A video acquisition device and analysis method for wear particles in flowing oil

Abstract

A video acquisition device for wear particles in flowing oil, comprising an industrial camera, a host computer, a telecentric lens, a bowl-shaped light source, and an oil flow detection module; the industrial camera is installed at the eye position of the telecentric lens, the outer ring of the observation end of the telecentric lens is installed with a bowl-shaped light source, and the signal output end of the camera is connected to the signal acquisition end of the host computer; the center of the telecentric lens and the bowl-shaped light source is the same axis, and the telecentric lens and the bowl-shaped light source face the oil flow detection module; it also includes a reflective background plate, and the side of the oil flow detection module facing away from the bowl-shaped light source is provided with a reflective background plate, and the reflective background plate corresponds to the bowl-shaped light source; the oil flow detection module includes an oil circuit block and a microcrystalline heating plate, the oil circuit block is transparent, and the oil circuit block body is provided with a microchannel, and the microchannel is located in the observation range of the telecentric lens and the bowl-shaped light source. The imaging lens of this scheme can improve the depth of field of the imaging object, and the combination of the two can clearly collect the image information of wear particles in the oil.

Description

Technical Field

[0001] The invention relates to oil detection technology, in particular to a video acquisition device and an analysis method for wear particles in flowing oil. Background Art

[0002] The most typical wear particle analysis technologies and devices are the Online Visual Ferrograph (OLVF) and the Laser Net Fines (LNF). The principle of the Online Visual Ferrograph (OLVF) system is to use magnetism to adsorb the wear particles in the oil flow to a plane, and then use a camera to photograph the wear particles. This system is similar to an analytical ferroscope. Since the wear particles are adsorbed by magnetism, they always form aggregate chains. Although researchers have tried to separate these aggregated wear particles, it is still a problem to extract the characteristics of individual wear particles. In contrast, the automatic wear particle analyzer (LNF) system uses a circulation pool. Although it does not use magnetism to obtain the moving wear particles, the imaging lens of this type of analyzer cannot perform all-round imaging of the flowing oil, resulting in blur and shadow problems in the imaging. Improvements need to be made to the above problems. Summary of the invention

[0003] The first purpose of the present invention is to provide a video acquisition device for wear particles in flowing oil. The imaging lens of this scheme can improve the depth of field of the imaging object. The combination of the two can clearly capture the image information of wear particles in the oil.

[0004] The second purpose of the present invention is to propose a method for using a video acquisition device for wear particles in flowing oil, which captures and tracks the particle information captured in the oil, collects information from multiple perspectives generated by a single particle tumbling in the oil, and finally counts the wear particles to perform wear diagnosis analysis.

[0005] To achieve this object, the present invention adopts the following technical solutions:

[0006] A video acquisition device for wear particles in flowing oil, comprising an industrial camera, a host computer, a telecentric lens, a bowl-shaped light source, and an oil flow detection module;

[0007] The industrial camera is installed at the eye position of the telecentric lens, a bowl-shaped light source is installed on the outer ring of the observation end of the telecentric lens, and the signal output end of the camera is connected to the signal acquisition end of the host computer;

[0008] The telecentric lens and the bowl-shaped light source have a single common axis, and the telecentric lens and the bowl-shaped light source face the oil flow detection module;

[0009] It also includes a reflective background plate, and the reflective background plate is provided on a side of the oil flow detection module facing away from the bowl-shaped light source, and the reflective background plate corresponds to the bowl-shaped light source;

[0010] The oil flow detection module includes an oil circuit block and a microcrystalline heating plate. The oil circuit block is transparent. The oil circuit block body is provided with a microchannel. The microchannel is located within the observation range of the telecentric lens and the bowl-shaped light source. The microcrystalline heating plate is arranged on both sides of the oil circuit block. The heating direction of the microcrystalline heating plate is toward both sides of the oil circuit block.

[0011] Preferably, the inner wall of the microchannel is coated with a nano-silicon dioxide coating, and the microchannel is relatively perpendicular to the observation plane of the telecentric lens;

[0012] The microchannel is connected to a multi-stage pressure balancing pool, and the diameter of the pressure balancing pool is larger than the diameter of the microchannel.

[0013] Preferably, the spherical inner concave surface of the bowl-shaped light source is a reflective surface, and a luminous ring is provided at the opening position of the spherical inner concave surface, and the irradiation direction of the luminous ring is toward the spherical inner concave surface.

[0014] Preferably, the oil inlet of the oil circuit block is connected to the oil pool through a pipeline, and the oil outlet of the oil circuit block is connected to a metering pump through a pipeline.

[0015] Preferably, the oil circuit block is made of quartz or quartz glass.

[0016] Preferably, the reflecting surface includes a mirror reflecting device and a diffuse reflecting device.

[0017] Preferably, the depth of the microchannel is less than 400 μm.

[0018] A video analysis method for wear particles in flowing oil comprises the following steps:

[0019] S100: Connect the oil inlet of the oil block to the oil pool through a pipeline, connect the oil outlet to the metering pump, connect the industrial camera to the host computer, and connect the bowl-shaped light source and the metering pump to the external control circuit;

[0020] S200: Turn on the video acquisition system, and control the on / off and brightness of the LED lamp beads of the bowl-shaped light source by the given voltage of the control circuit to illuminate the microchannel of the oil block and provide video images for the industrial camera; at the same time, the software controls the flow rate and flow velocity of the metering pump to ensure that the oil sample passes through the microchannel smoothly during collection;

[0021] S300: Adjust the distance between the telecentric lens and the microfluidic device so that a clear image can be presented on the camera target surface, set the camera's exposure time, acquisition frame rate and other parameters, and start continuous and dynamic image acquisition;

[0022] S400: The collected video stream is input into the host computer for further processing. The wear particle extraction based on the video stream file can make full use of the continuity characteristics of the video stream and use the target tracking algorithm to extract the contour and image of the particles; after all particles are collected, the target recognition algorithm is used to identify and classify the particles; through the above steps, the pixel-level extraction of wear particles in the video is completed, and the number, area, size, etc. of wear particles can be calculated, and the wear of the test oil can be diagnosed and analyzed; finally, all particle images are arranged and displayed according to certain rules to complete the visualization of the results, and the wear diagnosis results are issued;

[0023] S500: The metering pump uses a higher flow rate to flush the oil circuit and micro-flow to prevent abrasive particles from adhering and affecting the detection accuracy of the next experiment;

[0024] S600: Turn off the LED lamp beads of the bowl-shaped light source and the metering pump in the video acquisition system to end the analysis and test of the oil sample.

[0025] Furthermore, in step 4:

[0026] Taking the processing of the current t-th frame result as an example, the specific steps are as follows:

[0027] S410: Use the YOLO algorithm to identify wear particles in the tth frame of the video, and combine it with the MOG2 algorithm to specifically capture dynamic small particles in the image, so as to supplement the deficiency of the YOLO algorithm in extracting small particles;

[0028] S420: using Kalman filtering technology to predict possible positions of particles in the current frame, i.e., the tth frame, based on the motion information of the particles in the previous frame, i.e., the t-1 frame;

[0029] S430: using the Hungarian matching algorithm, the particles in the t-th frame, that is, the current frame, are associated with the particles in the t-1-th frame, that is, the previous frame, and corresponding IDs are assigned to the particles in the current frame, thereby achieving continuous tracking of wear particles between frames, and recording image information of each tracked particle;

[0030] S440: When all video frames have been tracked, a convolutional neural network is used to identify the types and materials of all tracked wear particles, and calculate the area and size of the particles;

[0031] S450: Performing statistical analysis based on the model category, material, area, and size information of all tracked target particles, and performing wear diagnosis analysis.

[0032] One of the above technical solutions includes the following beneficial effects: 1. The bowl-shaped light source cooperates with the telecentric objective lens and the reflective surface to identify the color of the abrasive particles in the oil, and can identify non-ferrous metals such as Cu, Al, Babbitt alloy, etc.; it can identify iron oxides such as black Fe3O4 and red Fe2O3, etc. The wear type and wear position of the abrasive particles can be determined by combining the morphology and color of the abrasive particles; 2. The design of the telecentric objective lens in the oil abrasive particle monitoring system obtains that the lateral magnification of the object side is approximately constant, and a greater depth of field is obtained, so that the whole system has a better imaging effect; 3. The microchannel design uses quartz glass as the processing material in the material selection and processing of the microchannel, and uses the optical glue integrated process to form it in one time, and the overall glass surface is clear and flat. It ensures that the entire microchannel sample pool is clear and transparent, and there is no interference from bubbles in the glass and uneven glass surface on the imaging. In the design of the microchannel, a pressure balance pool is designed. After the oil enters the microchannel through the pipeline, it first passes through the pressure balance pool to reduce the pressure of the oil sample entering, so that the flow rate of the oil sample is slower and more stable when passing through the detection area, ensuring excellent imaging effects. In the installation of the oil circuit, a vertical installation oil inlet method is adopted to avoid the blockage of the microchannel by abrasive deposition and the interference of abrasive particles by the channel boundary; in addition, a non-interference heating method is adopted for the microchannel, so that the microchannel has temperature, which is convenient for the flow of oil, reduces the adhesion of cold oil on the inner wall of the channel, and improves the observation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a schematic diagram of the overall structure of an embodiment of the present invention;

[0034] Figure 2 is a schematic structural diagram of a microfluidic channel according to an embodiment of the present invention;

[0035] Figure 3 is a flow chart of a method according to an embodiment of the present invention;

[0036] Figure 4 is a flowchart of specific steps in step S400 of an embodiment of the present invention;

[0037] Figure 5 It is a single frame of video of copper, iron, fiber and plastic mixture collected by the system without color for comparison;

[0038] Figure 6 It is a single frame picture of the copper, iron, fiber and plastic mixture video collected by this system in color;

[0039] Figure 7 It is a schematic diagram of the conventional imaging of objects of equal height for comparison;

[0040] Figure 8This is the imaging principle diagram of the telecentric objective lens used in this scheme;

[0041] Among them: 1-reflective background plate, 2-oil block, 3-bowl-shaped light source, 4-telecentric lens, 5-industrial camera, 6-oil pool, 7-metering pump, 8-host computer, 9-oil inlet, 10-pressure balance tank, 11-microchannel, 12-oil outlet. DETAILED DESCRIPTION

[0042] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0043] like Figure 1 As shown, a video acquisition device for wear particles in flowing oil includes an industrial camera 5, a host computer 8, a telecentric lens 4, a bowl-shaped light source 3, and an oil flow detection module;

[0044] The industrial camera 5 is installed at the eye position of the telecentric lens 4, the outer ring of the observation end of the telecentric lens 4 is installed with a bowl-shaped light source 3, and the signal output end of the camera is connected to the signal collection end of the host computer 8;

[0045] The telecentric lens 4 and the bowl-shaped light source 3 have a single common axis, and the telecentric lens 4 and the bowl-shaped light source 3 face the oil flow detection module;

[0046] It also includes a reflective background plate 1, and the reflective background plate 1 is provided on a side of the oil-through detection module facing away from the bowl-shaped light source 3, and the reflective background plate 1 corresponds to the bowl-shaped light source 3;

[0047] The oil flow detection module includes an oil circuit block 2 and a microcrystalline heating plate. The oil circuit block 2 is transparent. The body of the oil circuit block 2 is provided with a microchannel 11. The microchannel 11 is located within the observation range of the telecentric lens 4 and the bowl-shaped light source 3. The microcrystalline heating plate is arranged on both sides of the oil circuit block 2. The heating direction of the microcrystalline heating plate is toward both sides of the oil circuit block.

[0048] The characteristic of this scheme is that the flowing oil is monitored by the telecentric lens 4 in cooperation with the bowl-shaped light source 3, and the wear particles that can display color are captured. When the oil passes through the oil circuit block 2, the light irradiated by the bowl-shaped light source 3 fully irradiates the wear particles, and cooperates with the reflective background plate 1 to form a combination of transparent and reflected light sources to illuminate the oil circuit block 2, so that the wear particles can better present the shape and color, and finally be photographed by the industrial camera 5; because the scheme adopts a telecentric objective lens, the imaging magnification of the wear particles at different heights in the oil is consistent, and different sizes of pictures will not be taken at different heights for a particle, so the shape of the wear particles can be identified, so as to perform tracking shooting. Through tracking shooting and color imaging, more information about the wear particles can be obtained more accurately, such as the size, color, material, etc. of the wear particles. Accurately and clearly obtaining information about the wear particles is conducive to evaluating the degree and condition of equipment wear. The scheme uses the oil circuit block 2 in conjunction with the microcrystalline heating plate to make the microchannel 11 have a temperature, thereby improving the fluidity of the oil, reducing the adhesion of oil to the inner wall of the microchannel 11 and the adhesion of wear particles to the inner wall of the channel, making the inner wall of the microchannel 11 clean, and providing a good observation environment, which can further improve the observation effect; the temperature of the microcrystalline heating plate can be adjusted, and quartz glass has good thermal conductivity. The combination of the two can create a microchannel 11 with a temperature that is convenient for the flow of engine oil, and will not affect the shape and color of the oil, wear particles and the oil circuit block 2. The microcrystalline heating plate is infrared heating and does not generate light, and will not affect the imaging of the color of the wear particles themselves.

[0049] Telecentric objective lens imaging explanation, assuming there are two identical targets A and B with the same height, both y. These two targets are imaged through an objective lens, and the aperture is in the middle of the objective lens, such as Figure 7 As shown. Since the object distances of the two targets are not equal (lA≠lB), the principal rays of the maximum field of view are inconsistent, or in other words, their corresponding field of view angles w are not equal, and the corresponding image heights on the image plane are also not equal (y'A≠y'B). If this system is used to measure the linear dimensions of an object, errors will occur. Although the heights of the objects are the same, different object distances correspond to different principal rays, which in turn lead to different image heights, ultimately resulting in errors in magnification measurement. It can be seen that the inconsistency of the principal rays of objects of equal height is a key factor in inaccuracy. If the aperture stop is moved back to the rear focal plane, the entrance pupil is pushed forward to infinity in the object space, then the principal rays of objects A and B coincide and are parallel to the optical axis, forming telecentric imaging, such as Figure 8As shown. Since the entrance pupil is at infinity, the principal ray in the object space is parallel to the optical axis. Although objects A and B have different object distances, the image heights of the two objects are the same, and the difference in the longitudinal position of the objects does not cause errors in the magnification measurement, that is, the lateral magnification in the object space is approximately constant. Therefore, the selection of telecentric lens 4 not only ensures the same magnification but also obtains a greater depth of field, so that the entire system has a better imaging effect.

[0050] like Figure 2 As shown, the inner wall of the microchannel 11 is coated with a nano-silicon dioxide coating, and the microchannel 11 is relatively perpendicular to the observation plane of the telecentric lens 4;

[0051] The microchannel 11 is connected to a multi-stage pressure balancing pool 10 , and the diameter of the pressure balancing pool 10 is larger than the diameter of the microchannel 11 .

[0052] The design of the microfluidic channel 11 uses quartz glass as the processing material in the material selection and processing of the microfluidic channel 11, and uses the optical glue integrated process to form it in one time, and the overall glass surface is clear and flat. The clarity and transparency of the entire microfluidic channel 11 sample pool is guaranteed, and there are no bubbles in the glass and the uneven surface of the glass interfere with the imaging. In the design of the microfluidic channel 11, a pressure balance tank 10 is designed. After the oil enters the microfluidic channel 11 through the pipeline, it first passes through the pressure balance tank 10 to reduce the pressure of the oil sample entering, so that the flow rate of the oil sample is slower and more stable when passing through the detection area, ensuring excellent imaging effects. In the installation of the oil circuit, a vertical installation oil inlet method is adopted to avoid the abrasive deposition blocking the microfluidic channel 11 and the abrasive particles being disturbed by the flow channel boundary. The solution is coated with a nano-silicon dioxide coating on the inner wall of the microfluidic channel 11 to reduce the adhesion of the oil to the inner wall of the microfluidic channel 11, ensure a clean visual environment in the microfluidic channel 11, and facilitate the cleaning of the microfluidic channel 11.

[0053] The spherical inner concave surface of the bowl-shaped light source 3 is a reflective surface, and a luminous ring is provided at the opening position of the spherical inner concave surface, and the irradiation direction of the luminous ring is toward the spherical inner concave surface.

[0054] By cooperating with the telecentric lens 4 and the reflected light from the bowl-shaped light source 3, the color of the abrasive particles can be well displayed. Figure 6 A picture of a mixture of copper, iron, fiber and plastic taken during oil dynamic measurement is shown in the figure.

[0055] The particles marked as copper have a bright yellow luster, the particles marked as iron have a bright silver luster, the particles marked as plastic have a pure black property, and the fibers can be identified by morphological characteristics. It can be seen that the wear type and wear position of the abrasive particles can be determined more quickly by combining the morphology and color of the abrasive particles.

[0056] like Figure 1 As shown, the oil inlet 9 of the oil circuit block 2 is connected to the oil pool 6 through a pipeline, and the oil outlet 12 of the oil circuit block 2 is connected to the metering pump 7 through a pipeline.

[0057] The speed of the oil passing through the microchannel 11 is controlled by the metering pump 7, and the oil flow rate is in a state that is easy to observe.

[0058] Wherein, the oil circuit block 2 is made of quartz or quartz glass.

[0059] Wherein, the reflecting surface includes a mirror reflection device and a diffuse reflection device.

[0060] Wherein, the depth of the microchannel 11 is less than 400 μm.

[0061] like Figure 3 As shown, a video analysis method for wear particles in flowing oil includes the following steps:

[0062] S100: Connect the oil inlet 99 of the oil block 2 to the oil pool 66 through a pipeline, connect the oil outlet 1212 to the metering pump 77, connect the industrial camera 55 to the host computer 88, and connect the bowl-shaped light source 33 and the metering pump 77 to the external control circuit;

[0063] S200: Turn on the video acquisition system, and control the on / off and brightness of the LED lamp beads of the bowl-shaped light source 33 by the given voltage of the control circuit to illuminate the microchannel 11 of the oil block 2 and provide a video image for the industrial camera 5; at the same time, the software controls the flow rate and flow velocity of the metering pump 7 to ensure that the oil sample passes through the microchannel 11 smoothly during the acquisition;

[0064] S300: adjusting the distance between the telecentric lens 4 and the microfluidic channel 11 device so that a clear image can be presented on the camera target surface, setting the camera exposure time, acquisition frame rate and other parameters, and starting continuous and dynamic image acquisition;

[0065] S400: The collected video stream is input into the host computer 8 for further processing. The wear particle extraction based on the video stream file can make full use of the continuity characteristics of the video stream and use the target tracking algorithm to extract the contour and image of the particles; after all particles are collected, the target recognition algorithm is used to identify and classify the particles; through the above steps, the pixel-level extraction of wear particles in the video is completed, and the number, area, size, etc. of wear particles can be calculated, and the wear of the test oil can be diagnosed and analyzed; finally, all particle images are arranged and displayed according to certain rules to complete the visualization of the results, and the wear diagnosis results are issued;

[0066] S500: the metering pump 7 uses a relatively high flow rate to flush the oil path and the micro-flow to prevent the abrasive particles from adhering and affecting the detection accuracy of the next experiment;

[0067] S600: Turn off the LED lamp beads of the bowl-shaped light source 33 and the metering pump 7 in the video acquisition system, and end the analysis and test of the oil sample.

[0068] This method uses a wear particle video acquisition device to track and extract the contours and images of wear particles flowing in the oil, and then compares and analyzes the extracted images to understand the specific wear location of the equipment; this method can output visual results by collecting pictures and videos; wear particle information can be accurately collected through visual collection combined with tracking algorithms, and the wear diagnosis speed can be improved by combining with statistical algorithms.

[0069] like Figure 4 As shown, the above technology is further explained, in step 4:

[0070] Taking the processing of the current t-th frame result as an example, the specific steps are as follows:

[0071] S410: Use the YOLO algorithm to identify wear particles in the tth frame of the video, and combine it with the MOG2 algorithm to specifically capture dynamic small particles in the image, so as to supplement the deficiency of the YOLO algorithm in extracting small particles;

[0072] The use of MOG2 algorithm makes up for the shortcomings of YOLO in extracting small particles and improves the comprehensiveness of wear particle information collection.

[0073] S420: using Kalman filtering technology to predict possible positions of particles in the current frame, i.e., the tth frame, based on the motion information of the particles in the previous frame, i.e., the t-1 frame;

[0074] Predict the movement position of particles, improve the accuracy of particle tracking collection, reduce tracking loss, and improve the collection effect of particle information at different positions;

[0075] S430: using the Hungarian matching algorithm, the particles in the t-th frame, that is, the current frame, are associated with the particles in the t-1-th frame, that is, the previous frame, and corresponding IDs are assigned to the particles in the current frame, thereby achieving continuous tracking of wear particles between frames, and recording image information of each tracked particle;

[0076] Archive the information of the same wear particle and compare and analyze it through an information set. More information can improve the accuracy of diagnostic analysis.

[0077] S440: When all video frames have been tracked, a convolutional neural network is used to identify the types and materials of all tracked wear particles, and calculate the area and size of the particles;

[0078] S450: Performing statistical analysis based on the model category, material, area, and size information of all tracked target particles, and performing wear diagnosis analysis.

[0079] This method uses a video acquisition device for wear particles in flowing oil to collect the wear particles in the oil. First, various wear particles are collected and calibrated by a comprehensive horizontal collection method, and then the whole process of various particles is followed and filmed, and in-depth collection is performed vertically. Finally, the collected image information is quantified in terms of area size; finally, the above information is uniformly diagnosed and analyzed. In the present invention, the detection of abrasive particles can be achieved without an external magnetic field, avoiding the difficulty in image recognition caused by the "abrasive particle chain" formed under magnetic induction conditions, especially the problem of small particle adsorption and agglomeration being difficult to identify; at the same time, it also avoids the problem of incomplete release of abrasive particles caused by incomplete demagnetization. The present invention does not use a magnetic field to adsorb ferromagnetic abrasive particles, but collects the most realistic flow effect of abrasive particles in oil, and can clearly image abrasive particles of different sizes. The abrasive particles are not limited to ferromagnetic abrasive particles, but all abrasive particles in the oil, including ferromagnetic and non-ferromagnetic particles, are detected.

[0080] The technical principle of the present invention is described above in conjunction with specific embodiments. These descriptions are only for explaining the principle of the present invention and cannot be interpreted as limiting the scope of protection of the present invention in any way. Based on the explanations herein, those skilled in the art can associate other specific implementations of the present invention without paying creative labor, and these methods will fall within the scope of protection of the present invention.

Claims

1. A video analysis method for wear particles in flowing oil, characterized in that: The method adopts a wear particle video acquisition device in flowing oil, including an industrial camera, a host computer, a telecentric lens, a bowl-shaped light source, and an oil flow detection module; The industrial camera is installed at the eye position of the telecentric lens, a bowl-shaped light source is installed on the outer ring of the observation end of the telecentric lens, and the signal output end of the camera is connected to the signal acquisition end of the host computer; The telecentric lens and the bowl-shaped light source have a single common axis, and the telecentric lens and the bowl-shaped light source face the oil flow detection module; It also includes a reflective background plate, and the reflective background plate is provided on a side of the oil flow detection module facing away from the bowl-shaped light source, and the reflective background plate corresponds to the bowl-shaped light source; The oil flow detection module includes an oil passage block and a microcrystalline heating plate, the oil passage block is transparent, a microchannel is provided on the oil passage block body, the microchannel is located in the observation range of the telecentric lens and the bowl-shaped light source, the microcrystalline heating plate is arranged on both sides of the oil passage block, and the heating direction of the microcrystalline heating plate is toward both sides of the oil passage block; The steps include: S100: Connect the oil inlet of the oil block to the oil pool through a pipeline, connect the oil outlet to the metering pump, connect the industrial camera to the host computer, and connect the bowl-shaped light source and the metering pump to the external control circuit; S200: Turn on the video acquisition device, and control the on / off and brightness of the LED lamp beads of the bowl-shaped light source through the given voltage of the external control circuit to illuminate the microchannel of the oil block and provide video images for the industrial camera; at the same time, the software controls the flow rate and flow velocity of the metering pump to ensure that the oil sample passes through the microchannel smoothly during collection; S300: Adjust the distance between the telecentric lens and the microchannel so that a clear image can be presented on the target surface of the industrial camera, set the exposure time and acquisition frame rate parameters of the industrial camera, and start continuous and dynamic image acquisition; S400: The collected video stream is input into the host computer for further processing. The wear particle extraction based on the video stream file can make full use of the continuity characteristics of the video stream and use the target tracking algorithm to extract the contour and image of the wear particles; after all the wear particles are collected, the target recognition algorithm is used to identify and classify the wear particles; through the above steps, the pixel-level extraction of the wear particles in the video is completed, the number, area and size of the wear particles can be calculated, and the wear of the test oil can be diagnosed and analyzed; finally, all the wear particle images are arranged and displayed according to certain rules to complete the visualization of the results, and the wear diagnosis results are issued; S500: The metering pump uses a higher flow rate to flush the oil block and microchannel to prevent wear particles from adhering and affecting the detection accuracy of the next experiment; S600: Turn off the LED lamp beads of the bowl-shaped light source and the metering pump in the video acquisition system to end the analysis and test of the oil sample.

2. The video analysis method for wear particles in flowing oil according to claim 1, characterized in that: The inner wall of the microchannel is coated with a nano-silicon dioxide coating, and the microchannel is relatively perpendicular to the observation plane of the telecentric lens; The microchannel is connected to a multi-stage pressure balancing pool, and the diameter of the pressure balancing pool is larger than the diameter of the microchannel.

3. The video analysis method for wear particles in flowing oil according to claim 1, characterized in that: The spherical inner concave surface of the bowl-shaped light source is a reflective surface, and a luminous ring is provided at the opening position of the spherical inner concave surface, and the irradiation direction of the luminous ring is toward the spherical inner concave surface.

4. The video analysis method for wear particles in flowing oil according to claim 1, characterized in that: The oil circuit block is made of quartz or quartz glass.

5. The video analysis method for wear particles in flowing oil according to claim 3, characterized in that: The reflecting surface includes a mirror reflecting device and a diffuse reflecting device.

6. The video analysis method for wear particles in flowing oil according to claim 1, characterized in that: The depth of the microchannel is less than 400 μm.

7. The video analysis method for wear particles in flowing oil according to claim 1, characterized in that: In step S400: The specific steps are as follows: S410: Use the YOLO algorithm to identify wear particles in the tth frame of the video, and combine it with the MOG2 algorithm to specifically capture dynamic small particles in the image, so as to supplement the deficiency of the YOLO algorithm in extracting small particles; S420: using Kalman filtering technology to predict possible positions of wear particles in a current frame, i.e., the tth frame, based on motion information of the wear particles in a previous frame, i.e., the t-1 frame; S430: using the Hungarian matching algorithm, the wear particles in the t-th frame, that is, the current frame, are associated with the wear particles in the t-1-th frame, that is, the previous frame, and the wear particles in the current frame are assigned corresponding IDs, thereby achieving continuous tracking of the wear particles between frames, and recording the image information of each tracked wear particle; S440: When all video frames have been tracked, a convolutional neural network is used to identify the types and materials of all tracked wear particles, and calculate the area and size of the wear particles; S450: Performing a wear diagnosis analysis based on statistical analysis of the model category, material, area, and size information of all tracked target wear particles.

Images