Displacement monitoring method adopting visual analysis technology

Through visual analysis technology, the functional relationship between displacement and water level height and time is established, and the problem of low stability and accuracy of hardware monitoring equipment is solved, real-time accuracy and cost savings of displacement monitoring are achieved.

CN118966460APending Publication Date: 2024-11-15DALIAN DAKAI SEWAGE TREATMENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411159034.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

In the existing displacement monitoring methods, the hardware monitoring equipment has poor stability and flexibility, low accuracy, inconvenient installation and maintenance, and high cost.

Method used

Visual analysis technology is used to monitor the video data of the drainage pipeline, establish a model to identify the water level height, and combine the displacement data to establish a functional relationship between displacement and water level height and time, and use a multivariate nonlinear regression algorithm for prediction.

Benefits of technology

It realizes real-time, accurate, stable and efficient displacement monitoring, reduces equipment maintenance and procurement costs, and has the characteristics of life-long maintenance-free deployment, and the monitoring results are accurate and reliable.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118966460A_ABST
    Figure CN118966460A_ABST
Patent Text Reader

Abstract

The invention discloses a displacement monitoring method adopting a visual analysis technology, and belongs to the technical field of image recognition. According to the method, a visual analysis model is adopted for pipeline drainage water quantity monitoring, the visual analysis model is established mainly according to pipeline drainage historical image data, the pipeline water level height serves as a recognition target, and the displacement after the T moment is judged through the current water level height recognized by the model and time. And calculating the displacement at a certain moment in the future according to a function relation between the displacement and the water level height as well as the duration, and using the displacement as an early warning forecast. And the model belongs to a software system, is easy to install and maintain, can be tightly combined with a supervision platform, and realizes real-time, accurate, stable and efficient displacement monitoring.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a displacement monitoring method using visual analysis technology, and belongs to the technical field of image recognition, in which a model is established according to historical image data to perform real-time recognition on monitoring images. Background Art

[0002] Visual analysis, an artificial intelligence model, has been widely used in many fields, such as vehicle recognition models in intelligent traffic control, human recognition models in intelligent security, garbage classification models and sewage recognition models in intelligent environmental protection, etc. The application of these visual analysis models in the field of water quality monitoring and control is limited to real-time recognition of regulatory targets in the waters of monitoring points.

[0003] The current displacement monitoring method is to use various types of flow monitoring instruments / equipment for flow monitoring. The instruments and equipment need to be installed in the pre-pipeline, which is inconvenient to install, repair and replace. The service life is limited, and the installation and replacement are inconvenient, requiring a large amount of procurement funds. The flow monitoring instruments and equipment currently used have monitoring results that are greatly affected by the operating status of the instruments and equipment. Existing hardware monitoring products have problems such as poor stability and flexibility, and low accuracy. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a displacement monitoring method using visual analysis technology. According to the functional relationship between the displacement, water level height and duration, the displacement at a certain moment in the future is calculated, which can be used as an early warning forecast.

[0005] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: a displacement monitoring method using visual analysis technology, comprising the following steps:

[0006] S1. Install video surveillance equipment: The surveillance camera is installed facing the drainage pipe to obtain real-time monitoring images and save the surveillance video data;

[0007] S2. Accumulate raw data: Raw data is divided into two categories: image data and numerical data.

[0008] (1) Image data: images that can show the different water levels in the drainage pipes. At least 3,000 surveillance images of different water levels in the pipes are accumulated to train the visual parsing model;

[0009] (2) Numerical data: Accumulate known displacement data corresponding to different times and use them to combine with pipeline water level data to generate the corresponding functional relationship between water level, time and displacement.

[0010] S3. Use image data to build a visual model. The modeling process follows three sub-steps: labeling image data - encoding training model - saving model.

[0011] (1) Marking image data: Mark the pipeline water level monitoring images, use the water level as the identification target, and generate a water level marking file for each image;

[0012] (2) Encoding training model: Use the general yolov5 algorithm framework to load image data and labeling files, adjust the algorithm model preset parameters, and establish a visual parsing model.

[0013] (3) Save the model: After the model training is completed, save the model locally as a file for future loading and calling.

[0014] S4. After the visual model training is completed, the accumulated raw video data is screenshotted every minute. The water level is identified in each screenshot through the visual model, and the current water level height H of each picture is obtained, which is the ratio of the water level height to the pipe diameter, and the value range is between 0 and 1;

[0015] S5. Combine the current water level data with the known displacement data at different times to create a data table; use a multivariate nonlinear regression algorithm to perform mathematical modeling and fit the functional relationship between displacement, water level and duration:

[0016] V = β0 + β1H 2 + β2H + β3T 2 + β4T + β5HT + ε

[0017] Where: V represents the discharge volume after T time in the future according to the current water level height H, β0, β1, β2, β3, β4, β5 are regression model coefficients, and ε is the error constant.

[0018] The effects and benefits of the present invention are as follows: This solution uses a visual analysis model to monitor the amount of water discharged from pipelines. The visual analysis model is mainly established based on the historical image data of pipeline drainage, and the pipeline water level height is used as the identification target. The displacement size after time T is determined by the current water level height and time identified by the model. The monitoring pictures are intercepted at fixed frequencies (every second or every minute, etc.); the visual model is used to identify the pipeline water level height in the picture, and the pipeline water level height is identified; the displacement at a certain moment in the future is calculated based on the functional relationship between the displacement, the water level height and the duration, which can be used as an early warning forecast.

[0019] In addition, with the popularization of artificial intelligence technology, video surveillance equipment is widely used, and visual analysis models are easy to install and deploy, so that a set of monitoring equipment can be equipped with multiple visual analysis models, replacing the data collection functions of various types of detection instruments and equipment, and saving a lot of funds. The visual analysis model has the characteristics of one-time deployment and lifelong maintenance-free. The monitoring results are highly accurate and stable. The model is a software system, easy to install and maintain, and can be closely integrated with the supervision platform to achieve real-time, accurate, stable and efficient displacement monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart of a displacement monitoring method using visual analysis technology.

[0021] Figure 2 It is a flow chart of the application of monitoring method. DETAILED DESCRIPTION

[0022] The technical problem to be solved by the present invention is to solve the problem that existing hardware monitoring products have poor stability and flexibility and low accuracy.

[0023] To this end, this solution uses a visual analysis model to monitor the amount of water discharged from pipelines. The visual analysis model is mainly established based on the historical image data of pipeline drainage, and the pipeline water level height is used as the recognition target. The current water level height and time recognized by the model are used to determine the amount of water discharged after time T.

[0024] Figure 1 The figure is a flow chart of the displacement monitoring method using visual analysis technology. The specific scheme of the monitoring method is as follows:

[0025] S1. Install video surveillance equipment: The surveillance camera is installed facing the drainage pipe to obtain real-time monitoring images and save the surveillance video data;

[0026] S2. Accumulate raw data: Raw data is divided into two categories: image data and numerical data.

[0027] (1) Image data: images that can show the different water levels in the drainage pipes. At least 3,000 surveillance images of different water levels in the pipes are accumulated to train the visual parsing model;

[0028] (2) Numerical data: Accumulate known displacement data corresponding to different times and use them to combine with pipeline water level data to generate the corresponding functional relationship between water level, time and displacement.

[0029] Time (min) Displacement (V) 2020-01-01 10:00:00 15 2020-01-01 10:01:00 45 2020-01-01 10:02:00 75 2020-01-01 10:03:00 103 ... ...

[0030] The accumulation of raw data usually takes a long time, and depending on the actual emission conditions, it usually takes more than several months.

[0031] S3. After the original data is accumulated, the image data is used to build the visual model. The modeling process follows three basic sub-steps: labeling image data - encoding training model - saving model.

[0032] (1) Marking image data: Mark the pipeline water level monitoring images, use the water level as the identification target, and generate a water level marking file for each image;

[0033] (2) Encoding training model: Use the general YOLOv5 algorithm framework to write business logic code, load image data and labeling files, adjust the algorithm model preset parameters, and establish a visual parsing model.

[0034] (3) Save the model: After the model training is completed, save the model locally as a file for future loading and calling.

[0035] S4. After the visual model training is completed, a screenshot of the accumulated original video data is taken every minute. The water level in each screenshot is identified through the visual model, and the current water level height H of each picture is obtained. This height is the relative height relative to the pipe diameter, which is the ratio of the water level height to the pipe diameter. The value range is between 0-1, and the recognition result is saved as a data table file.

[0036] Time (min) Water level (H) 2020-01-01 10:00:00 0.1 2020-01-01 10:01:00 0.1 2020-01-01 10:02:00 0.2 2020-01-01 10:03:00 0.2 ... ...

[0037] S5. Combine the water level data with the known displacement data at different times to create a data table

[0038] Water level (H) Duration (min) Displacement (V) 0.1 2 60 0.2 3 115 ... ... ...

[0039] S6. Based on the above table data, a multivariate nonlinear regression algorithm is used to perform mathematical modeling and fit the functional relationship between discharge volume, water level height and duration:

[0040] V = β0 + β1H 2 + β2H + β3T 2 + β4T + β5HT + ε

[0041] Where: V represents the discharge volume after T time in the future according to the current water level height H, β, ε are the regression model coefficients and error constants.

[0042] The specific application of the monitoring method is as follows (such as Figure 2 shown):

[0043] (1) Capture surveillance images at a fixed frequency (every second or every minute, etc.);

[0044] (2) Use the visual model to identify the water level in the pipe in the image;

[0045] (3) Based on the functional relationship between discharge volume, water level height and duration, the discharge volume at a certain moment in the future can be calculated and used as an early warning forecast.

Claims

1. A displacement monitoring method using visual analysis technology, characterized in that: The following steps are involved: S1. Install video surveillance equipment: The surveillance camera is installed facing the drainage pipe to obtain real-time monitoring images and save the surveillance video data; S2. Accumulate raw data: Raw data can be divided into two categories: image data and numerical data; (1) Image data: images that can show the different water levels in the drainage pipes. At least 3,000 surveillance images of different water levels in the pipes are accumulated to train the visual parsing model; (2) Numerical data: accumulate known displacement data corresponding to different times, and use it to combine with pipeline water level data to generate the corresponding functional relationship between water level, time and displacement; S3. Use image data to build a visual model. The modeling process follows three sub-steps: labeling image data - encoding training model - saving model; (1) Marking image data: Mark the pipeline water level monitoring images, use the water level as the identification target, and generate a water level marking file for each image; (2) Encoding training model: Use the general YOLOv5 algorithm framework to load image data and label files, adjust the algorithm model preset parameters, and establish a visual parsing model; (3) Save the model: After the model training is completed, save the model locally in the form of a file for future loading and calling; S4. After the visual model training is completed, the accumulated raw video data is screenshotted every minute. The water level is identified in each screenshot through the visual model, and the current water level height H of each picture is obtained, which is the ratio of the water level height to the pipe diameter, and the value range is between 0 and 1; S5. Combine the current water level data with the known displacement data at different times to create a data table; use a multivariate nonlinear regression algorithm to perform mathematical modeling and fit the functional relationship between displacement, water level and duration: V = β0+ β1H 2 + β2H + β3T 2 + β4T + β5HT + e Where: V represents the discharge volume after T time in the future according to the current water level height H, β0, β1, β2, β3, β4, β5 are regression model coefficients, and ε is the error constant.