Lubricating system monitoring method and system based on machine vision

Abstract

The invention relates to the field of image processing, in particular to a lubricating system monitoring method and system based on machine vision, and the method comprises the steps: obtaining an abrasive particle image of oil in a lubricating system; obtaining a closed edge of the abrasive particle image, taking a region enclosed by the closed edge as a suspected abrasive particle region, and calculating a detail significance degree and an edge smoothness coefficient of the suspected abrasive particle region; calculating an improved value of a contrast gain based on the detail obvious degree and the edge smoothing coefficient to obtain an improved contrast gain, and performing enhancement processing on the abrasive particle image by using the improved contrast gain; and identifying the types of the abrasive particles by using the enhanced abrasive particle image, and judging whether the lubrication system is abnormal or not according to the types of the abrasive particles, so as to complete monitoring of the lubrication system. By accurately identifying the types of the abrasive particles, the monitoring accuracy of the whole lubricating system is improved, and it is ensured that possible problems in the lubricating system can be accurately found in time.

Description

Technical Field

[0001] The present invention relates to the field of image processing. More specifically, the present invention relates to a lubrication system monitoring method and system based on machine vision. Background Art

[0002] The lubrication system is the "blood circulation system" of mechanical equipment, and its core task is to continuously, moderately and cleanly provide lubrication for friction parts. By collecting oil samples in the lubrication system and analyzing the types of abrasive particles in the oil (such as metal chips, pollutants, etc.), the wear state of the machine or the health status of the lubrication system can be judged. The morphology, size and material of abrasive particles can reflect the types of wear inside the equipment (such as normal wear, fatigue wear, cutting wear, etc.).

[0003] In order to analyze abrasive particles more precisely, the prior art performs image processing on abrasive particles (such as denoising, contrast enhancement or edge sharpening, etc.) to highlight the detailed features of abrasive particles (such as surface texture, shape, cracks, etc.), which is convenient for subsequent classification or diagnosis.

[0004] However, there are significant differences in the morphological features of different abrasive particles (for example, metal abrasive particles need to highlight sharp edge features, while fiber pollutants need to maintain their soft texture structure). If fixed parameters are used for image enhancement processing, two types of problems will occur: First, the key features of some abrasive particles may be over-enhanced (such as artifacts generated due to excessive sharpening of metal edges); Second, the subtle features of other types of abrasive particles may have information loss. In addition, there is also the problem of interference from non-target signals during the enhancement process. Irrelevant information such as stray particles in the oil background and image sensor noise may be inappropriately enhanced, which will seriously interfere with the effective extraction of real abrasive particle features.

[0005] In summary, over-enhancement or noise will reduce the image quality and cannot achieve effective monitoring of the lubrication system. Summary of the Invention

[0006] In order to solve the technical problem that the effective monitoring of the lubrication system cannot be achieved due to the limitations of the image enhancement algorithm in the lubrication system monitoring, the present invention provides solutions in the following aspects.

[0007] In a first aspect, a lubrication system monitoring method based on machine vision includes: Obtaining an abrasive particle image of the oil in the lubrication system; Obtaining the closed edge of the abrasive particle image, the area enclosed by which is used as a suspected abrasive particle area, and calculating the detail saliency degree and the edge smoothness coefficient of the suspected abrasive particle area; Calculating an improved value of the contrast gain based on the detail saliency degree and the edge smoothness coefficient to obtain an improved contrast gain, and using the improved contrast gain to perform enhancement processing on the abrasive particle image; Using the enhanced abrasive particle images, identify the types of abrasive particles, and based on the types of abrasive particles, determine whether the lubrication system is abnormal, thereby completing the monitoring of the lubrication system.

[0008] First, by obtaining the abrasive particle images of the oil in the lubrication system, it is possible to visually detect whether there are abrasive particles in the oil. Also, since different types of abrasive particles have different morphological, dimensional, and compositional characteristics, it is necessary to identify the types of abrasive particles to distinguish between normal wear and abnormal wear. Among them, by calculating the detail saliency and edge smoothness coefficient of the suspected abrasive particle region, and based on these parameters, calculating the improved value of the contrast gain to enhance the abrasive particle image, the characteristics of the abrasive particles can be presented more clearly, reducing misjudgments caused by factors such as poor image quality. Accurate abrasive particle identification can achieve effective monitoring of the lubrication system.

[0009] Preferably, the enhancement process uses an adaptive contrast enhancement algorithm.

[0010] Preferably, the process of obtaining the detail saliency includes: Calculate the texture distinctiveness factor and the high-frequency saliency factor of the suspected abrasive particle region, and use the normalized value of the product of the texture distinctiveness factor and the high-frequency saliency factor as the detail saliency of the suspected abrasive particle region.

[0011] By calculating the texture distinctiveness factor (reflecting local texture features) and the high-frequency saliency factor (reflecting high-frequency information such as edges and gradients), the multi-scale features of the abrasive particle region can be comprehensively captured. For example, the texture distinctiveness factor can highlight texture differences such as scratches and edges on the surface of the abrasive particle, while the high-frequency saliency factor can enhance the clarity of the abrasive particle boundary. The normalized product of the two can effectively distinguish different types such as cutting abrasive particles and fatigue abrasive particles.

[0012] Preferably, the process of obtaining the edge smoothness coefficient includes: Obtain the corner points on the edge of the suspected abrasive particle region, construct its neighborhood with each corner point as the center, and obtain the continuously adjacent edge pixel points on both sides of the corner point in the neighborhood to form two edge line segments, and calculate the included angle between these two edge line segments; Normalize the included angle to obtain the roundness contribution degree of the corner point, and then use the mean value of the roundness contribution degrees of all corner points as the edge smoothness coefficient of the suspected abrasive particle region.

[0013] Particles generated by normal wear usually have smoother edges, while particles generated by abnormal wear (such as fatigue and fracture) may have sharper or more irregular edges. By focusing on the corner points (local curvature maximum points) on the edge, the key geometric features are captured. The included angle of the corner points directly reflects the local sharpness, and the normalized contribution degree converts the discrete corner point characteristics into continuous numerical values, avoiding the excessive influence of a single corner point on the overall evaluation.

[0014] Preferably, the improved contrast gain satisfies the following relationship: ; where is the improved contrast gain, is the initial contrast gain, is the mean value of the detail clarity of all suspected abrasive regions, is the mean value of the edge smoothness coefficient of all suspected abrasive regions.

[0015] Preferably, the process of obtaining the texture clarity factor includes: Extracting features from the suspected abrasive regions, including the mean value of the gradient magnitude and the variance of the gradient direction; Taking the product of the mean value of the gradient magnitude and the variance of the gradient direction as the texture clarity factor of the suspected abrasive region.

[0016] The gradient magnitude reflects the speed of pixel gray value change in the image. In the abrasive region, due to factors such as the edges and surface textures of the abrasives, the gray value change is usually relatively obvious. Calculating the mean value of the gradient magnitude can quantify the overall intensity of this gray value change; the gradient direction represents the direction of gray value change. The texture of the abrasives often has a certain directionality, such as scratches and texture lines on the abrasive surface. The variance of the gradient direction measures the diversity or consistency of this directionality. If the gradient directions in a region are relatively consistent, it indicates that the region may have relatively regular textures. If the gradient directions are relatively scattered, it may indicate that the region has more complex textures or is not an abrasive.

[0017] Taking the product of the mean value of the gradient magnitude and the variance of the gradient direction as the texture clarity factor comprehensively considers the intensity and directionality of the texture. This is because a high mean value of the gradient magnitude alone may only be due to large local gray value changes, but with chaotic directionality, it does not necessarily represent an abrasive; while a small variance of the gradient direction alone, without sufficient gradient magnitude, may also be just some noise or irrelevant textures in the image. By taking the product, the texture clarity factor will be large only when both the mean value of the gradient magnitude and the variance of the gradient direction reach a certain level, thus more accurately representing whether the texture characteristics of the suspected abrasive region are obvious.

[0018] Preferably, the process of obtaining the high-frequency significance factor includes: Performing a Fourier transform on the suspected abrasive region, masking the low-frequency region at the center of the spectrum through a mask, calculating the sum of the squares of the amplitudes of all pixels in the high-frequency region as the high-frequency energy; calculating the sum of the squares of the amplitudes of the entire spectrum as the total energy; taking the ratio of the high-frequency energy to the total energy as the high-frequency significance factor of the suspected abrasive region.

[0019] The Fourier transform converts an image from the spatial domain to the frequency domain. The high-frequency components correspond to detailed information such as edges and textures in the image, while the low-frequency components correspond to smooth regions (such as backgrounds or large uniformly-colored areas). Calculate the sum of the squared magnitudes of all pixels in the high-frequency region as the high-frequency energy, and calculate the sum of the squared magnitudes of the entire spectrum as the total energy. By making a comparison, the significance level of the high-frequency features in the suspected abrasive particle region can be quantitatively represented.

[0020] Preferably, the process of obtaining the texture distinctiveness factor includes: Calculate the LBP value for each pixel point in the suspected abrasive particle region to obtain an LBP histogram, and use the variance of the LBP histogram as the texture distinctiveness factor for the suspected abrasive particle region.

[0021] Preferably, the process of obtaining the edge smoothness coefficient includes: Use an edge detection algorithm to extract the edges of the suspected abrasive particle region, calculate the magnitude of the edge gradient to obtain a gradient image, calculate the variance of the gradient change based on the gradient image, and use the reciprocal of the variance of the gradient change as the edge smoothness coefficient.

[0022] In a second aspect, a lubrication system monitoring system based on machine vision includes: a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, any one of the lubrication system monitoring methods based on machine vision is implemented.

[0023] The beneficial effects of the present invention are: Through a series of fine processing and analysis of the abrasive particle image, including obtaining closed edges to determine the suspected abrasive particle region, calculating the significance level of details and the edge smoothness coefficient, etc., the types of abrasive particles can be more accurately identified. Different types of abrasive particles often correspond to different types of faults or abnormal conditions in the lubrication system. Accurately identifying the types of abrasive particles can provide strong evidence for judging whether the lubrication system is abnormal, thereby improving the accuracy of judging abnormal conditions in the lubrication system and effectively avoiding misjudgment or missed judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flowchart of the method from step S1 to step S4 in the lubrication system monitoring method based on machine vision according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention.

[0026] The application scenario of the present invention is: using an improved image enhancement processing technology to process the collected image in order to obtain a high-quality image for effective monitoring of the lubrication system.

[0027] Refer to Figure 1 , the machine vision-based lubrication system monitoring method includes steps S1-S4, specifically as follows: S1: Obtain the abrasive particle image of the oil in the lubrication system.

[0028] In one embodiment, the oil used in the lubrication system is sampled and detected regularly. When sampling, ensure that the sampling position is clean to avoid external contamination, and then filter the oil through a filter membrane to separate the wear particles.

[0029] Place the obtained wear particles on a ferrogram for depositing particles and heating and curing. Use a ferroscope to obtain the ferrogram of the wear particles, and use a scanning camera to obtain the image in the ferroscope microscope. Perform grayscale processing on the obtained image to obtain the abrasive particle image of the oil in the lubrication system.

[0030] S2: Obtain the closed edge of the abrasive particle image. The area enclosed by it is used as the suspected abrasive particle area, and calculate the detail saliency and edge smoothness coefficient of the suspected abrasive particle area.

[0031] The surface texture of the abrasive particles reflects their wear types. The surface of normal wear particles is relatively smooth and the texture is not obvious; while the surface texture of abnormal wear particles (such as cutting wear, corrosion wear) is more, and the gray value changes significantly. By analyzing the texture information, the type of abrasive particles can be judged, which provides a reference for image enhancement.

[0032] In one embodiment, first use the Canny edge detection algorithm or the Sobel operator to extract all the edges in the above-mentioned abrasive particle image to locate the contour boundary of the object in the image. Then, by judging whether the start and end coordinates of the edge coincide, the coincident ones are used as the closed edges. The continuous area enclosed by these closed edges is defined as the suspected abrasive particle area, and further identify the independent area with a complete contour in the image.

[0033] Furthermore, perform HOG (Histogram of Oriented Gradients) feature extraction on each suspected abrasive particle area to obtain two key parameters, namely the mean gradient amplitude (reflecting the overall intensity of pixel gray value change in the area or the sharpness of the edge) and the variance of gradient direction (characterizing the degree of dispersion of texture directions in the area or the texture complexity).

[0034] Finally, the product of the mean gradient magnitude and the variance of the gradient direction calculated above is used as the texture distinctiveness factor of the suspected abrasive particle region. This texture distinctiveness factor, as a composite index, takes into account both the edge strength and the texture complexity, and is used to quantitatively evaluate the texture significance of the region. The greater the texture significance, the more obvious the texture of the suspected abrasive particle region. Therefore, when performing image enhancement processing, in order to prevent over-enhancement of the texture, the contrast gain parameter should be reduced at this time; conversely, increase the contrast gain parameter.

[0035] It should be noted that the high-frequency information in the abrasive particle image also needs to be considered. High-frequency information usually reflects the noise and details in the image. If there is more high-frequency information in the abrasive particle region, it means that the region contains more noise or details. When performing image enhancement, if the high-frequency information is not controlled, it is easy to cause over-enhancement, making the noise more obvious and thus affecting the identification of abrasive particles.

[0036] Perform a discrete Fourier transform (DFT) on the suspected abrasive particle region to convert the spatial domain image into a frequency domain spectrum. The low-frequency components (central region) in the frequency domain represent the overall structure and contour of the image, and the high-frequency components (edge region) correspond to details, edges, and noise.

[0037] Then, mask the central low-frequency region of the spectrum to retain only the high-frequency region (similar to high-pass filtering in the frequency domain), and then calculate the sum of the squares of the amplitudes of all pixels in the retained high-frequency region as the high-frequency energy; calculate the sum of the squares of the amplitudes of the entire spectrum (i.e., low-frequency plus high-frequency) as the total energy.

[0038] Furthermore, the ratio of the high-frequency energy to the total energy is used as the high-frequency significance factor of the suspected abrasive particle region. The higher this high-frequency significance factor, the more high-frequency noise or details are contained in the suspected abrasive particle region. Therefore, reduce the contrast gain parameter to avoid misjudgment of abrasive particles caused by over-enhancing the noise; conversely, the lower this high-frequency significance factor, the smoother the suspected abrasive particle region and the lack of details. Therefore, increase the contrast gain parameter to improve the visibility of details for accurate identification of abrasive particles.

[0039] The texture distinctiveness factor and the high-frequency significance factor obtained through the above calculations can quantify the detailed information of the abrasive particle region.

[0040] Further, the normalized value of the product of the texture distinctiveness factor and the high-frequency significance factor is used as the detail significance degree of each suspected abrasive particle region.

[0041] Generally speaking, by combining the texture distinctiveness factor and the high-frequency significance factor, a detail distinctiveness coefficient is obtained, thereby dynamically adjusting the contrast gain parameter in the image enhancement algorithm. When the detail distinctiveness coefficient is large, it indicates that there are more details in the abrasive particle region, and the contrast gain needs to be reduced to avoid over-enhancement; conversely, the contrast gain can be appropriately increased to enhance the contrast of the image.

[0042] In another embodiment, the LBP value of each pixel point in the suspected abrasive region is calculated to obtain an LBP histogram, and the variance of the LBP histogram is used as the texture distinctiveness factor of the suspected abrasive region.

[0043] The edge features of abrasive particles are important bases for judging their wear types. The edges of normal wear particles are usually relatively smooth, while the edges of abnormal wear particles (such as cutting wear, corrosive wear) are more jagged and contain more high-frequency information. Therefore, by analyzing the smoothness of the edges, the types of abrasive particles can be identified more accurately.

[0044] In one embodiment, corner points on the edge of the suspected abrasive region are obtained through a corner detection algorithm. An n×n neighborhood is constructed centered on each corner point (exemplarily, n is taken as 3 and can be adjusted according to the actual resolution in other embodiments). In this neighborhood, the edge pixel points continuously adjacent to both sides of the corner point are identified to form two edge line segments, and the included angle between these two edge line segments is calculated. The more corner points and the smaller the included angle, the sharper the edge and the more detailed information it contains.

[0045] Then, the calculated included angle is normalized to obtain the bluntness contribution degree of the corner point, that is, the relational expression is satisfied as:

[0046] In the formula, is the bluntness contribution degree of the corner point, is the included angle between the edge line segments on both sides centered on the corner point.

[0047] Next, the value is calculated for all corner points in the suspected abrasive region, and then the mean value is taken as the edge smoothness coefficient of the suspected abrasive region. When the edge smoothness coefficient is large, it indicates that the abrasive edge is relatively smooth and there is less detailed information. In this case, the contrast gain parameter can be appropriately increased to enhance the overall contrast of the image and make the features of the abrasive particles more obvious; when the edge smoothness coefficient is small, it indicates that the abrasive edge contains more detailed information and the edge change is more obvious. In this case, the contrast gain parameter needs to be appropriately reduced to prevent over-enhancement of the detailed part, resulting in an increase in noise and affecting the recognition effect of the abrasive particles.

[0048] In another implementation, first use an edge detection algorithm (such as Sobel, Canny, etc.) to extract the edge of the suspected abrasive region, calculate the modulus value of the edge gradient to obtain a gradient image, perform statistical analysis on the gradient image, calculate the variance or standard deviation of the gradient change, and use the reciprocal of the variance or standard deviation of the gradient change as the edge smoothness coefficient. The smaller the variance or standard deviation, the smoother the gradient change and the smoother the edge.

[0049] Although the detail obviousness coefficient can reflect the texture and high-frequency information of the abrasive region, it mainly focuses on the details inside the abrasive grains. The edge smoothness coefficient, from another perspective, that is, the characteristics of the abrasive grain edges, provides supplementary information. By comprehensively considering these two coefficients, the detail information of the abrasive grains can be evaluated more comprehensively, and thus the image enhancement parameters can be adjusted more accurately.

[0050] According to the above operations, the detail salience and edge smoothness coefficients of all suspected abrasive regions can be calculated.

[0051] S3: Calculate the improved value of the contrast gain based on the detail obviousness and edge smoothness coefficients to obtain the improved contrast gain, and use the improved contrast gain to perform enhancement processing on the abrasive grain image.

[0052] According to the values of the detail obviousness coefficient and the edge smoothness coefficient, adjust the contrast gain parameter in the ACE adaptive contrast enhancement algorithm to avoid the problem of over-enhancement resulting in noise when enhancing the image containing abrasive grains.

[0053] Specifically, the improved contrast gain satisfies the relational expression as:

[0054] In the formula, is the improved contrast gain, is the initial contrast gain, is the mean value of the detail obviousness of all suspected abrasive regions, is the mean value of the edge smoothness coefficients of all suspected abrasive regions.

[0055] Through the improvement of the contrast gain parameter, targeted enhancement of the image containing abrasive grains can be achieved. During the enhancement process, the detail information and edge features on the surface of the abrasive grains are fully considered, avoiding the problems of over-enhancement or under-enhancement caused by fixed parameters in traditional image enhancement algorithms. The improved image enhancement algorithm can better highlight the features of the abrasive grains, providing a clearer and more accurate image basis for subsequent identification of abrasive grain types and determination of lubrication system abnormalities.

[0056] Furthermore, perform enhancement processing on the abrasive grain image according to the above improved contrast gain parameter to obtain the enhanced abrasive grain image.

[0057] It should be noted that in the embodiment of the present invention, the image enhancement technology used is: the ACE adaptive contrast enhancement algorithm.

[0058] S4: Use the enhanced abrasive grain image to identify the types of abrasive grains, and judge whether the lubrication system is abnormal according to the types of abrasive grains, and then complete the monitoring of the lubrication system.

[0059] In one embodiment, the clear abrasive particle images after enhancement processing using the above S3 are utilized to identify the types of abrasive particles, and whether there is an abnormality in the lubrication system is judged according to the categories of the abrasive particles, so as to realize the effective monitoring of the lubrication system.

[0060] Specifically, an abrasive particle type identification method (such as methods based on feature matching, machine learning, etc.) is used to analyze the enhanced abrasive particle images to identify the types of abrasive particles, including: Normal wear particles: Such as particles with regular shapes, smooth surfaces, small and uniform sizes, usually indicating that the lubrication system is in normal operation.

[0061] Abnormal wear particles: Such as particles with irregular shapes, rough surfaces, large sizes and uneven distributions, which may indicate that there is an abnormality in the lubrication system.

[0062] Common abnormal wear particles include: Cutting wear particles: Usually in the shape of flakes or blocks, with obvious cutting marks on the surface, which may indicate metal contact or insufficient lubrication in the lubrication system.

[0063] Corrosion wear particles: Usually in the shape of spheres or irregular shapes, with corrosion marks on the surface, which may indicate the presence of moisture or corrosive substances in the lubrication system.

[0064] Abrasive wear particles: Usually in the shape of particles, with abrasion marks on the surface, which may indicate the presence of hard particles in the lubrication system, resulting in component wear.

[0065] Fatigue wear particles: Usually in the shape of flakes or blocks, with fatigue cracks on the surface, which may indicate the presence of periodic loads or insufficient lubrication in the lubrication system.

[0066] Furthermore, the state of the lubrication system can be determined according to the types and quantities of the abrasive particles.

[0067] Specifically, an abnormal threshold of the abrasive particle quantity is set. When there are a small number of abnormal wear abrasive particles but their quantity does not exceed the set threshold, a warning signal can be sent to prompt the staff to pay attention to the operation state of the lubrication system. When the quantity of the abnormal wear abrasive particles exceeds the set threshold, it is determined that there is an abnormality in the lubrication system, and further inspection and maintenance are required. The oil level and oil quality of the lubrication system can be checked. If the oil level is found to be low, lubricating oil can be replenished in time, and the lubrication system can be maintained to ensure the normal operation of the motor.

[0068] The system includes a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the lubrication system monitoring method based on machine vision according to the first aspect of the present invention is realized.

[0069] The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be elaborated here.

[0070] It should be noted that, for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent shall be subject to the appended claims.

Claims

1. A monitoring method for a lubrication system based on machine vision, characterized in that Including: Obtain the abrasive particle image of the oil in the lubrication system; Obtain the closed edge of the abrasive particle image, and the area enclosed by it is used as the suspected abrasive particle area. Calculate the detail significance degree and the edge smoothness coefficient of the suspected abrasive particle area; Calculate the improved value of the contrast gain based on the detail significance degree and the edge smoothness coefficient to obtain the improved contrast gain, and use the improved contrast gain to enhance the abrasive particle image; Use the enhanced abrasive particle image to identify the type of abrasive particle, and judge whether the lubrication system is abnormal according to the type of abrasive particle, so as to complete the monitoring of the lubrication system.

2. The machine vision-based lubrication system monitoring method according to claim 1, wherein The enhancement process uses an adaptive contrast enhancement algorithm.

3. The machine vision-based lubrication system monitoring method according to claim 2, wherein The process of obtaining the detail significance degree includes: Calculate the texture distinctness factor and the high-frequency significance factor of the suspected abrasive particle area, and take the normalized value of the product of the texture distinctness factor and the high-frequency significance factor as the detail significance degree of the suspected abrasive particle area.

4. The method for monitoring a lubrication system based on machine vision according to claim 3, wherein The process of obtaining the edge smoothness coefficient includes: Obtain the corner points on the edge of the suspected abrasive particle area, construct its neighborhood centered on each corner point, obtain the edge pixel points continuously adjacent to both sides of the corner point in the neighborhood to form two edge line segments, and calculate the included angle between these two edge line segments; Normalize the included angle to obtain the roundness contribution degree of the corner point, and then take the mean value of the roundness contribution degrees of all corner points as the edge smoothness coefficient of the suspected abrasive particle area.

5. The machine vision-based lubrication system monitoring method according to claim 4, characterized in that, The improved contrast gain satisfies the relational expression: ; wherein, is the improved contrast gain, is the initial contrast gain, is the mean value of the detail obviousness of all suspected abrasive particle regions, is the mean value of the edge smoothness coefficient of all suspected abrasive particle regions.

6. The method for monitoring a lubrication system based on machine vision according to claim 5, characterized in that, The process of obtaining the texture distinctness factor includes: Extract features from the suspected abrasive particle area, including the mean value of the gradient amplitude and the variance of the gradient direction; Take the product of the mean value of the gradient amplitude and the variance of the gradient direction as the texture distinctness factor of the suspected abrasive particle area.

7. The machine vision-based lubrication system monitoring method according to claim 6, wherein The process of obtaining the high-frequency significance factor includes: Perform Fourier transform on the suspected abrasive particle area, shield the low-frequency area at the center of the spectrum through a mask, calculate the sum of the squares of the amplitudes of all pixels in the high-frequency area as the high-frequency energy; calculate the sum of the squares of the amplitudes of the entire spectrum as the total energy; take the ratio of the high-frequency energy to the total energy as the high-frequency significance factor of the suspected abrasive particle area.

8. The machine vision-based lubrication system monitoring method according to claim 3, characterized in that The process of obtaining the texture distinctness factor includes: Calculate the LBP value of each pixel point in the suspected abrasive particle area to obtain the LBP histogram, and take the variance of the LBP histogram as the texture distinctness factor of the suspected abrasive particle area.

9. The method for monitoring a lubrication system based on machine vision according to claim 3, wherein, The process of obtaining the edge smoothness coefficient includes: Use an edge detection algorithm to extract the edge of the suspected abrasive particle area, calculate the modulus of the edge gradient to obtain the gradient image, calculate the variance of the gradient change based on the gradient image, and take the reciprocal of the variance of the gradient change as the edge smoothness coefficient.

10. A lubrication system monitoring system based on machine vision, characterized in that, Including: A processor and a memory, and the memory stores computer program instructions, which implement the machine vision-based lubrication system monitoring method according to any one of claims 1-9 when the computer program instructions are executed by the processor.

Images