Industrial ash emission detection method and system based on target identification

Abstract

The invention discloses an industrial dust emission detection method and system based on target recognition, and relates to the technical field of target recognition, and the method comprises the steps: deploying a camera at an industrial pipeline, collecting a pipeline image to obtain a video stream, and carrying out the first recognition of the video stream based on deep learning and a target recognition model of a neural network; cutting the picture which is identified for the first time and has ash emission or air leakage, extracting key feature points in an image area in which the ash emission or air leakage part is acquired, and calculating motion vectors of the feature points in continuous frames of images; analyzing the motion trend of dust emission or air leakage, and judging whether the dust emission or air leakage phenomenon exists or not; and if the phenomenon of dust emission or air leakage exists, evaluating the emergency degree based on the movement speed, the direction consistency and the area change of the diffusion region. According to the invention, deep learning and computer vision technologies are adopted to analyze the motion vector speed, direction consistency and diffusion area change, the severity of dust emission and gas leakage events is intelligently evaluated, and intelligentization and automation of industrial pipeline monitoring are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of target recognition, in particular to an industrial dust emission detection method and system based on target recognition. Background Art

[0002] In industrial production, dust and gas leakage in pipelines are common problems. Traditional detection methods mainly rely on manual inspections or simple sensor monitoring, which have the following technical problems:

[0003] Low detection efficiency: Manual inspections are time-consuming and labor-intensive, and sensor coverage is limited;

[0004] High false alarm rate: The sensor is easily affected by environmental interference, resulting in false alarms;

[0005] Unable to respond in real time: Traditional methods have delays and it is difficult to detect problems in time;

[0006] Unable to assess urgency: It can only detect anomalies but cannot determine the extent of spread and urgency;

[0007] Insufficient intelligence: Reliance on manual judgment and lack of intelligent analysis capabilities.

[0008] To address these issues, the present invention provides a method for detecting dust and gas leaks in industrial pipelines based on target recognition and optical flow. This method uses a target recognition model to identify dust and gas leaks in the pipeline in real time, and uses optical flow to perform dynamic verification, reducing false alarm rates. Furthermore, the method calculates the urgency level based on the spread of dust and gas leaks, enabling intelligent, real-time pipeline safety monitoring. Summary of the Invention

[0009] In view of the above-mentioned problems, the present invention is proposed.

[0010] Therefore, the problem to be solved by the present invention is that the traditional detection method has the problems of low detection efficiency, high false alarm rate, inability to respond in real time, inability to assess the degree of urgency and lack of intelligence.

[0011] In order to solve the above technical problems, the present invention provides the following technical solutions: an industrial dust detection method based on target recognition, which includes: deploying a camera at an industrial pipeline, collecting pipeline images to obtain a video stream, and performing a first recognition of the video stream based on a target recognition model of deep learning and a neural network; cropping the first identified picture of dust or air leakage, extracting key feature points in the image area of ​​the dust or air leakage part, and calculating the motion vector of the feature point in the continuous frame image; analyzing the movement trend of dust or air leakage to determine whether dust or air leakage exists; if dust or air leakage exists, evaluating the urgency based on the movement speed, directional consistency and change in the diffusion area.

[0012] As a preferred solution of the industrial dust detection method based on target recognition described in the present invention, the deployment of cameras at industrial pipelines includes installing high-definition industrial cameras at key areas of industrial pipelines, including elbows, valves and connections, covering all locations where dust and air leakage may occur; deploying a server in the camera for running target recognition and motion analysis programs in real time, and interconnecting with the network.

[0013] As a preferred solution of the industrial dust detection method based on target recognition described in the present invention, the target recognition model based on deep learning and neural network performs the first recognition of the video stream, including collecting historical industrial pipeline dust and gas leakage image data, which must cover different lighting conditions, pipeline materials and background environments; annotating the historical image data, and the annotation content includes the location and category of the dust and gas leakage areas, expanding the historical image data through data enhancement technology, using the YOLOv5 (YouOnly LookOnceversion5) framework to train the target recognition model, and inputting the annotated historical image data; preprocessing the input image data, filtering the low-confidence false detection results according to the confidence threshold, and training the model.

[0014] As a preferred solution of the industrial dust detection method based on target recognition described in the present invention, the target recognition model based on deep learning and neural network performs the first recognition of the video stream, and also includes inputting the video stream into the trained target recognition model for the first recognition, and inputting the picture with the dust or air leakage identified into the motion analysis program to perform motion analysis on the picture.

[0015] As a preferred solution of the industrial dust detection method based on target recognition of the present invention, wherein: the calculation of the motion vector of the feature point in the continuous frame image includes: for the continuous frame I t and I t+1 , let the motion vector of the feature point p(x,y) be (u,v), and the optical flow constraint equation is expressed as,

[0016] I x u+I y v+I t =0

[0017] Among them, I x , I y It is represented as the image gradient, u is represented as the direction of the motion vector, and v is represented as the speed of the motion vector. A window W is selected around the feature point, and the error is minimized as follows:

[0018]

[0019] By least squares solution:

[0020]

[0021] After detecting multiple areas of dust or air leakage, calculate the optical flow vectors (u i , v i ).

[0022] As a preferred solution of the industrial dust detection method based on target recognition of the present invention, the determination of whether dust or air leakage exists includes calculating the average movement direction of the feature points. and average speed Expressed as,

[0023]

[0024] If the motion direction in adjacent frames is consistent, it indicates that there is motion in the abnormal area rather than static interference; the diffusion degree D of the feature point is calculated as,

[0025]

[0026] Where N represents the total number of feature points. If D exceeds the diffusion threshold, it means that the abnormal area is spreading, which is consistent with the characteristics of air leakage or dust emission.

[0027] As a preferred solution of the industrial dust detection method based on target recognition of the present invention, wherein: the evaluation of the urgency includes calculating the average optical flow velocity V expressed as,

[0028]

[0029] The final urgency determination formula E is expressed as:

[0030] E=αV+βD+γ(A t+1 -A t )

[0031] Among them, α, β, and γ are weight coefficients, A t It is expressed as the diffusion area of ​​dust or gas leakage at time t, A t+1 It is expressed as the diffusion area of ​​dust or air leakage at time t+1; E is compared with the preset threshold to evaluate the urgency.

[0032] In order to solve the above technical problems, the present invention provides the following technical solutions: an industrial dust detection system based on target recognition, comprising: an image acquisition module, a first recognition module and a motion analysis module; the image acquisition module obtains a video stream based on the pipeline image captured by the camera; the first recognition module performs a first recognition on the video stream based on the target recognition model of deep learning and neural network; the motion analysis module crops the picture with dust or air leakage identified for the first time, extracts key feature points in the image area of ​​the dust or air leakage part, calculates the motion vector of the feature point in the continuous frame image, analyzes the motion trend of dust or air leakage, and determines whether dust or air leakage exists. If dust or air leakage exists, the urgency is evaluated based on the movement speed, direction consistency and change in the diffusion area.

[0033] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the industrial dust emission detection method based on target recognition when executing the computer program.

[0034] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the industrial dust emission detection method based on target recognition as described above.

[0035] The beneficial effects of the present invention are as follows: the present invention automatically identifies dust and air leakage through a target recognition model (such as YOLO), thereby avoiding the time-consuming and labor-intensive problems of traditional manual inspections and greatly improving the detection speed and coverage.

[0036] The present invention verifies the target movement trend, effectively eliminates interference sources such as static stains and light and shadow changes, and significantly reduces the false alarm rate.

[0037] The present invention intelligently evaluates the severity of dust emission and air leakage events by analyzing the motion vector speed, directional consistency and diffusion area changes, providing a quantitative basis for subsequent operation and maintenance decisions.

[0038] The entire process uses deep learning and computer vision technologies to realize the intelligent and automated monitoring of industrial pipelines, reduce dependence on manual experience, and improve the overall reliability and adaptability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1This is a flow chart of the industrial dust detection method based on target recognition in Example 1.

[0041] Figure 2 This is a dust emission or air leakage phenomenon identification diagram of the industrial dust emission detection method based on target recognition in Example 1.

[0042] Figure 3 This is the first moment diagram of the movement trend of the industrial dust detection method based on target recognition in Example 2.

[0043] Figure 4 This is the second moment diagram of the movement trend of the industrial dust detection method based on target recognition in Example 2.

[0044] Figure 5 This is the third moment diagram of the movement trend of the industrial dust detection method based on target recognition in Example 2.

[0045] Figure 6 This is the fourth moment diagram of the movement trend of the industrial dust detection method based on target recognition in Example 2.

[0046] Figure 7 This is the fifth moment diagram of the movement trend of the industrial dust detection method based on target recognition in Example 2. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0049] Example 1, with reference to Figure 1-Figure 7 , which is the first embodiment of the present invention, provides an industrial dust detection method based on target recognition, including: Figure 1 As shown:

[0050] Step 1: Deploy cameras on industrial pipelines to capture pipeline images and generate video streams.

[0051] First, in terms of hardware, high-definition industrial cameras must be installed in key areas of industrial pipelines (such as elbows, valves, and joints) to ensure coverage of all locations where dust and gas leaks may occur. Servers must also be deployed to run real-time target recognition and motion analysis programs, ensuring network connectivity.

[0052] Step 2: Perform the first recognition of the video stream based on the target recognition model of deep learning and neural network.

[0053] In order to ensure that the images of dust emission and air leakage can be accurately identified in a specific environment, this embodiment collects a large amount of historical industrial pipeline dust emission and air leakage image data, covering different lighting conditions, pipeline materials and background environments. The image data is labeled, and the annotation content includes the location and category of the dust emission and air leakage area. The diversity of the material is improved through data enhancement technology (such as rotation, scaling, adding noise, etc.). The target recognition model is trained using the YOLOv5 framework and the labeled data set is input. During model training, the input image is preprocessed (such as normalization, scaling, etc.) to ensure that the input data meets the model requirements. Finally, the low-confidence false detection results are filtered according to the confidence threshold to ensure the accuracy of the detection results. Images recognized by the target recognition model are as follows: Figure 2 shown.

[0054] Step 3: crop the first identified image with dust or air leakage, extract key feature points from the image area of ​​the dust or air leakage portion, and calculate the motion vectors of the feature points in the continuous frame images.

[0055] The extracted key feature points can include:

[0056] Dust / leakage edge contour feature points: Select locations with strong contrast at the edges of dust masses or air leakage groups. Changes in edge positions can clearly reflect changes in the morphology of dust / leakage, helping to accurately track movement and diffusion trends.

[0057] Texture change feature points are selected within the leakage area where the texture suddenly changes. The subtle texture in the leakage will be displaced with the movement of the airflow. Tracking these points can effectively capture the leakage activity.

[0058] Brightness gradient extreme point (grayscale mutation point): Select the area where the brightness changes sharply (such as the initial point of dust emission or the center of air leakage). The grayscale gradient mutation point represents the starting position of the movement and helps to determine the source of the movement and the initial diffusion speed.

[0059] Representative points in areas with dense feature distribution: In the motion area, selecting several representative points in local areas with high density of feature points can reduce the amount of calculation. At the same time, it represents the overall motion trend and helps to accurately infer the overall diffusion motion.

[0060] Point groups with strong dynamic consistency: Prioritize the selection of feature point groups with small changes in motion direction and speed in consecutive frames. This can be used to further verify the authenticity of the motion and eliminate environmental disturbances.

[0061] Calculating the motion vector of the feature point in the continuous frame image includes: t and It+1 , let the motion vector of the feature point p(x,y) be (u,v), and the optical flow constraint equation is expressed as,

[0062] I x u+I y v+I t =0

[0063] Among them, I x , I y It is represented as image gradient, u is the direction of motion vector, and v is the speed of motion vector.

[0064] Select a window W around the feature point and minimize the error expressed as,

[0065]

[0066] By least squares solution:

[0067]

[0068] After detecting multiple areas of dust or air leakage, calculate the optical flow vectors (u i , v i ).

[0069] Step 4: Analyze the movement trend of dust emission or air leakage to determine whether dust emission or air leakage exists.

[0070] Calculate the average motion direction of feature points and average speed Expressed as,

[0071]

[0072] If the motion direction in adjacent frames is consistent, it indicates that there is motion in the abnormal area rather than static interference.

[0073] The diffusion degree D of the feature point is calculated as follows:

[0074]

[0075] Where N represents the total number of feature points. If D exceeds the diffusion threshold, it means that the abnormal area is spreading, which is consistent with the characteristics of air leakage or dust emission.

[0076] Step 5: If dust or air leakage exists, assess the urgency based on the movement speed, directional consistency, and changes in the diffusion area.

[0077] The average optical flow velocity V is calculated as:

[0078]

[0079] The final urgency determination formula E is expressed as:

[0080] E=αV+βD+γ(A t+1 -A t )

[0081] Among them, α, β, and γ are weight coefficients, A t It is expressed as the diffusion area of ​​dust or gas leakage at time t, A t+1 It is expressed as the diffusion area of ​​dust or air leakage at time t+1; E is compared with the preset threshold to evaluate the urgency.

[0082] Further explanation: the preset threshold can be a single threshold to determine whether to issue an early warning, or it can be multiple thresholds, such as a low-level alarm threshold and a high-level alarm threshold. If the corresponding threshold is exceeded, an alarm will be issued. The specific number and value of the thresholds can be set and changed according to different environmental conditions.

[0083] This embodiment intercepts Figure 2 The dust and air leakage areas identified by the target are cut out 5 times in about 1 to 2 seconds (such as Figure 3-Figure 7 After dynamic analysis is performed to determine whether it is a dynamic scene and the motion trend, the speed and diffusion range are calculated, and an alarm is issued. As shown in the following scenario Table 1, we can determine that the current scene is experiencing dust emission and air leakage, but the motion situation is relatively low (the low-level alarm threshold set in this embodiment is 0.3), and an alarm can be issued according to the configured low level.

[0084] Table 1: Motion area ratio scenario table

[0085]

[0086]

[0087] Example 2 is the second embodiment of the present invention, which is different from the first embodiment in that it includes an image acquisition module, a first recognition module and a motion analysis module; the image acquisition module obtains a video stream based on the camera acquisition pipeline image; the first recognition module performs a first recognition on the video stream based on deep learning and a target recognition model of a neural network; the motion analysis module crops the picture of the first recognition of the existence of dust or air leakage, extracts key feature points in the image area of ​​the dust or air leakage part, calculates the motion vector of the feature point in the continuous frame image, analyzes the motion trend of dust or air leakage, and determines whether dust or air leakage exists. If dust or air leakage exists, the urgency is evaluated based on the movement speed, direction consistency and change in the area of ​​the diffusion area.

[0088] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0089] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0090] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0091] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, it can be implemented using a combination of any of the following technologies known in the art: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0092] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Industrial dust detection method based on target recognition, characterized by: include, Deploy cameras on industrial pipelines to capture pipeline images and generate video streams. This video stream is then first identified using deep learning and neural network-based object recognition models. The first identified image with dust or air leakage is cropped, key feature points are extracted from the image area of ​​the dust or air leakage portion, and the motion vectors of the feature points in the continuous frame images are calculated; Analyze the movement trend of dust emission or air leakage to determine whether dust emission or air leakage exists; If dust or air leakage exists, assess the urgency based on the speed of movement, directional consistency, and changes in the area of ​​the diffusion zone.

2. The industrial dust detection method based on target recognition according to claim 1, characterized in that: Deploying cameras on industrial pipelines includes installing high-definition industrial cameras at key areas of the pipelines, including elbows, valves, and joints, to cover all locations where dust and gas leaks may occur. A server is deployed in the camera to run target recognition and motion analysis programs in real time and connect to the network.

3. The industrial dust emission detection method based on target recognition according to claim 2, characterized in that: The deep learning and neural network-based target recognition model performs the first recognition of the video stream, including collecting historical image data of dust and gas leaks from industrial pipelines, covering different lighting conditions, pipeline materials, and background environments; The historical image data was annotated, including the location and category of dust and air leakage areas. The historical image data was expanded using data augmentation technology. The object recognition model was trained using the YOLOv5 framework and input into the annotated historical image data. The input image data is preprocessed, low-confidence false detection results are filtered according to the confidence threshold, and the model is trained.

4. The industrial dust emission detection method based on target recognition according to claim 3 is characterized in that: The target recognition model based on deep learning and neural networks performs the first recognition on the video stream, and also includes inputting the video stream into the trained target recognition model for the first recognition, inputting the picture with dust or air leakage identified into the motion analysis program, and performing motion analysis on the picture.

5. The industrial dust emission detection method based on target recognition according to claim 4 is characterized in that: The calculation of the motion vectors of the feature points in the continuous frame images includes: t and I t+1 , let the motion vector of the feature point p(x,y) be (u,v), and the optical flow constraint equation is expressed as, I x u+I y υ+I t =0 Among them, I x , I y It is represented as image gradient, u is the direction of motion vector, and v is the speed of motion vector; Select a window W around the feature point and minimize the error expressed as, By least squares solution: After detecting multiple areas of dust or air leakage, calculate the optical flow vectors (u i ,υ i ).

6. The industrial dust emission detection method based on target recognition according to claim 5, characterized in that: The determination of whether dust emission or air leakage occurs includes calculating the average movement direction of the feature points. and average speed Expressed as, If the motion direction in adjacent frames is consistent, it indicates that there is motion in the abnormal area rather than static interference; The diffusion degree D of the feature point is calculated as follows: Where N represents the total number of feature points. If D exceeds the diffusion threshold, it means that the abnormal area is spreading, which is consistent with the characteristics of air leakage or dust emission.

7. The industrial dust emission detection method based on target recognition according to claim 6, characterized in that: The evaluation of the urgency level includes calculating the average optical flow velocity V as: The final urgency determination formula E is expressed as: E=αV+βD+γ(A t+1 -A t ) Among them, α, β, and γ are weight coefficients, A t It is expressed as the diffusion area of ​​dust or gas leakage at time t, A t+1 Expressed as the dust or gas leakage diffusion area at time t+1; Compare E with the preset threshold to assess the urgency.

8. An industrial dust emission detection system based on target recognition, applying the industrial dust emission detection method based on target recognition according to any one of claims 1 to 7, characterized in that: It includes image acquisition module, first recognition module and motion analysis module; The image acquisition module acquires a video stream based on the camera acquisition pipeline image; The first recognition module performs a first recognition on the video stream based on a target recognition model of deep learning and a neural network; The motion analysis module crops the first image identified as having dust or air leakage, extracts key feature points in the image area of ​​the dust or air leakage part, calculates the motion vector of the feature points in the continuous frame image, analyzes the movement trend of the dust or air leakage, and determines whether the dust or air leakage phenomenon exists. If the dust or air leakage phenomenon exists, the urgency is evaluated based on the movement speed, direction consistency and change in the area of ​​the diffusion area.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the industrial dust emission detection method based on target recognition according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the industrial dusting detection method based on target recognition according to any one of claims 1 to 7 are implemented.

Images