtering, and then performs feature mapping through a 3×3 convolutional kernel to extract the shape features, size features and gray-level distribution features of the image, and outputs a 256-dimensional feature vector. The defect classification unit uses the Softmax classifier to classify the extracted feature vectors into three types of defects: cracks, siltation, and interface damage. The structured report generated by the report generation unit includes pipeline mileage location, defect type, defect size, defect percentage, and repair priority. This structured report supports export in PDF and Excel formats.

9. The municipal drainage pipeline image recognition and analysis system according to claim 1, characterized in that, The adaptive iteration module has a preset defect recognition rate threshold of 95%, which is calibrated based on the industry accuracy requirements for defect recognition in municipal drainage pipelines. When the defect recognition rate is lower than this threshold for 10 consecutive times in a certain comprehensive scenario, parameter iterative optimization is triggered. Iterative optimization is performed using the gradient descent method, with a step size of 0.

05. This step size is calibrated based on the optimization convergence speed and accuracy requirements of the image enhancement collaboration module parameters. The technical parameter weights of each sub-module in the image enhancement collaboration module are adjusted using this method, and the optimized technical parameters are stored in the scene parameter library, overwriting the original corresponding parameters in the library.

Images