Oil and gas cylinder drainage method and system based on machine vision

Abstract

The present invention relates to a method and system for draining liquid from an oil and gas cylinder based on machine vision. The method comprises: constructing a training sample set, the training sample set including a number of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, then constructing an oil-water discrimination model, and using the training sample set to train the oil-water discrimination model; providing an optical window on the side wall of the liquid collecting chamber of the oil and gas cylinder, and obtaining an image of the liquid in the liquid collecting chamber through the optical window; performing image preprocessing on the liquid image to extract image features; inputting the image features into the trained oil-water discrimination model and outputting a discrimination result; and controlling the liquid discharge from the liquid collecting chamber based on the discrimination result. The present invention realizes non-contact oil-water emulsion discrimination with high discrimination accuracy and strong real-time performance, which is conducive to accurately controlling the liquid discharge in the liquid collecting chamber and solving the problem of lubricating oil waste.

Description

Technical Field

[0001] The present invention relates to the technical field of air compressors, and in particular to a method and system for draining oil and gas cylinders based on machine vision. Background Art

[0002] In industrial power systems, air compressors serve as core air supply equipment, and their stable operation is directly related to production efficiency and equipment safety. Lubricating oil in air compressors performs functions such as lubrication, sealing, cooling, and rust prevention. However, during operation, lubricating oil frequently comes into contact with water, resulting in increasingly prominent oil-water emulsification, causing equipment wear, increased energy consumption, and even breakdowns and shutdowns. During the operation of the air compressor, the moisture that the lubricating oil comes into contact with mainly comes from the ambient air and phase changes during the compression process. For example, the ambient air inhaled by the air compressor contains a large amount of water vapor, the air volume shrinks during the compression process, and the water vapor saturates and condenses into liquid water. In addition, the high-temperature compressed air enters the oil and gas cylinder or aftercooler for cooling, and the temperature drops below the ambient dew point, further condensing and precipitating free water. This process is exacerbated in low-temperature environments or high-humidity conditions.

[0003] After condensed water mixes with lubricating oil, it forms a stable oil-water emulsion (milky white or yellow-brown viscous) under the action of mechanical shear force (such as rotor stirring). This oil-water emulsion has the following hazards:

[0004] 1) Decreased lubrication performance: The viscosity of the emulsified oil decreases and the oil film strength is insufficient, resulting in increased wear of bearings, rotors and other components;

[0005] 2) Corrosion risk: Moisture causes metal parts (bearings, pipes) to rust; if oil oxidation products (acidic substances) coexist with water, they will form a more corrosive acidic emulsion;

[0006] 3) System blockage: The emulsion has poor fluidity and is easy to clog the oil filter and oil separator; water evaporation at high temperature may leave carbon deposits, aggravating equipment aging.

[0007] Currently, the determination of emulsified air compressor lubricant oil relies primarily on manual experience, such as visual observation and touch. However, since water is dispersed in the oil as micron-sized droplets during the initial emulsification phase, misjudgment is high. Furthermore, oil oxidation, discoloration, and impurities can interfere with judgment, leading to the misinterpretation of yellowish-white emulsified oil as simply aged. Furthermore, manual touch requires a significant decrease in viscosity or increased friction to detect the condition, at which point the equipment's bearings have already entered a period of accelerated wear. Furthermore, periodic oil sampling is sometimes performed and sent to specialized laboratories for analysis (such as capacitance and dew point methods). While this method can accurately quantify water content, the inspection cycle can take up to 24 to 72 hours, making real-time early warning impossible.

[0008] In addition, the current drainage method used for the liquid collecting chamber in the oil and gas cylinder of the air compressor is manual drainage or timed discharge by the solenoid valve. Due to the misjudgment or lag of manual detection, there is a large waste of lubricating oil. Summary of the Invention

[0009] The purpose of the present invention is to provide an oil and gas cylinder drainage method and system based on machine vision, which can realize non-contact oil-water emulsification judgment with high judgment accuracy and strong real-time performance, which is conducive to accurately controlling the liquid discharge in the liquid collecting chamber and solving the problem of lubricating oil waste.

[0010] In order to achieve the above object, the present invention provides a method for draining an oil and gas cylinder based on machine vision, comprising:

[0011] Constructing a training sample set, the training sample set including a plurality of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, then constructing an oil-water discrimination model, and using the training sample set to train the oil-water discrimination model;

[0012] An optical window is provided on the side wall of the liquid collecting chamber of the oil and gas cylinder, and an image of the liquid in the liquid collecting chamber is obtained through the optical window;

[0013] performing image preprocessing on the liquid image to extract image features;

[0014] Input the image features into a trained oil-water discrimination model and output a discrimination result;

[0015] The liquid discharge from the liquid collecting chamber is controlled based on the discrimination result.

[0016] Optionally, the liquid image includes a pure infrared band image and a white light RGB image, and preprocessing the liquid image includes:

[0017] Performing denoising on the pure infrared band image to obtain a denoised image, and selecting a central area of ​​the denoised image as a feature extraction area;

[0018] Grayscale processing is performed on the white light RGB image to obtain a grayscale image, and the average brightness of the grayscale image is calculated.

[0019] Optionally, median filtering is used to perform denoising on the pure infrared band image.

[0020] Optionally, the image features include near-infrared absorbance and water droplet size of the pure infrared band image and average brightness, transparency, transmittance, dynamic viscosity and texture entropy of the white light RGB image.

[0021] Optionally, the oil-water discrimination model is used to output the discrimination result:

[0022] When the near-infrared absorbance is greater than a first threshold, the average brightness is greater than a second threshold, and the transparency is less than a third threshold, the determination result is a pure water phase;

[0023] When the transmittance is greater than a fourth threshold value and the dynamic viscosity is within a first preset range, the determination result is a pure oil phase;

[0024] When the water droplet diameter is within the second preset range and the texture entropy is within the third preset range, the determination result is an oil-water emulsion phase.

[0025] Optionally, the bottom drain port of the liquid collecting chamber is provided with a piezoelectric ceramic micro pump for controlling liquid discharge, and the piezoelectric ceramic micro pump has a pulse mode and an intermittent mode;

[0026] When the phase is identified as a water phase, the pulse mode is started to control liquid discharge;

[0027] When the emulsified phase is identified, the intermittent mode is activated to control liquid discharge.

[0028] Based on the same inventive concept, the present invention also provides an oil and gas cylinder drainage system based on machine vision, comprising:

[0029] An oil and gas cylinder, wherein an optical window is provided on the side wall of the liquid collecting chamber inside the oil and gas cylinder;

[0030] a visual detection module, configured to obtain an image of the liquid in the liquid collecting chamber through the optical window, and perform image preprocessing on the liquid image to extract image features;

[0031] a recognition module for constructing a training sample set, the training sample set including a plurality of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, and for constructing an oil-water discrimination model, training the oil-water discrimination model using the training sample set, and inputting the image features into the trained oil-water discrimination model to output a discrimination result;

[0032] An actuator is used to control the liquid discharge from the liquid collecting chamber based on the discrimination result.

[0033] Optionally, the visual inspection module includes a shooting head and a fixing tube. A through hole is opened on the side wall of the oil and gas cylinder at a position corresponding to the optical window. One end of the fixing tube extends obliquely downward from the outside of the through hole to the optical window, and the other end is used to install the shooting head. The shooting head is aligned with the optical window at a preset angle.

[0034] Optionally, the optical window is made of sapphire glass.

[0035] Optionally, the surface of the optical window is coated with a super oleophobic nano coating.

[0036] In the oil and gas cylinder drainage method and system based on machine vision provided by the present invention, an optical window is set to collect the liquid image in the liquid collecting chamber in real time, and the features of the liquid image are discriminated by using a trained oil-water discrimination model, and the discrimination results of pure oil phase, pure water phase or oil-water emulsion phase are output, thereby realizing non-contact oil-water emulsion discrimination with high discrimination accuracy and strong real-time performance, which is conducive to the precise control of liquid discharge in the liquid collecting chamber and solves the problem of lubricating oil waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Those skilled in the art will appreciate that the accompanying drawings are provided for a better understanding of the present invention and do not constitute any limitation on the scope of the present invention.

[0038] Figure 1 A flow chart of a method for draining an oil and gas cylinder based on machine vision provided in an embodiment of the present invention;

[0039] Figure 2 A schematic structural diagram of an oil and gas cylinder drainage system based on machine vision provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, advantages and features of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the drawings are in a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention. In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, please refer to the accompanying drawings. It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the conditions for the implementation of the present invention. Any modification of the structure, change in the proportional relationship or adjustment of the size, under the condition that the effect produced by the present invention and the purpose that can be achieved are the same or similar, should still fall within the scope of the technical content disclosed by the present invention.

[0041] As used herein, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. As used herein, the term "or" is generally used in a sense including "and / or" unless the context clearly dictates otherwise.

[0042] Please refer to Figure 1 This embodiment provides a method for draining an oil and gas cylinder based on machine vision, comprising:

[0043] S1. Constructing a training sample set, which includes image samples of a pure water phase, a pure oil phase, and oil-water emulsion phases with different emulsification degrees, and then constructing an oil-water discrimination model, and using the training sample set to train the oil-water discrimination model;

[0044] S2. An optical window is provided on the side wall of the liquid collecting chamber of the oil and gas cylinder to obtain an image of the liquid in the liquid collecting chamber through the optical window;

[0045] S3, performing image preprocessing on the liquid image and extracting image features;

[0046] S4, input the image features into the trained oil-water discrimination model and output the discrimination results;

[0047] S5. Control the liquid discharge from the liquid collecting chamber based on the discrimination result.

[0048] By setting an optical window to collect the liquid image in the liquid collecting chamber in real time, and using the trained oil-water discrimination model to discriminate the characteristics of the liquid image, the discrimination results of pure oil phase, pure water phase or oil-water emulsion phase are output, realizing non-contact oil-water emulsion discrimination with high discrimination accuracy and strong real-time performance, which is conducive to the precise control of liquid discharge in the liquid collecting chamber and solves the problem of lubricating oil waste.

[0049] First, execute S1 to construct a training sample set. The training sample set includes several image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees. Then, an oil-water discrimination model is constructed and trained using the training sample set. It should be noted that the samples in this image sample set can be acquired through image acquisition or generated using a generative adversarial model (GAN) as known in the art, and the present invention is not limited to this. Labels are also set to indicate pure water, pure oil, and oil-water emulsion regions.

[0050] Then, execute S2, set an optical window on the side wall of the oil-gas cylinder's liquid collection chamber, and obtain an image of the liquid in the liquid collection chamber through the optical window. The oil-gas cylinder (including intelligent oil-gas cylinders) is the core separation equipment in compressed air systems (such as screw air compressors). Its core function is to achieve efficient separation of oil-gas mixtures. The cylinder body is usually divided into three functional areas: a cyclone separation chamber, a filter element fine separation chamber, and a liquid collection chamber, which work together through a multi-stage physical separation mechanism. The liquid collection chamber is located at the bottom of the cylinder body and is used to collect the lubricating oil separated from the cyclone chamber and the filter element chamber. The oil is returned to the main engine lubrication system through the return oil pipe at the bottom of the liquid collection chamber to prevent secondary mixing. When pure water or oil-water emulsion accumulates at the bottom of the liquid collection chamber, it needs to be discharged in a timely manner to avoid affecting the normal operation of the air compressor.

[0051] Preferably, the optical window is made of sapphire glass, the surface of which is coated with a SiO2 superoleophobic nano-coating with a coating thickness of 100-200 nm and a contact angle of >150°.

[0052] In this embodiment, an existing visual inspection module is used to acquire an image of the liquid in the liquid collection chamber. The visual inspection module includes a camera head and a fixed tube. A through-hole is provided on the side wall of the oil and gas cylinder at a position corresponding to the optical window. One end of the fixed tube extends obliquely downward from the through-hole to the optical window, and the other end is used to mount the camera head, which is aligned with the optical window at a preset angle. The camera head is, for example, a 2-megapixel industrial camera equipped with two light sources: an 850nm infrared LED and a white light LED, for capturing pure infrared band images and white light RGB images, respectively. Furthermore, a fill light source can be configured as required, which is not limited by the present invention.

[0053] Preferably, the fixing cylinder is arranged at 60 degrees to the side wall of the liquid collecting chamber to avoid direct reflection of the oil film when taking images.

[0054] When capturing an image of the liquid in the liquid collection chamber using the camera head, the 850nm infrared light source is first turned on, while other light sources are turned off to avoid interference. This captures a pure infrared band image. The 850nm infrared LED light source is then turned off, and the white light LED light source is turned on, immediately capturing a white light RGB image. Alternatively, the white light RGB image can be captured first, followed by the pure infrared band image, and this is not a limitation of the present invention.

[0055] Then, S3 is executed to perform image preprocessing on the liquid image and extract image features. In this embodiment, the liquid image includes a pure infrared band image and a white light RGB image. The preprocessing of the liquid image includes:

[0056] The pure infrared band image is denoised to obtain a denoised image, and the central area of ​​the denoised image is selected as the feature extraction area;

[0057] The white light RGB image is grayscaled to obtain a grayscale image, and the average brightness of the grayscale image is calculated.

[0058] In this embodiment, median filtering is used to perform denoising on the pure infrared band image. Specifically, for each pixel in the liquid image , the output of the median filter Defined as:

[0059]

[0060] Where W is the filter window, median represents the median operation, and (i, j) is the offset relative to the center point within the filter window.

[0061] Then the central area of ​​the denoised image is taken as the feature extraction area to avoid the influence of edge reflection on the discrimination result.

[0062] The calculation formula for the average brightness µ of a grayscale image is as follows:

[0063]

[0064] Where N is the total number of image pixels and gray(m) is the grayscale value of the mth pixel.

[0065] In this embodiment, the image features include near-infrared absorbance and water droplet size of pure infrared band images, and average brightness, transparency, transmittance, dynamic viscosity, and texture entropy of white light RGB images.

[0066] The near-infrared absorbance I is obtained as follows:

[0067] First calculate the current infrared light intensity I, the formula is as follows:

[0068]

[0069] in, is the pixel value at the coordinate (x, y) of the pure infrared band image after denoising, ROI is the region of interest, and N is the total number of pixels in the region of interest.

[0070] Then calculate the near-infrared absorbance A 850 , the formula is as follows:

[0071]

[0072] Where I0 is the reference light intensity, which can be pre-calibrated and calculated from the pure infrared band image taken when the liquid collection cavity is empty. The greater the absorbance, the more infrared light is absorbed by the water molecules.

[0073] Water droplet diameter d i Using edge detection (Canny algorithm), the calculation formula is as follows:

[0074]

[0075] Among them, A i is the pixel area of ​​the i-th water droplet.

[0076] Transparency σ can be expressed by the grayscale standard deviation of the white light RGB image, and its calculation formula is as follows:

[0077]

[0078] Where gray(x, y) is the pixel value of the grayscale image at coordinate (x, y), and N is the total number of image pixels.

[0079] The calculation formula of transmittance T is as follows:

[0080]

[0081] Among them, mean(gray) is the average value of all pixels in the grayscale image.

[0082] The dynamic viscosity η can be estimated by reverse calculation of the transmittance, and the calculation formula is as follows:

[0083]

[0084] Among them, k and n are oil characteristic constants, such as the typical values ​​of mineral oil: k=120, n=1.3).

[0085] Texture entropy E is calculated based on the gray-level co-occurrence matrix and is used to reflect the complexity of image texture. Its calculation formula is as follows:

[0086]

[0087] Where p(x, y) is the joint probability that grayscale values ​​i and j are adjacent in the image.

[0088] Then, S4 is executed to input the image features into the trained oil-water discrimination model and output the discrimination result.

[0089] Specifically, the oil-water discrimination model outputs the discrimination results as follows:

[0090] When the near-infrared absorbance is greater than the first threshold, the average brightness is greater than the second threshold, and the transparency is less than the third threshold, the judgment result is a pure water phase;

[0091] When the transmittance is greater than the fourth threshold value and the dynamic viscosity is within the first preset range, the determination result is a pure oil phase;

[0092] When the water droplet size is within the second preset range and the texture entropy is within the third preset range, the determination result is an oil-water emulsion phase.

[0093] For example, when the near-infrared absorbance A 850 >0.5, average brightness μ>220, transparency σ<10, the result is pure water phase, near infrared absorbance A 850 >0.5 indicates that water molecules "eat up" a large amount of infrared light, which can exclude oil and air. The average brightness µ>220 indicates that the liquid is as transparent as glass, which excludes emulsions (whitish and turbid). Transparency σ<10 indicates that the image is uniform and free of particles, which can exclude air bubbles and impurities.

[0094] When both transmittance T>80% and dynamic viscosity are within the range of 40-100 cSt, the result is a pure oil phase. Transmittance T>80% indicates that the oil is as transparent as glass, excluding emulsions (turbidity) or contaminants. 40≤η≤100 ensures that the lubricant is not aged, excluding fuel or degraded oil (abnormal viscosity).

[0095] When both the water droplet size is within the 1-10μm range and the texture entropy E is within the 2.5-4.0 range, the phase is identified as an oil-water emulsion. Generally speaking, in pure water or pure oil, oil and water separate spontaneously due to their polarity differences (water is highly polar, oil is non-polar), resulting in two phases. One phase is characterized by complete separation after standing due to the density difference, resulting in no dispersed phase droplets and a clear interface. Mechanical disturbance (such as stirring) may briefly form large oil droplets, but due to high interfacial tension, the droplets easily coalesce (low coalescence barrier), rapidly floating or sinking, and becoming unstable. Consequently, the water droplet size is either zero or larger than 50μm. Oil-water emulsions, on the other hand, are more likely to form microdroplets, typically ranging from 1-10μm. Furthermore, pure water or pure oil are relatively homogeneous, with a texture entropy E generally less than 2. Oil-water emulsions are more granular, with a texture entropy E generally ranging from 2.5-4.0.

[0096] Preferably, the bottom drain port of the liquid collecting chamber is provided with a piezoelectric ceramic micro pump for controlling liquid discharge, and the piezoelectric ceramic micro pump has a pulse mode and an intermittent mode;

[0097] When the water phase is identified, the pulse mode is started to control the liquid discharge (such as single opening time 0.2-0.5 seconds, interval time 1-3 seconds);

[0098] When an emulsified phase is identified, the intermittent mode is started to control liquid discharge (such as a single opening time of 1-2 seconds and an interval time of 10-30 seconds).

[0099] Based on the same invention concept, please refer to Figure 2 The embodiment of the present invention further provides an oil and gas cylinder drainage system based on machine vision, comprising:

[0100] The oil and gas cylinder 100 has an optical window provided on the side wall of the liquid collecting chamber 110 therein;

[0101] The visual detection module 200 is used to obtain the liquid image in the liquid collecting chamber 110 through the optical window, and perform image preprocessing on the liquid image to extract image features;

[0102] Recognition module 300 is used to construct a training sample set, which includes a number of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, and is used to construct an oil-water discrimination model, train the oil-water discrimination model using the training sample set, input image features into the trained oil-water discrimination model, and output discrimination results;

[0103] The actuator 400 is used to control the discharge of liquid from the liquid collecting chamber 110 based on the discrimination result.

[0104] In this embodiment, the actuator 400 is preferably a piezoelectric ceramic micropump.

[0105] The visual inspection module 200 includes a shooting head 210 and a fixed tube 220. A through hole is opened on the side wall of the oil and gas cylinder 100 at a position corresponding to the optical window. One end of the fixed tube 220 extends obliquely downward from the outside of the through hole to the optical window, and the other end is fixed with the shooting head 210, which is aligned with the optical window at a preset angle.

[0106] Preferably, the fixing cylinder 220 is arranged at an angle of 60° to the side wall of the liquid collecting chamber 110 to avoid direct reflection of the oil film when taking images.

[0107] Preferably, the optical window is made of sapphire glass, the surface of which is coated with a SiO2 superoleophobic nano-coating with a coating thickness of 100-200 nm and a contact angle of >150°.

[0108] Since the oil and gas cylinder drainage system based on machine vision provided in the embodiment of the present invention and the oil and gas cylinder drainage method based on machine vision described above belong to the same inventive concept, the oil and gas cylinder drainage system based on machine vision provided by the present invention has all the advantages of the oil and gas cylinder drainage method based on machine vision described above, and therefore the beneficial effects of the oil and gas cylinder drainage system based on machine vision provided by the present invention will not be described one by one here.

[0109] In summary, the embodiments of the present invention provide a machine vision-based method and system for draining oil and gas cylinders. This system uses an optical window to capture real-time images of the liquid in the liquid collection chamber. A trained oil-water discrimination model is then used to identify the characteristics of the liquid image and output the results of the pure oil phase, pure water phase, or oil-water emulsion phase. This achieves non-contact oil-water emulsion discrimination with high accuracy and real-time performance. Furthermore, a piezoelectric ceramic micropump is used to precisely control liquid discharge from the collection chamber, addressing the issues of lubricant waste and high maintenance costs.

[0110] Furthermore, it should be recognized that although the present invention has been disclosed above with reference to preferred embodiments, the above embodiments are not intended to limit the present invention. Any person skilled in the art can utilize the above disclosed technical content to make many possible changes and modifications to the technical solution of the present invention, or modify it into equivalent embodiments with equivalent variations, without departing from the scope of the technical solution of the present invention. Therefore, any simple modifications, equivalent variations, and modifications made to the above embodiments based on the technical essence of the present invention, without departing from the content of the technical solution of the present invention, shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A method for draining oil and gas cylinders based on machine vision, characterized in that: include: Constructing a training sample set, the training sample set including a plurality of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, then constructing an oil-water discrimination model, and using the training sample set to train the oil-water discrimination model; An optical window is provided on the side wall of the liquid collecting chamber of the oil and gas cylinder, and an image of the liquid in the liquid collecting chamber is obtained through the optical window; performing image preprocessing on the liquid image to extract image features; Input the image features into a trained oil-water discrimination model and output a discrimination result; The liquid discharge from the liquid collecting chamber is controlled based on the discrimination result.

2. The oil and gas cylinder drainage method based on machine vision according to claim 1, characterized in that: The liquid image includes a pure infrared band image and a white light RGB image, and preprocessing the liquid image includes: Performing denoising on the pure infrared band image to obtain a denoised image, and selecting a central area of ​​the denoised image as a feature extraction area; Grayscale processing is performed on the white light RGB image to obtain a grayscale image, and the average brightness of the grayscale image is calculated.

3. The oil and gas cylinder drainage method based on machine vision according to claim 2, characterized in that: The pure infrared band image is denoised using median filtering.

4. The oil and gas cylinder drainage method based on machine vision according to claim 2, characterized in that: The image features include the near-infrared absorbance and water droplet size of the pure infrared band image and the average brightness, transparency, transmittance, dynamic viscosity and texture entropy of the white light RGB image.

5. The oil and gas cylinder drainage method based on machine vision according to claim 4 is characterized in that: The oil-water discrimination model is used to output the discrimination results: When the near-infrared absorbance is greater than a first threshold, the average brightness is greater than a second threshold, and the transparency is less than a third threshold, the determination result is a pure water phase; When the transmittance is greater than a fourth threshold value and the dynamic viscosity is within a first preset range, the determination result is a pure oil phase; When the water droplet diameter is within the second preset range and the texture entropy is within the third preset range, the determination result is an oil-water emulsion phase.

6. The oil and gas cylinder drainage method based on machine vision according to claim 5, characterized in that: The bottom drain port of the liquid collecting chamber is provided with a piezoelectric ceramic micro pump for controlling liquid discharge, and the piezoelectric ceramic micro pump has a pulse mode and an intermittent mode; When the phase is identified as a water phase, the pulse mode is started to control liquid discharge; When the emulsified phase is identified, the intermittent mode is activated to control liquid discharge.

7. A machine vision-based oil and gas cylinder drainage system, characterized in that: include: An oil and gas cylinder, wherein an optical window is provided on the side wall of the liquid collecting chamber inside the oil and gas cylinder; a visual detection module, configured to obtain an image of the liquid in the liquid collecting chamber through the optical window, and perform image preprocessing on the liquid image to extract image features; a recognition module for constructing a training sample set, the training sample set including a plurality of image samples of pure water phase, pure oil phase, and oil-water emulsion phases with different emulsification degrees, and for constructing an oil-water discrimination model, training the oil-water discrimination model using the training sample set, and inputting the image features into the trained oil-water discrimination model to output a discrimination result; An actuator is used to control the liquid discharge from the liquid collecting chamber based on the discrimination result.

8. The oil and gas cylinder drainage system based on machine vision according to claim 7 is characterized in that: The visual inspection module includes a shooting head and a fixing tube. A through hole is opened on the side wall of the oil and gas cylinder at a position corresponding to the optical window. One end of the fixing tube extends obliquely downward from the outside of the through hole to the optical window, and the other end is used to install the shooting head. The shooting head is aligned with the optical window at a preset angle.

9. The oil and gas cylinder drainage system based on machine vision according to claim 8, characterized in that: The optical window is made of sapphire glass.

10. The oil and gas cylinder drainage system based on machine vision according to claim 7, characterized in that: The surface of the optical window is coated with a super oleophobic nano coating.

Images