A sewer pipeline fault classification method, device, medium and product

Abstract

The application discloses a sewage pipeline fault classification method and device, medium and product, and relates to the technical field of sewage pipeline fault classification. The method comprises the following steps: acquiring a pipeline internal video, then performing pretreatment and definition screening to obtain a plurality of clear images; determining a plurality of key frame images according to the key frame confidence of all clear images; for each key frame image, inputting the key frame image into a trained first fault classification model and a trained second fault classification model to obtain the confidence of each category; and finally determining the final fault category of the target sewage pipeline according to the first category confidence, the second category confidence, the third category confidence, the fourth category confidence, the fifth category confidence and the sixth category confidence corresponding to all key frame images. The application can guarantee the safety of the detection personnel and improve the accuracy of the sewage pipeline fault classification.

Description

Technical Field

[0001] This invention relates to the field of sewage pipeline fault classification technology, and in particular to a sewage pipeline fault classification method, device, medium and product based on deep learning. Background Technology

[0002] With the continuous development of cities and the advancement of urbanization, the scale of urban drainage facilities is constantly expanding, making their maintenance increasingly difficult. Traditional manual maintenance methods require professional inspectors to enter the pipes, which is inefficient and may pose a safety hazard to the inspectors due to contaminants inside the pipes. While video-based inspection methods acquire video footage of the pipe interior using various cameras, allowing inspectors to visually diagnose faults based on the video, this method is susceptible to subjective biases and has certain limitations. Summary of the Invention

[0003] The purpose of this invention is to provide a method, device, medium, and product for classifying sewage pipeline faults. This method can classify sewage pipeline faults using deep learning algorithms while ensuring the safety of inspection personnel. By learning the patterns of sewage pipeline fault classification through deep learning, it has strong objectivity and avoids the influence of subjective human factors, thereby improving the accuracy of sewage pipeline fault classification.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A method for classifying sewage pipeline faults, comprising:

[0006] Acquire video footage of the inside of the pipeline; the video footage of the target sewage pipeline is obtained by filming; the video footage of the inside of the pipeline includes multiple video frames;

[0007] The video frames in the video inside the pipe are preprocessed to obtain preprocessed video frames; the sharpness of all the preprocessed video frames is filtered to obtain several clear images.

[0008] For each of the aforementioned clear images, the clear image is input into the trained keyframe extraction model to obtain the keyframe confidence of the clear image;

[0009] A number of keyframe images are determined based on the keyframe confidence of all the clear images; the keyframe images are images that simultaneously contain all the faults contained in the video inside the pipe;

[0010] For each keyframe image, the keyframe image is input into a trained first fault classification model to obtain the first category confidence score, second category confidence score, and third category confidence score corresponding to the keyframe image; the first category confidence score is the confidence score that the keyframe image is predicted to be classified as a deposition fault; the second category confidence score is the confidence score that the keyframe image is predicted to be classified as a comprehensive fault; the third category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe misalignment fault; the comprehensive fault includes root intrusion faults, pipe wall corrosion faults, and pipe wall scaling faults;

[0011] The keyframe image is input into the trained second fault classification model to obtain the fourth, fifth, and sixth category confidence scores corresponding to the keyframe image; the fourth category confidence score is the confidence score that the keyframe image is predicted to be classified as a tree root intrusion fault; the fifth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall corrosion fault; and the sixth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall scaling fault.

[0012] The final fault category of the target sewage pipeline is determined based on the first category confidence score, second category confidence score, third category confidence score, fourth category confidence score, fifth category confidence score, and sixth category confidence score corresponding to all the keyframe images.

[0013] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described sewage pipeline fault classification method.

[0014] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described sewage pipeline fault classification method.

[0015] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described sewage pipeline fault classification method.

[0016] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The present invention provides a method, apparatus, medium, and product for classifying sewage pipeline faults. First, video of the inside of the pipeline is acquired. Then, the video frames in the video of the inside of the pipeline are preprocessed to obtain preprocessed video frames. All preprocessed video frames are filtered for clarity to obtain several clear images. For each clear image, the clear image is input into a trained keyframe extraction model to obtain the keyframe confidence score of the clear image. Several keyframe images are determined based on the keyframe confidence scores of all clear images. For each keyframe image, the keyframe image is input into a trained first fault classification model to obtain the first category confidence score, second category confidence score, and third category confidence score corresponding to the keyframe image. Simultaneously, the keyframe image is input into a trained second fault classification model to obtain the fourth category confidence score, fifth category confidence score, and sixth category confidence score corresponding to the keyframe image. The final fault category of the target sewage pipeline is determined based on the first category confidence score, second category confidence score, third category confidence score, fourth category confidence score, fifth category confidence score, and sixth category confidence score corresponding to all keyframe images. This invention can classify sewage pipeline faults using deep learning algorithms while ensuring the safety of inspection personnel. By learning the patterns of sewage pipeline fault classification through deep learning, it has strong objectivity and avoids the influence of subjective human factors, thereby improving the accuracy of sewage pipeline fault classification. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the sewage pipeline fault classification method provided in Embodiment 1 of the present invention;

[0019] Figure 2 This is a schematic diagram illustrating the specific implementation process of the sewage pipeline fault classification method provided in Embodiment 1 of the present invention;

[0020] Figure 3 An internal structural diagram of a computer device is provided for embodiments of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Current automatic sewage pipeline fault classification schemes require the use of large datasets to cover as many possible sewage pipeline environments as possible, but in reality, it is difficult to obtain sewage pipeline fault data frames. Furthermore, existing schemes are designed for sewage pipeline images, while in actual detection, equipment usually acquires sewage pipeline videos rather than images, which makes many automatic sewage pipeline fault classification schemes difficult to apply in practice.

[0023] The purpose of this invention is to provide a method, device, medium, and product for classifying sewage pipeline faults. It aims to ensure the safety of inspection personnel while employing deep learning algorithms for sewage pipeline fault classification. By using deep learning to identify the patterns in sewage pipeline fault classification, it achieves strong objectivity, avoiding the influence of subjective human factors and thus improving the accuracy of sewage pipeline fault classification. This invention also presents a sewage pipeline fault classification scheme based on transfer learning. Specifically, it uses transfer learning methods to train a keyframe extraction model, a first fault classification model, and a second fault classification model, solving the problem of low accuracy in fault classification schemes when acquiring sewage pipeline fault data frames is difficult.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] Example 1

[0026] like Figure 1 As shown, a sewage pipeline fault classification method in this embodiment includes:

[0027] S1: Acquire video of the inside of the pipe; the video of the inside of the pipe is a video obtained by shooting the target sewage pipe; the video of the inside of the pipe includes multiple video frames.

[0028] S2: Preprocess the video frames in the video inside the pipe to obtain preprocessed video frames; perform clarity filtering on all the preprocessed video frames to obtain several clear images.

[0029] S3: For each of the clear images, input the clear image into the trained keyframe extraction model to obtain the keyframe confidence of the clear image; determine a number of keyframe images based on the keyframe confidence of all the clear images; the keyframe images are images that simultaneously contain all the faults contained in the video inside the pipeline.

[0030] S4: For each keyframe image, input the keyframe image into the trained first fault classification model to obtain the first category confidence, second category confidence, and third category confidence corresponding to the keyframe image; the first category confidence is the confidence that the keyframe image is predicted to be classified as a deposition fault; the second category confidence is the confidence that the keyframe image is predicted to be classified as a comprehensive fault; the third category confidence is the confidence that the keyframe image is predicted to be classified as a pipe misalignment fault; the comprehensive fault classification model... The fault classification model includes tree root intrusion faults, pipe wall corrosion faults, and pipe wall scaling faults. The keyframe images are input into a trained second fault classification model to obtain the fourth, fifth, and sixth category confidence scores corresponding to the keyframe images. The fourth category confidence score is the confidence score that the keyframe image is predicted to be classified as a tree root intrusion fault; the fifth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall corrosion fault; and the sixth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall scaling fault.

[0031] S5: Determine the final fault category of the target sewage pipeline based on the first category confidence, second category confidence, third category confidence, fourth category confidence, fifth category confidence and sixth category confidence corresponding to all the keyframe images.

[0032] This embodiment provides a sewage pipeline fault classification method that identifies fault types from sewage pipeline videos (i.e., videos of the pipeline interior). The method consists of three steps: extracting clear frames from the video, extracting keyframes, and identifying the fault type. The keyframe extraction and fault type identification steps utilize a neural network model. This embodiment achieves sewage pipeline fault classification based on a neural network model using transfer learning on small datasets.

[0033] 1. Extract clear frames from the video: including steps S1-S2.

[0034] Before preprocessing the video frames in the video inside the pipe in step S2, the method further includes: sampling the video frames in the video inside the pipe to obtain several video frame groups; each video frame group includes a set number of consecutive video frames. The set number can be 3.

[0035] Specifically, continuous frames are extracted from the video inside the pipe at equal time intervals. In this embodiment, three consecutive frames can be obtained at 0.5-second intervals as a group of video frames, thereby sampling the viewpoint image (i.e., video frame) that appears for more than 0.5 seconds in the video inside the pipe.

[0036] In step S2, the preprocessing is convolution processing; preprocessing the video frames in the video inside the pipe to obtain preprocessed video frames specifically includes:

[0037] Using the Laplacian operator, each video frame in each video frame group is convolved to obtain the preprocessed video frame corresponding to the video frame.

[0038] In step S2, the sharpness of all the preprocessed video frames is filtered to obtain several clear images, specifically including:

[0039] The clarity of the preprocessed video frame is calculated based on the image information in the preprocessed video frame;

[0040] For each video frame group, a clear image in the video frame group is determined based on the sharpness of the preprocessed video frame corresponding to all the video frames in the video frame group and a set sharpness.

[0041] A 3×3 Laplacian operator is used to convolve the grayscale image of each video frame in each of the aforementioned video frame groups. The variance of the convolved image (i.e., the preprocessed video frame) is then used as the sharpness of that preprocessed video frame. The sharpness of the preprocessed video frame is then used as the sharpness of the corresponding video frame in the video frame group. Finally, the video frame with the highest sharpness exceeding a set sharpness in each video frame group is extracted as the selected sharp image. In this embodiment, the set sharpness can be 40. The sharpness calculation formula is as follows:

[0042]

[0043] Where s represents the sharpness of the image after convolution; x i,j This represents the gray value at position (i,j) in the image after convolution; This represents the mean grayscale value of the image after convolution, and n is the number of pixels in the image after convolution.

[0044] II. Extracting keyframes: including step S3.

[0045] Each clear image obtained in step S2 is fed into the trained keyframe extraction model. The trained keyframe extraction model is used to predict the confidence level of the clear image as a keyframe, and the keyframe confidence level of the clear image is obtained.

[0046] The keyframes mentioned above refer to frames in the video that simultaneously contain all types of faults (including deposition faults, root intrusion faults, pipe wall corrosion faults, pipe wall scaling faults, and pipe misalignment faults).

[0047] The trained keyframe extraction model is a ResNet101 model. Before inputting the clear image into the trained keyframe extraction model, the process further includes: training the keyframe extraction model, specifically including:

[0048] First, the ResNet101 model is pre-trained using the ImageNet dataset to obtain the first pre-trained model. Then, the first pre-trained model is fine-tuned using transfer learning on a dataset of manually extracted keyframes from sewage pipe videos to obtain a trained keyframe extraction model.

[0049] After obtaining the keyframe confidence score for each clear image, several keyframe images are determined based on the keyframe confidence scores of all the clear images. That is, a preset number of clear images with the highest keyframe confidence scores are selected as keyframe images. In this embodiment, the five clear images with the highest confidence scores (above 0.6) can be selected and output as keyframe images to step S4.

[0050] III. Identify the fault type: including steps S4-S5.

[0051] Step S4 specifically includes: inputting the preset number of keyframe images obtained in step S3 into fault classification model A (i.e., the trained first fault classification model) and fault classification model B (i.e., the trained second fault classification model). Each trained fault classification model will obtain confidence scores for 3 categories, and the two trained fault classification models will obtain a total of 6 category confidence scores p. k (k = 1, 2, 3, 4, 5, 6). Where p1 is the first category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a deposition fault; p2 is the second category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a combined fault; p3 is the third category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a pipe misalignment fault; p4 is the fourth category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a tree root intrusion fault; p5 is the fifth category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a pipe wall corrosion fault; and p6 is the sixth category confidence score for the keyframe image, i.e., the confidence score that the keyframe image is predicted to be classified as a pipe wall scaling fault.

[0052] The trained first fault classification model and the trained second fault classification model are Query2Label models. The backbone network of the Query2Label model is a ResNet101 model. That is, in step S4, fault classification model A and fault classification model B are both ResNet101 models used as the backbone Query2Label models. The backbone ResNet101 model is pre-trained using the ImageNet dataset to obtain the second pre-trained model. In subsequent training, the second pre-trained model is fine-tuned using transfer learning with asymmetric loss. The dataset used by fault classification model A (first fault classification model) contains data frames containing deposition faults, root intrusion faults, pipe wall corrosion faults, pipe wall scaling faults, and pipe misalignment faults. Root intrusion faults, pipe wall corrosion faults, and pipe wall scaling faults are labeled as the same class. That is, the dataset used for training fault classification model A (first fault classification model) includes data frames labeled as deposition faults, combined category faults, and pipe misalignment faults. Fault classification model B (second fault classification model) is trained using only data frames labeled as root intrusion fault, pipe wall corrosion fault, and pipe wall scaling fault.

[0053] In step S5, the final fault category of the target sewage pipeline is determined based on the first category confidence score, second category confidence score, third category confidence score, fourth category confidence score, fifth category confidence score, and sixth category confidence score corresponding to all the keyframe images. Specifically, this includes:

[0054] For each central fault category, the average confidence score of the central category corresponding to all keyframe images is calculated to obtain the average confidence score corresponding to the central fault category; the central fault category is deposition fault, comprehensive category fault, pipe misalignment fault, tree root intrusion fault, pipe wall corrosion fault, or pipe wall scaling fault; the central category confidence score is the category confidence score corresponding to the keyframe image being predicted to be classified as the fault category.

[0055] The final fault category of the target sewage pipeline is determined based on the average confidence level corresponding to all the central fault categories.

[0056] Calculate p for the category confidence of a preset number of keyframe images. k mean p k (k = 1, 2, 3, 4, 5, 6), where each fault category is taken as the central fault category, and the category confidence corresponding to the central fault category is taken as the central category confidence. Under this central fault category, the mean of the central category confidence corresponding to all keyframe images classified as the central fault category is calculated to obtain the average confidence p corresponding to this central fault category.ik (k = 1, 2, 3, 4, 5, 6). For example, when the fault category is deposition fault, i.e., k = 1, the average confidence level p corresponding to deposition fault is... j1 = First category confidence score of all keyframe images divided by the preset number.

[0057] The P obtained above jk Substituting into equation (2) yields the final fault category ( Figure 2 (Classification results in the text):

[0058]

[0059] Where class represents the final fault category, class = 1 indicates that the fault belongs to this category, for example, class 沉积 =1 indicates that the pipeline has a deposition fault, meaning the final fault category is deposition fault. P j2 This represents the average confidence level for the second category, i.e., the average confidence level when the fault category is a composite fault; P j3 P represents the average confidence level for the third category, i.e., the average confidence level corresponding to the fault category of pipe misalignment fault; j4 This represents the average confidence level for the fourth category, i.e., the average confidence level corresponding to the fault category of tree root intrusion fault; P j5 P represents the average confidence level for the fifth category, i.e., the average confidence level corresponding to the fault category of pipe wall corrosion fault; j6 This represents the average confidence level for the sixth category, i.e., the average confidence level when the fault category is pipe wall scaling fault.

[0060] This embodiment has the following beneficial effects:

[0061] 1. This invention addresses the problems existing in current automatic sewage pipeline fault classification schemes by proposing a sewage pipeline fault classification scheme based on transfer learning, which solves the problem of low accuracy of fault classification schemes when it is difficult to obtain sewage pipeline fault data frames.

[0062] 2. This invention uses methods such as extracting keyframes to effectively identify fault categories from sewage pipeline videos, solving the problem that automatic sewage pipeline fault classification systems usually need to detect faults from videos rather than images.

[0063] Example 2

[0064] A computer device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of a sewage pipe fault classification method in Embodiment 1.

[0065] Example 3

[0066] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a sewage pipeline fault classification method in Embodiment 1.

[0067] Example 4

[0068] A computer program product includes a computer program that, when executed by a processor, implements the steps of a sewage pipeline fault classification method in Embodiment 1.

[0069] Example 5

[0070] A computer device, which may be a database, may have an internal structure diagram as shown below. Figure 3 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores pending transactions. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a sewage pipeline fault classification method as described in Embodiment 1.

[0071] It should be noted that the object information (including but not limited to object device information, object personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this invention are all information and data authorized by the object or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided by this invention may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided by this invention may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0073] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0074] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for classifying sewage pipeline faults, characterized in that, include: Acquire video footage of the inside of the pipeline; the video footage of the target sewage pipeline is obtained by filming; the video footage of the inside of the pipeline includes multiple video frames; The video frames in the video inside the pipe are preprocessed to obtain preprocessed video frames; the sharpness of all the preprocessed video frames is filtered to obtain several clear images. For each of the aforementioned clear images, the clear image is input into the trained keyframe extraction model to obtain the keyframe confidence of the clear image; A number of keyframe images are determined based on the keyframe confidence of all the clear images; the keyframe images are images that simultaneously contain all the faults contained in the video inside the pipe; For each keyframe image, the keyframe image is input into the trained first fault classification model to obtain the first category confidence, second category confidence, and third category confidence corresponding to the keyframe image; the first category confidence is the confidence that the keyframe image is predicted to be classified as a deposition fault. The second category confidence level is the confidence level that the keyframe image is predicted to be classified as a comprehensive category fault; the third category confidence level is the confidence level that the keyframe image is predicted to be classified as a pipe misalignment fault; the comprehensive category faults include root intrusion faults, pipe wall corrosion faults, and pipe wall scaling faults; The keyframe image is input into the trained second fault classification model to obtain the fourth, fifth, and sixth category confidence scores corresponding to the keyframe image; the fourth category confidence score is the confidence score that the keyframe image is predicted to be classified as a tree root intrusion fault; the fifth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall corrosion fault; and the sixth category confidence score is the confidence score that the keyframe image is predicted to be classified as a pipe wall scaling fault. The final fault category of the target sewage pipeline is determined based on the first category confidence score, second category confidence score, third category confidence score, fourth category confidence score, fifth category confidence score, and sixth category confidence score corresponding to all the keyframe images, specifically including: For each central fault category, the average confidence score of the central category corresponding to all keyframe images is calculated to obtain the average confidence score corresponding to the central fault category; the central fault category is deposition fault, comprehensive category fault, pipe misalignment fault, tree root intrusion fault, pipe wall corrosion fault, or pipe wall scaling fault; the central category confidence score is the category confidence score corresponding to the keyframe image being predicted to be classified as the fault category. The final fault category of the target sewage pipeline is determined based on the average confidence level corresponding to all the central fault categories.

2. The sewage pipeline fault classification method according to claim 1, characterized in that, Before preprocessing the video frames in the video inside the pipe, the process also includes: The video frames inside the pipeline are sampled to obtain several video frame groups; each video frame group includes a set number of consecutive video frames.

3. The sewage pipeline fault classification method according to claim 2, characterized in that, The preprocessing is a convolution process; The video frames inside the pipe are preprocessed to obtain preprocessed video frames, specifically including: Using the Laplacian operator, each video frame in each video frame group is convolved to obtain the preprocessed video frame corresponding to the video frame.

4. The sewage pipeline fault classification method according to claim 3, characterized in that, The sharpness of all the preprocessed video frames is filtered to obtain several sharp images, specifically including: The sharpness of the preprocessed video frame is calculated based on the image information in the preprocessed video frame; For each video frame group, a clear image in the video frame group is determined based on the sharpness of the preprocessed video frame corresponding to all the video frames in the video frame group and a set sharpness.

5. The sewage pipeline fault classification method according to claim 1, characterized in that, The trained first fault classification model and the trained second fault classification model are Query2Label models; the backbone network of the Query2Label model is a ResNet101 model.

6. The sewage pipeline fault classification method according to claim 1, characterized in that, The trained keyframe extraction model is a ResNet101 model.

7. A computer device, comprising: The memory and processor contain a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the steps of the sewage pipeline fault classification method according to any one of claims 1-6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of a sewage pipeline fault classification method according to any one of claims 1-6.

9. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of a sewage pipeline fault classification method as described in any one of claims 1-6.

Images