Method for detecting solid content image of gas-liquid-solid dilute phase annular flow in vertical pipeline

Abstract

The invention relates to the technical field of multiphase flow parameter detection, and particularly discloses a vertical pipeline gas-liquid-solid dilute phase annular flow solid content image detection method, which comprises the following steps: S1, carrying out multi-view image acquisition; s2, preprocessing the acquired image; s21, dynamic background differencing is carried out, and gas-liquid interface reflection noise is removed; s22, non-local mean filtering denoising is adopted, noise is removed, and detail information of the image is reserved; s23, carrying out local Otsu segmentation, and segmenting the particles from the background; s3, analyzing the preprocessed image; s31, performing spatial correlation verification, and eliminating multi-view repeated particles; s32, time correlation analysis is carried out, and a displacement vector of the particles is calculated; s33, correcting a displacement error by using an optical flow correction algorithm; and S4, carrying out particle data statistics, and calculating the solid content. According to the detection method, dynamic interference can be effectively resisted, a single-visual-angle blind area is eliminated, repeated particle counting is accurately eliminated, and the solid rate measurement precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of multiphase flow parameter detection, and specifically provides an image detection method for solid holdup in vertical pipeline gas-liquid-solid dilute-phase annular flow. Background Art

[0002] Vertical pipeline gas-liquid-solid dilute-phase annular flow is a complex multiphase flow state. In this flow state, gas mainly forms a continuous gas core in the center of the pipeline, liquid forms a liquid film flow on the pipeline wall, and at the same time, part of the liquid in the liquid film is entrained into the gas core by the shearing action of the gas flow, forming liquid droplets and moving with the gas flow. Solid particles are sparsely distributed in the gas-liquid two-phase flow and there are complex interactions with the gas and liquid phases.

[0003] In the actual application scenarios of vertical pipeline gas-liquid-solid dilute-phase annular flow, such as oil well bores in oil extraction, reaction pipelines in chemical production, and transportation pipelines in the energy industry, the accurate measurement of solid holdup plays a crucial role. Solid holdup is directly related to the optimization of the process flow. For example: 1. In chemical production, an appropriate solid holdup can ensure the full progress of the reaction, improve product quality and production efficiency; 2. For the safe operation of equipment, accurate solid holdup data helps prevent problems such as pipeline blockage and equipment wear, and ensure the long-term stable operation of the equipment; 3. In terms of environmental protection control, accurate mastery of solid holdup can effectively control pollutant emissions and meet the increasingly strict environmental protection regulations.

[0004] However, traditional detection methods, such as conductivity probes and gamma densitometers, have many defects. First, as a contact detection means, conductivity probes are prone to corrosion and wear, which affect the measurement accuracy and service life. Moreover, based on the principle of conductivity change, for complex gas-liquid-solid multiphase systems, the measurement results are easily interfered. Second, although gamma densitometers can achieve non-contact measurement, there are radiation safety hazards, high equipment costs, complex maintenance, and it is also difficult to meet the requirements of modern industrial production in terms of measurement accuracy and real-time performance.

[0005] Specifically, the traditional methods have the following problems: 1. There is a problem of single - perspective limitation: A sensor or camera device with a fixed perspective can only obtain information about a partial cross - section of the pipeline and cannot cover the entire cross - section, which will lead to a large number of particles being missed. For example, in some corner or edge areas of the pipeline cross - section, due to the perspective limitation, it is difficult for the sensor to detect the presence of particles, resulting in a large deviation between the measurement result and the actual solid - content ratio. 2. There is a problem of dynamic interference: In the case of high - speed flow, particles overlap with each other, move at high speed, and are prone to motion blur. At the same time, the reflection noise at the gas - liquid interface will also seriously interfere with the detection signal. Taking an oil - transportation pipeline as an example, when the three - phase of oil, gas, and solid flows at high speed, the overlap and motion blur of particles make it difficult for the sensor to accurately identify and count particles, and the reflected light at the gas - liquid interface will generate false signals, thus greatly reducing the detection accuracy. 3. There is a problem of repeated counting: When particles pass through the detection area multiple times, traditional detection methods usually cannot effectively distinguish them, resulting in repeated statistics and significant data deviation. In some circulating transportation systems, particles may circulate through the detection point multiple times in the pipeline. If the detection method cannot accurately eliminate repeated counting, it will seriously affect the accuracy of solid - content ratio measurement.

[0006] In summary, it is an urgent problem to provide a detection method for the solid - content ratio of gas - liquid - solid dilute - phase annular flow in vertical pipelines that can cover the entire cross - section, resist dynamic interference, and effectively exclude repeated particles. Summary of the Invention

[0007] To solve the above problems, the present invention provides an image - based detection method for the solid - content ratio of gas - liquid - solid dilute - phase annular flow in vertical pipelines.

[0008] An image - based detection method for the solid - content ratio of gas - liquid - solid dilute - phase annular flow in vertical pipelines provided by the present invention is detected through an image detection system, and specifically includes the following steps: S1. Perform multi - perspective image acquisition; S2. Pre - process the acquired images; S21. Perform dynamic background difference to remove the reflection noise at the gas - liquid interface; S22. Use non - local means filtering for denoising to remove noise and retain the detailed information of the image; S23. Perform local Otsu segmentation to separate particles from the background; S3. Analyze the pre - processed images; S31. Perform spatial correlation verification: Perform polar - coordinate transformation and matching to exclude multi - perspective repeated particles; S32. Perform temporal correlation analysis: Use particle image velocimetry displacement tracking technology to calculate the displacement vector of particles; S33. Correct the displacement error: Use the optical flow correction algorithm to correct the displacement error and ensure the accuracy of the particle movement trajectory. S4. Conduct particle data statistics and calculate the solids content; update the solids content and particle distribution map every second.

[0009] Furthermore, the multi-view image acquisition in step S1 includes the following sub-steps: S11. Initialize the rotating platform, calibrate the camera and the laser. The laser hits the vertical pipe to form a laser plane, and the camera is perpendicular to the laser plane. The rotating platform drives the camera and the laser to rotate while the vertical pipe remains stationary. The camera faces the laser plane for shooting; the camera and the laser rotate synchronously around the vertical pipe. Set the angular interval of the rotation of the camera and the laser to ≤5°, that is, take an image every time the rotation does not exceed 5°. S12. Configure the trigger system and start the synchronous trigger of the laser and the camera; specifically, turn on the laser, adjust its wavelength to 532 nm, and control the pulse width to be below 10 ns; start the camera, set the resolution to 1280×1024, and adjust the frame rate to above 5000 fps. S13: Use a rotary encoder to record the angular position. Each time the camera rotates, record the current rotation angle and ensure that the angle of image acquisition matches the encoder data. S14: Store the image sequence and cover the entire cross-section of the pipe; for each captured image, save the image and its angle information. The number of images captured in a complete cycle is 72. Ensure that the positions at each angle are covered within a complete cycle to form the image data of the entire cross-section of the pipe.

[0010] Furthermore, in step S21, for dynamic background difference to remove the reflection noise at the gas-liquid interface, specifically: Extract moving objects from the video or image sequence through dynamic background difference and remove the background noise. Collect background images in the particle-free state at intervals of Δt. , during the actual detection process, by calculating the current frame image and the background image in the particle-free state to remove the static part in the background and the noise generated by the reflection at the gas-liquid interface, that is ; where represents the background image in the particle-free state; represents the current frame image, including the original data of particles, liquid, gas, and noise; represents the difference image, indicating the difference part between the current frame and the background; Δt represents time.

[0011] Further, when non-local means filtering is used for denoising in S22, non-local means filtering is performed on each pixel to find similar pixels in its neighborhood and perform weighted averaging. By adjusting the filtering parameters, the details and edges in the image are retained while the noise is removed; In S23, when performing local Otsu segmentation, the Otsu algorithm is used to calculate the gray histogram of the image and find the optimal threshold; the area in the image smaller than the threshold is classified as the background, and the area larger than the threshold is classified as particles; the segmentation result is used to extract the particle area from the background.

[0012] Further, the specific steps for verifying spatial correlation in step S31 are as follows: S311. Perform polar coordinate transformation and matching. Convert the positions of the particles in the image from the Cartesian coordinate system to the polar coordinate system, and use the polar coordinate formula to calculate the position of each particle; S312. Perform normalized cross-correlation matching to verify spatial consistency. For the image region that has been converted to polar coordinates, extract the local gray-scale image of the particles, and calculate the normalized cross-correlation coefficient 𝑅. The calculation formula is as follows:

[0013] where 𝑅 represents the normalized cross-correlation coefficient, and its value range is [-1, 1]. The larger the value, the higher the similarity; 、 represents the gray-scale value of the adjacent view image in the polar coordinate system; 𝑟 represents the radial distance, and 𝜃 represents the polar angle; if 𝑅 > 0.85, it is determined to be the same particle, and only the unique coordinates are retained; if 𝑅 ≤ 0.85, it is regarded as a different particle and is retained separately; in this way, the situation where the same particle is repeatedly detected due to the view change under multiple views is excluded.

[0014] Further, step S311 is specifically: match the images collected from multiple views, exclude duplicate particles, and ensure the uniqueness of each particle under multiple views; extract the two-dimensional Cartesian coordinate points (x, y) of the particles from two adjacent images and convert them into polar coordinate points (r, 𝜃), and unify them to the same reference view; specifically, through the formula , convert the Cartesian coordinate point (x, y) in the Cartesian coordinate system into the polar coordinate point (r, 𝜃) in the polar coordinate system, where 𝑟 is the radial distance, 𝜃 is the polar angle, and x and y are the coordinates of the particle in the Cartesian coordinate system.

[0015] Further, S32. Perform time correlation analysis as follows: For the particles in the image sequence, based on the maximum cross-correlation matching between consecutive frames, determine the displacement vector of the particles in time, and calculate the displacement vector of the particles according to the formula ; where, 、 represent two consecutive frames of images, represents the horizontal component of the particle displacement vector, represents the vertical component of the particle displacement vector, and y represent the position coordinates of the particle in the rectangular coordinate system, where is the coordinate value in the horizontal direction and y is the coordinate value in the vertical direction.

[0016] Further, in S33, an optical flow correction algorithm is used to correct the displacement error, including the following sub-steps: S331. Perform optical flow correction on the particle motion trajectory and calculate the optical flow vector of each pixel; S332. Correct the displacement error caused by camera movement or other factors to ensure the accuracy of particle motion; S333. Correct the displacement error of the particle motion trajectory according to the optical flow correction equation where represents the spatial gradient change in the horizontal direction in the image rectangular coordinate system, represents the spatial gradient change in the vertical direction in the image rectangular coordinate system, where is the coordinate value in the horizontal direction and y is the coordinate value in the vertical direction, represents the change in time, u represents the component of the optical flow in the direction, and v represents the component of the optical flow in the y direction.

[0017] Further, step S4 includes the following sub-steps: S41. Conduct particle data statistics: Obtain the data after averaging the multiple rotation statistics according to the formula where is the total effective number of particles obtained by statistics for each rotation period, is the number of rotation periods; Calculate through the formula where is the effective number of particles obtained after duplicate removal in each frame of the image, is the total number of pictures taken within a rotation period; S42: Conduct particle volume calculation: Obtain the average volume of a single particle according to the formula where r is the particle radius; S43: Conduct pipe cross-sectional area calculation: Obtain the cross-sectional area of the pipe according to the formula where D is the inner diameter of the pipe; S44: Calculate the solids content S: The calculation formula for the solids content S is 固 where N is the total number of particles after averaging the multiple rotation statistics, A is the cross-sectional area of the pipe, L is the length of the detection section, and V 固 is the average volume of a single particle.

[0018] Compared with the prior art, the present invention can achieve the following beneficial effects: The detection method in the present invention is applicable to non-contact real-time monitoring of high-dynamic multiphase flows in the petroleum, chemical, and energy industries. By rotating a laser high-speed camera 360° to take pictures, image processing is performed on the obtained pictures. The spatial-temporal correlation of solid particles in two adjacent pictures is verified to exclude duplicate particles, which can completely eliminate the single-view blind area, accurately identify and exclude duplicate particles between images from different perspectives, effectively avoid the problem of multiple counting of the same particle due to perspective changes, and improve the accuracy of particle counting. At the same time, through displacement tracking technology, the movement trajectory of particles in consecutive frame images can be analyzed, avoiding duplicate counting of the same particle in the time domain, and further improving the accuracy of solids content measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the overall flowchart of the image detection method provided by an embodiment of the present invention; Figure 2 is the background image in the particle-free state provided by an embodiment of the present invention; Figure 3 is the current frame image provided by an embodiment of the present invention; Figure 4 is the image after dynamic background difference processing provided by an embodiment of the present invention; Figure 5 is the denoised image provided by an embodiment of the present invention; Figure 6 is the image after local Otsu segmentation provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] In order to make the objectives, technical solutions, and advantages of the present invention clearer and more understandable, the following further details the present invention in conjunction with the Figure 1-6 accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation to the present invention.

[0021] An image detection method for solid holdup in vertical pipeline gas-liquid-solid dilute-phase annular flow is detected by an image detection system. Among them, the image detection system includes a laser high-speed camera module, and image acquisition is carried out through the laser high-speed camera module. The laser high-speed camera module includes a laser and a high-speed camera. The laser is a pulsed laser, the peak power of the laser is 1.2 kW, the pulse frequency is 1 kHz, the wavelength of the laser is 532 nm, and the pulse width ≤ 10 ns. The resolution of the high-speed camera ≥ 1280×1024, and the frame rate ≥ 5000 fps. The high peak power can instantaneously provide sufficient light intensity to illuminate the flow field in the pipeline, and the pulse frequency ensures that stable light can be continuously provided during high-speed photography, meeting the shooting requirements for high-speed dynamic multiphase flow.

[0022] The image detection system also includes a rotating platform. The laser high-speed camera module is installed on the rotating platform, and the rotating platform drives the camera in the laser high-speed camera module to continuously shoot 360° around the axis of the vertical pipeline, and an image is obtained every Δθ°. The image detection system also includes a synchronous control unit and a data processing unit. The synchronous control unit includes a rotary encoder. The data processing unit is based on FPGA hardware and is used to perform image preprocessing, spatio-temporal correlation fusion algorithm and real-time calculation of solid holdup.

[0023] This invention is experimentally verified based on a DN80 vertical pipeline. The experimental medium is selected as air-water-glass microspheres. The particle size of the glass microspheres is 80-150 μm, the density is 2500 kg / m³, the flow rate is set to 2 m / s, and the solid holdup range is 0.5%-3%. As Figure 1 shown, the image detection method for solid holdup in vertical pipeline gas-liquid-solid dilute-phase annular flow specifically includes the following steps: S1. Perform multi-view image acquisition. The multi-view image acquisition in step S1 includes the following sub-steps: S11. Initialize the rotating platform, calibrate the camera and the laser. The laser hits the vertical pipeline and forms a laser plane. The camera is perpendicular to the laser plane. The rotating platform drives the camera and the laser to rotate, and the vertical pipeline remains stationary. The camera faces the laser plane for shooting. A laser alignment instrument can be used for auxiliary calibration, and a scale or a laser alignment device is used to calibrate the alignment of the camera with the axis of the vertical pipeline to ensure that the error is within an acceptable range.

[0024] The camera and the laser rotate synchronously, in the same direction and at the same angular velocity around the vertical pipeline, and the vertical pipeline always remains stationary. The camera takes a photo every time it rotates Δθ°. The angular interval between adjacent shootings of the camera is Δθ°. The angular interval between the rotations of the camera and the laser is set to ≤ 5°, that is, an image is taken every time it rotates no more than 5°. In this way, the entire cross-section of the pipeline is fully covered to obtain rich image information. In this embodiment, the angular interval between rotations is 5°.

[0025] S12. Configure the triggering system to ensure synchronous triggering of the laser and the high-speed camera. Start the synchronous triggering of the laser and the camera. Specifically, turn on the laser, adjust its wavelength to 532 nm, and control the pulse width to be below 10 ns. Start the high-speed camera, set the resolution to 1280×1024, and adjust the frame rate to above 5000 fps.

[0026] S13: Use a rotary encoder to record the angular position. Each time the camera rotates, record the current rotation angle and ensure that the angle of image acquisition matches the encoder data. The accuracy of the integrated encoder is ±0.05°, which is used to record the angular position of the camera in real time and achieve spatial alignment of multi-view images. The positioning accuracy reaches ±0.1°. It is driven by a stepper motor, and the rotation speed is set to 0.5 r / s. This speed can not only ensure that the camera has enough time to obtain clear images but also complete 360° rotational shooting within a reasonable time. The rotary encoder can accurately measure the rotation angle of the rotating platform, providing an accurate angle basis for the subsequent spatial alignment of multi-view images, ensuring that the images taken from different perspectives can be accurately matched in space, and laying a foundation for subsequent data processing and analysis.

[0027] S14: Store the image sequence and cover the entire cross-section of the pipeline; for each captured image, save the image and its angle information. The number of images captured in a complete cycle is 72. Ensure that the positions at each angle are covered within a complete cycle to form image data of the entire cross-section of the pipeline.

[0028] S2. Preprocess the captured images.

[0029] S21. Perform dynamic background difference to remove the reflection noise at the gas-liquid interface. Specifically: Extract moving objects from the video or image sequence and remove background noise through dynamic background difference. The moving objects are changes such as the gas-liquid interface, and the background noise is the reflection noise at the gas-liquid interface.

[0030] To eliminate the influence of the reflection noise at the gas-liquid interface on particle detection, capture background images in the particle-free state at intervals of Δt. During the actual detection process, by calculating the difference between the current frame image and the background image in the particle-free state , remove the static part in the background and the noise generated by the reflection at the gas-liquid interface, highlighting the image features of the particles, which is convenient for subsequent extraction and positioning of the particles, that is ; where represents the background image in the particle-free state; represents the current frame image, including the original data of particles, liquid, gas, and noise; The representative differential image represents the different part between the current frame and the background; Δt represents time, for example, Δt can be set to 10.

[0031] The background image in the particle-free state is as Figure 2 shown, and the current frame image is as Figure 3 shown, and the image after dynamic background differential processing is as Figure 4 shown.

[0032] S22. Use non-local means filtering (NLM) for denoising to remove noise and retain the detailed information of the image. When using non-local means filtering for denoising, perform non-local means filtering on each pixel, find similar pixels in its neighborhood and perform weighted averaging. By adjusting the filtering parameters, retain the details and edges in the image and remove the noise. The denoised image is as Figure 5 shown.

[0033] S23. Perform local Otsu segmentation to segment the particles from the background. When performing local Otsu segmentation, use the Otsu algorithm to calculate the gray histogram of the image and find the optimal threshold; divide the area in the image smaller than the threshold into the background, and divide the area larger than the threshold into particles; use the segmentation result to extract the particle area from the background. The image after local Otsu segmentation is as Figure 6 shown.

[0034] Here, the local Otsu algorithm is used to perform binary processing on the differential image for adaptive segmentation. The local Otsu algorithm can adaptively determine the binary threshold according to the gray distribution characteristics of the local area of the image. Compared with the global threshold algorithm, it can better adapt to the complex illumination conditions and flow field characteristics in the pipeline. Through this method, the particle area can be accurately segmented from the background, and the extracted particle area is a coordinate set, thus providing basic data for subsequent particle analysis.

[0035] S3. Analyze the preprocessed image.

[0036] S31. Perform spatial correlation verification: perform polar coordinate transformation and matching to exclude multi-view duplicate particles. Specifically, it includes the following sub-steps: S311. Perform polar coordinate transformation and matching, convert the particle positions in the image from the rectangular coordinate system to the polar coordinate system, and calculate the position of each particle using the polar coordinate formula.

[0037] Convert the particle coordinates of two adjacent images to the same polar coordinate system. The angular difference between two adjacent images is Δθ°. In the polar coordinate system, the position of a particle is represented by the polar radius r and the polar angle θ. Step S311 is specifically as follows: Match the images collected from multiple perspectives, exclude duplicate particles, and ensure the uniqueness of each particle in multiple perspectives; Extract the two-dimensional rectangular coordinate points (x, y) of the particles from two adjacent images and convert them into polar coordinate points (r, θ), and unify them to the same reference perspective; Specifically, through the formula , convert the rectangular coordinate point (x, y) in the rectangular coordinate system into the polar coordinate point (r, θ) in the polar coordinate system, where r is the polar radius, θ is the polar angle, and x and y are the coordinates of the particle in the rectangular coordinate system.

[0038] S312. Normalized cross-correlation (NCC) matching is used to verify the spatial consistency. For the image region that has been converted to polar coordinates, extract the local grayscale image of the particle, and calculate the normalized cross-correlation (NCC) coefficient 𝑅. The calculation formula is as follows:

[0039] Among them, R represents the normalized cross-correlation coefficient, and the value range is [-1, 1]. The larger the value, the higher the similarity; , represents the grayscale value of the adjacent perspective images in the polar coordinate system; r represents the polar radius, and θ represents the polar angle; If R > 0.85, it is determined as the same particle, and only the unique coordinates are retained; If R ≤ 0.85, it is regarded as different particles and are retained separately; In this way, the situation that the same particle is repeatedly detected due to the perspective change under multiple perspectives is effectively excluded, greatly improving the accuracy of particle counting.

[0040] S32. Perform time correlation analysis: Adopt particle image velocimetry displacement tracking technology to calculate the displacement vector of the particle. The specific time correlation analysis is as follows: For the particles in the image sequence, based on the maximum cross-correlation matching between consecutive frames, determine the displacement vector of the particle in time, and calculate the displacement vector of the particle according to the formula Calculate the displacement vector of the particle; Among them, , represent two consecutive frames of images, represents the horizontal component of the particle displacement vector, represents the vertical component of the particle displacement vector, and y represent the position coordinates of the particle in the rectangular coordinate system, is the coordinate value in the horizontal direction, and y is the coordinate value in the vertical direction. Through this calculation, the displacement information of the particle between consecutive frames of images can be obtained, so as to realize the tracking of the particle motion trajectory.

[0041] S33. Correct the displacement error: Use the optical flow correction algorithm to correct the displacement error and ensure the accuracy of the particle motion trajectory. It includes the following sub-steps: S331. Perform optical flow correction on the particle motion trajectory and calculate the optical flow vector of each pixel.

[0042] S332. Correct the displacement error caused by camera movement or other factors to ensure the accuracy of particle motion.

[0043] S333. According to the optical flow correction equation correct the displacement error of the particle motion trajectory, where represents the spatial gradient change in the horizontal direction in the image rectangular coordinate system, represents the spatial gradient change in the vertical direction in the image rectangular coordinate system, x is the coordinate value in the horizontal direction, and y is the coordinate value in the vertical direction. represents the change over time, that is, the change rate of the image over time, which is the brightness difference of the same pixel position between two adjacent frames.

[0044] Based on the optical flow equation, correct the displacement error of the dynamic blur area. By solving the optical flow equation, the displacement of the dynamic blur area caused by the high-speed movement of particles can be accurately corrected, and abnormal trajectories can be eliminated. For example, when the particle movement speed is too fast and causes image blur, the actual displacement of the particle can be more accurately determined through optical flow correction, improving the accuracy of dynamic tracking and further avoiding duplicate counting of the same particle in the time domain.

[0045] S4. Conduct particle data statistics and calculate the solids content; update the solids content and particle distribution map every second. Specifically, it includes the following sub-steps: S41. Conduct particle data statistics: According to the formula obtain the data after averaging multiple rotation statistics, where is the total effective number of particles obtained from each rotation period statistics, is the number of rotation periods. The number of particles counted after the high-speed camera rotates around the vertical pipe for one week is N1, which is the relatively accurate total number of particles in the vertical pipe. However, multiple sets of data such as N1, N2, and N3 still need to be obtained through multiple experiments and averaged to get N.

[0046] Fuse multi-view and multi-frame data, and conduct total statistics on the effective particles obtained after spatial correlation verification and time correlation analysis. By comprehensively integrating the image information from different perspectives and different moments, the accuracy and integrity of particle counting can be further ensured.

[0047] Calculate through the formula where is the number of effective particles obtained after duplicate removal in each frame of the image. is the total number of pictures taken in one rotation period.

[0048] S42: Perform particle volume calculation: According to the formula obtain the average volume of a single particle, where r is the particle radius.

[0049] S43: Perform pipe cross-sectional area calculation: According to the formula obtain the cross-sectional area of the pipe, where D is the inner diameter of the pipe.

[0050] S44: Calculate the solids content S: The calculation formula for the solids content S is , where N is the total number of particles after averaging by multiple rotation statistics, A is the cross-sectional area of the pipe, L is the length of the detection section, and V 固 is the average volume of a single particle.

[0051] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. An image detection method for solid holdup in vertical pipe gas-liquid-solid dilute phase annular flow, characterized in that, Detection is performed by an image detection system, specifically including the following steps: S1. Perform multi-view image acquisition; S2. Preprocess the acquired images; S21. Perform dynamic background difference to remove the reflection noise at the gas-liquid interface; S22. Use non-local mean filtering for denoising to remove noise and retain the detailed information of the image; S23. Perform local Otsu segmentation to segment the particles from the background; S3. Analyze the preprocessed images; S31. Perform spatial correlation verification: perform polar coordinate transformation and matching to exclude multi-view duplicate particles; S32. Perform temporal correlation analysis: use particle image velocimetry displacement tracking technology to calculate the displacement vector of the particles; S33. Correct the displacement error: use an optical flow correction algorithm to correct the displacement error to ensure the accuracy of the particle motion trajectory; S4. Perform particle data statistics and calculate the solid content; update the solid content and particle distribution map every second.

2. The solid holdup image detection method for vertical pipeline gas-liquid-solid dilute phase annular flow according to claim 1, characterized in that The multi-view image acquisition in step S1 includes the following sub-steps: S11. Initialize the rotating platform, calibrate the camera and the laser. The laser is projected onto the vertical pipeline to form a laser plane. The camera is perpendicular to the laser plane. The rotating platform drives the camera and the laser to rotate, while the vertical pipeline remains stationary. The camera faces the laser plane for shooting; the camera and the laser rotate synchronously around the vertical pipeline. Set the rotation angle interval of the camera and the laser to ≤5°, that is, take an image every rotation of no more than 5°; S12. Configure the trigger system and start the synchronous trigger of the laser and the camera; specifically, turn on the laser, adjust its wavelength to 532 nm, and control the pulse width to be below 10 ns; start the camera, set the resolution to 1280×1024, and adjust the frame rate to above 5000 fps; S13: Use a rotary encoder to record the angular position. Each time the camera rotates, record the current rotation angle and ensure that the image acquisition angle matches the encoder data; S14: Store the image sequence and cover the entire cross-section of the pipeline; for each captured image, save the image and its angle information. The number of images captured in a complete cycle is 72. Ensure that the position of each angle is covered within a complete cycle to form the image data of the entire cross-section of the pipeline.

3. The method for detecting the solid holdup image of the gas-liquid-solid dilute-phase annular flow in a vertical pipe according to claim 1, characterized in that In step S21, when performing dynamic background difference to remove the reflection noise at the gas-liquid interface, specifically: extract moving objects from the video or image sequence through dynamic background difference and remove the background noise; Collect background images in a particle-free state at intervals of Δt , during the actual detection process, by calculating the current frame image and the background image in a particle-free state to remove the static part in the background and the noise generated by the reflection of the gas-liquid interface, that is ; where represents the background image in a particle-free state; represents the current frame image, containing the original data of particles, liquid, gas and noise; represents the difference image, indicating the difference part between the current frame and the background; Δt represents time.

4. The image detection method for solid holdup in vertical pipe gas-liquid-solid dilute-phase annular flow according to claim 3, wherein When using non-local mean filtering for denoising in S22, perform non-local mean filtering on each pixel, find similar pixels in its neighborhood and perform weighted averaging. By adjusting the filtering parameters, retain the details and edges in the image and remove the noise; When performing local Otsu segmentation in S23, use the Otsu algorithm to calculate the grayscale histogram of the image and find the optimal threshold; divide the area in the image smaller than the threshold into the background, and divide the area larger than the threshold into particles; use the segmentation result to extract the particle area from the background.

5. The method for detecting the solid holdup image of the gas-liquid-solid dilute-phase annular flow in a vertical pipe according to claim 1, wherein, The spatial correlation verification in step S31 specifically includes the following sub-steps: S311. Perform polar coordinate transformation and matching, convert the particle positions in the image from the rectangular coordinate system to the polar coordinate system, and calculate the positions of each particle using the polar coordinate formula; S312. Normalized cross-correlation matching to verify spatial consistency. For the image region that has been converted to polar coordinates, extract the local grayscale image of the particles, and calculate the normalized cross-correlation coefficient 𝑅. The calculation formula is as follows: ； Among them, R represents the normalized cross-correlation coefficient, and its value range is [-1, 1]. The larger the value, the higher the similarity. , represents the gray value of adjacent perspective images in the polar coordinate system; r represents the polar radius, and θ represents the polar angle. If R > 0.85, it is determined to be the same particle, and only the unique coordinates are retained; if R ≤ 0.85, it is regarded as different particles and retained separately; in this way, the situation where the same particle is repeatedly detected due to perspective changes under multiple perspectives is excluded.

6. The method for detecting the solid holdup image of the gas-liquid-solid dilute-phase annular flow in a vertical pipe according to claim 5, wherein Step S311 specifically includes: matching the images collected from multiple perspectives, excluding duplicate particles, and ensuring the uniqueness of each particle under multiple perspectives; extracting the two-dimensional rectangular coordinate points (x, y) of the particles from two adjacent images and converting them into polar coordinate points (r, θ), and unifying them to the same reference perspective; specifically through the formula , convert the rectangular coordinate point (x, y) in the rectangular coordinate system into the polar coordinate point (r, θ) in the polar coordinate system, where r is the polar radius, θ is the polar angle, and x and y are the coordinates of the particle in the rectangular coordinate system.

7. The method for detecting the solid holdup image of gas-liquid-solid dilute-phase annular flow in a vertical pipe according to claim 1, wherein S32. The time correlation analysis is carried out as follows: For the particles in the image sequence, based on the maximum cross-correlation matching between consecutive frames, the displacement vector of the particles in time is determined, and the displacement vector of the particles is calculated according to the formula ; where and represent two consecutive frames of images, represents the horizontal component of the particle displacement vector, represents the vertical component of the particle displacement vector, and y represent the position coordinates of the particle in the rectangular coordinate system, is the coordinate value in the horizontal direction, and y is the coordinate value in the vertical direction.

8. The method for detecting the solid holdup image of the gas-liquid-solid dilute-phase annular flow in a vertical pipe according to claim 1, wherein, In S33, an optical flow correction algorithm is used to correct the displacement error, including the following sub-steps: S331. Perform optical flow correction on the motion trajectories of the particles and calculate the optical flow vectors of each pixel; S332. Correct the displacement error caused by camera movement or other factors to ensure the accuracy of particle movement; S333. Correct the displacement error of the particle motion trajectory according to the optical flow correction equation where represents the spatial gradient change in the horizontal direction in the image rectangular coordinate system represents the spatial gradient change in the vertical direction in the image rectangular coordinate system x is the coordinate value in the horizontal direction, and y is the coordinate value in the vertical direction represents the change in time, u represents the component of the optical flow in the x direction, and v represents the component of the optical flow in the y direction 9. The method for detecting the solid holdup image of the vertical pipe gas-liquid-solid dilute-phase annular flow according to claim 1, wherein Step S4 includes the following sub-steps: S41. Perform particle data statistics: According to the formula Obtain the data after averaging the statistics of multiple rotations, where is the total effective number of particles obtained by statistics for each rotation period, is the number of rotation periods; Calculate through the formula where is the number of effective particles obtained after duplicate removal in each frame of the image, is the total number of pictures taken in one rotation period; S42: Perform particle volume calculation: According to the formula obtain the average volume of a single particle, where r is the particle radius; S43: Perform the calculation of the cross-sectional area of the pipeline: According to the formula obtain the cross-sectional area of the pipeline, where D is the inner diameter of the pipeline; S44: Calculate the solids content S: The calculation formula for the solids content S is , where N is the total number of particles after averaging through multiple rotation statistics, A is the cross-sectional area of the pipeline, L is the length of the detection section, and V 固 is the average volume of a single particle.

Images