CN117952958B - Drum filter screen intelligent health detection method based on machine learning - Google Patents
Drum filter screen intelligent health detection method based on machine learning Download PDFInfo
- Publication number
- CN117952958B CN117952958B CN202410323094.XA CN202410323094A CN117952958B CN 117952958 B CN117952958 B CN 117952958B CN 202410323094 A CN202410323094 A CN 202410323094A CN 117952958 B CN117952958 B CN 117952958B
- Authority
- CN
- China
- Prior art keywords
- curve
- suspected
- interval
- value
- dark area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000036541 health Effects 0.000 title claims abstract description 44
- 238000001514 detection method Methods 0.000 title claims abstract description 39
- 238000010801 machine learning Methods 0.000 title claims abstract description 19
- 238000000034 method Methods 0.000 claims abstract description 54
- 230000008859 change Effects 0.000 claims abstract description 35
- 230000003044 adaptive effect Effects 0.000 claims abstract description 28
- 238000004080 punching Methods 0.000 claims abstract description 20
- 230000003014 reinforcing effect Effects 0.000 claims abstract description 12
- 238000001914 filtration Methods 0.000 claims description 51
- 238000003708 edge detection Methods 0.000 claims description 14
- 230000002146 bilateral effect Effects 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000007621 cluster analysis Methods 0.000 claims description 9
- 238000003064 k means clustering Methods 0.000 claims description 9
- 238000005192 partition Methods 0.000 claims description 9
- 230000002708 enhancing effect Effects 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 230000002787 reinforcement Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 25
- 238000012216 screening Methods 0.000 abstract description 2
- 230000000737 periodic effect Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000013535 sea water Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of filter image enhancement, in particular to a drum filter screen intelligent health detection method based on machine learning, which comprises the steps of screening out suspected local dark area intervals according to the fluctuation condition of a gray scale fluctuation curve corresponding to each row of pixel points in a punched mesh gray scale image; combining the gray value change rule in the real local dark area, and obtaining the suspected degree of the dark area according to the gray change trend of the pixel points in each suspected local dark area interval; combining the suspected degree of the dark area and the position of each pixel point in the suspected local dark area interval to obtain an adaptive filter window; the method has the advantages that the punching net reinforcing image with better effect is obtained by combining the self-adaptive filter window, and the follow-up accuracy of the health detection of the drum filter screen according to the punching net reinforcing image is higher.
Description
Technical Field
The invention relates to the technical field of filter image enhancement, in particular to a drum filter screen intelligent health detection method based on machine learning.
Background
The drum filter screen is one of main filtering devices of water taking heads in a water taking system of a nuclear power plant and a circulating water cooling system of a thermal power plant, the corresponding drum filter screen device consists of a rotary cylindrical frame structure, and a punching net sheet is arranged on the outer surface of the cylindrical frame and used for intercepting dirt larger than the diameter of the net holes, so that the drum filter screen is usually used for removing solid impurities in seawater and ensuring normal operation of the device. For the punched mesh, the punched mesh is usually in liquid such as seawater for a long time in the daily use process, and corrosion of the surface of the mesh can occur, so that the service life of the filter mesh is influenced; and the health condition of the punched net sheet can directly influence the effect of removing sundries, so that the punched net sheet needs to be detected regularly.
In view of detection efficiency and cost factors, the existing method for health detection of the punched meshes generally performs health detection according to the punched mesh images. However, due to the mesh-shaped surface structure of the punched mesh and the influence of external environment, more noise usually exists in the acquired image, so that in order to ensure the accuracy of health detection, filtering enhancement processing needs to be performed on the acquired punched mesh image. In the prior art, a bilateral filtering method is generally adopted to carry out filtering enhancement on the punched mesh image. The bilateral filtering method adopts filtering windows with the same size for all pixel points; however, the mesh gap of the punched mesh is usually smaller, the corresponding part of edges generating diffraction effect is not obvious on the image, if the neighborhood window of the pixel point of the part is subjected to filtering treatment by adopting a filtering window with the same size as the normal edge, the part of the unobvious edges generating diffraction effect can be filtered, so that the effective edge information of the part of the punched mesh can be lost when the punched mesh enhanced image is obtained after the filtering enhancement, namely, the effect of filtering enhancement on the punched mesh image by adopting a bilateral filtering method in the prior art is poor, the obtained punched mesh enhanced image is inaccurate, and the accuracy of the subsequent drum filter screen health detection according to the punched mesh enhanced image is lower.
Disclosure of Invention
In order to solve the technical problems that the obtained punched mesh enhanced image is inaccurate due to the fact that the effect of filtering enhancement on the punched mesh image is poor in the prior art, and the accuracy of follow-up drum filter screen health detection according to the punched mesh enhanced image is low, the invention aims to provide a drum filter screen intelligent health detection method based on machine learning, and the adopted technical scheme is as follows:
the invention provides a machine learning-based intelligent health detection method for a drum filter screen, which comprises the following steps:
Acquiring a gray value fluctuation curve of each row of pixel points in a gray image of a punching net piece of drum-shaped filter screen equipment;
Obtaining all suspected local dark area intervals in the gray value fluctuation curve according to the slope value change condition and the gray value change condition in the gray value fluctuation curve; obtaining the suspected degree of the dark area of each suspected local dark area interval according to the gray value change trend of each pixel point in each suspected local dark area interval;
In each suspected local dark region interval of the gray value fluctuation curve, according to the position of each pixel point and the suspected degree of the dark region, the size of an adaptive filter window of each pixel point in each suspected local dark region interval is obtained; according to the self-adaptive filter window and a preset priori filter window of the pixel points outside the suspected local dark region, carrying out filter enhancement on the gray level image of the punching net sheet to obtain an enhanced image of the punching net sheet;
and carrying out health detection on the drum filter screen according to the punched mesh reinforcing image.
Further, the method for acquiring the suspected local dark region interval comprises the following steps:
In the gray value fluctuation curve, a curve segment between each extreme point and the adjacent next extreme point is used as an initial sub-curve segment of each extreme point; taking the average value of the absolute values of the tangential slopes of all curve data points on each initial sub-curve segment as the reference slope value of each initial sub-curve segment;
performing cluster analysis on the reference slope values of all the initial sub-curve segments through a k-means clustering algorithm to obtain two initial sub-curve segment clusters; taking all initial sub-curve segments in the initial sub-curve segment cluster with the largest mean value of the reference slope values of all corresponding initial sub-curve segments as reference sub-curve segments;
Arranging all the reference sub-curve segments in the sequence of the gray value fluctuation curves to obtain a sequence of the reference sub-curve segments; in the sequence of reference sub-curve segments, according to the slope distribution difference and the amplitude distribution condition on the gray value fluctuation curve between each reference sub-curve segment and the adjacent next reference curve segment, obtaining the distance characteristic value of each reference sub-curve segment;
Performing cluster analysis on the distance characteristic values of all the reference sub-curve segments through a k-means clustering algorithm to obtain two reference sub-curve segment clusters; taking maximum value points of all reference sub-curve segments in the reference sub-curve segment cluster with the maximum mean value of the distance characteristic values of all corresponding reference sub-curve segments as reference interval points; dividing the gray value fluctuation curve into at least two reference division sections by taking reference interval points as intervals; taking the gray value average value of all pixel points in each reference division interval as the gray reference value of each reference division interval;
Carrying out cluster analysis on gray scale reference values of all reference partition regions through a k-means clustering algorithm to obtain at least two reference partition region clusters; and taking all the reference dividing intervals in the reference dividing interval cluster with the minimum mean value of the gray reference values of all the corresponding reference dividing intervals as suspected local dark area intervals.
Further, the method for acquiring the suspected degree of the dark area comprises the following steps:
in the gray value fluctuation curve, a curve corresponding to the suspected local dark region interval is used as a suspected dark region curve; taking the fitting curve of the suspected dark area curve as a first fitting curve of a suspected local dark area interval; taking the fitting curve of all maximum points of the suspected dark area curve as a second fitting curve of the suspected local dark area interval; taking the fitting curve of all minimum points of the suspected dark area curve as a third fitting curve of the suspected local dark area interval;
Sequentially taking the first fitting curve, the second fitting curve and the third fitting curve as target fitting curves;
obtaining a reference suspected degree of the target fitting curve according to the slope distribution symmetry trend and the fluctuation stability condition of the target fitting curve;
and obtaining the dark region suspected degree of the suspected local dark region interval according to the reference suspected degree among the first fitted curve, the second fitted curve and the third fitted curve.
Further, the calculation formula of the size of the adaptive filter window includes:
Wherein, For/>The first/>, corresponding to each suspected local dark region intervalThe size of the adaptive filter window of each pixel point; /(I)The size of the prior filtering window is preset; /(I)The size of the preset minimum filter window is smaller than that of the preset priori filter window; /(I)For/>Sequentially from left to right in each suspected locally dark region interval/>Index values of the individual pixels; /(I)For/>The number of pixels in each suspected local dark region interval; For/> A dark region suspected level for each suspected local dark region interval; /(I)A minimum value of the suspected degree of the dark region in all the suspected local dark region intervals; /(I)Is a normalization function; /(I)Is an exponential function with a natural constant as a base; /(I)Is an absolute value symbol; /(I)To round the symbol up.
Further, the method for carrying out filtering enhancement on the punched mesh gray level image is a bilateral filtering method.
Further, the method for detecting the health of the drum filter according to the punched mesh reinforcing image comprises the following steps:
Performing edge detection on the punched mesh reinforced image by an edge detection method to obtain a punched mesh edge image; and performing template matching on the edge image of the punching net sheet and the edge image of the healthy punching net sheet, and performing drum filter screen health detection according to a template matching result.
Further, the calculation formula of the distance characteristic value includes:
Wherein, For the/>, on the gray value fluctuation curveDistance characteristic values of the reference sub-curve segments; /(I)For the/>, on the gray value fluctuation curveOrdinate value of maximum point of each reference sub-curve segment, and the/>Minimum values between the ordinate values of the maximum points of the respective reference sub-curve segments; /(I)The mean value of the ordinate values of the maximum points of all the reference sub-curve segments on the gray value fluctuation curve is taken as the mean value; /(I)For the/>, on the gray value fluctuation curveReference slope value and the/>, of each reference sub-curve segmentDifferences between the reference slope values of the reference sub-curve segments; /(I)Is an exponential function with a natural constant as a base; /(I)And the preset first adjusting parameter is larger than 0.
Further, the method for acquiring the reference suspected degree comprises the following steps:
And taking a negative correlation mapping value of the product between the number of extreme points in the target fitting curve and the absolute value of the accumulated value of the tangential slope values of all curve data points as the reference suspected degree of the target fitting curve.
Further, the method for obtaining the dark region suspected degree of the suspected local dark region interval according to the reference suspected degree among the first fitted curve, the second fitted curve and the third fitted curve comprises the following steps:
taking the standard deviation of the reference suspected degree among the first fitting curve, the second fitting curve and the third fitting curve and the sum value of a preset second adjusting parameter as the suspected degree consistency of a suspected local dark area interval, wherein the preset second adjusting parameter is larger than 0;
And taking the ratio of the average value of the reference suspected degrees among the first fitted curve, the second fitted curve and the third fitted curve to the consistency of the suspected degrees as the suspected degree of the dark area in the suspected local dark area interval.
Further, the edge detection method adopts Canny edge detection.
The invention has the following beneficial effects:
The reason that the effect of filtering and enhancing the punched mesh image is poor is that the bilateral filtering method adopts the filtering window with the same size for all pixel points, the corresponding area of the part of edges on the punched mesh, which generate diffraction effect, on the image is darker, the corresponding edge textures are not obvious, when the filtering window with the same size as the normal edge is adopted for filtering and enhancing the darker area, the corresponding edge textures are excessively processed, and the edge information is filtered, so that the accuracy of the obtained punched mesh enhanced image is lower. Therefore, if the darker area is analyzed, an adaptive filtering window capable of removing noise and retaining edge information is obtained, so that the subsequent filtering enhancement effect is better; it is therefore an object of embodiments of the present invention to calculate an adaptive filter window for pixel points in a locally dark region.
To calculate the adaptive filter window of the pixel points in the local dark area, the local dark area needs to be acquired first. For the punched mesh, a large number of similar square block-shaped structures exist in the corresponding punched mesh gray image, so that the gray fluctuation of each row of pixel points in the corresponding punched mesh gray image has obvious periodic characteristics under the condition of normal noise interference, and the local dark area is darker than the normal area, so that larger gray fluctuation appears at the junction of the local dark area and the normal area, larger gray change appears on a gray value fluctuation curve, namely the corresponding periodic characteristics change at the place, and the gray value of the corresponding highlight edge of the local dark area is reduced to a certain extent due to the influence of diffraction effect, so that the gray value of the corresponding local dark area is smaller than that of the normal area; therefore, the method and the device have the characteristics that all suspected local dark area sections in the gray value fluctuation curve are obtained according to the slope value change condition and the gray value change condition in the gray value fluctuation curve.
Further, considering the feature that the darker the center of the local dark area is on the punched mesh gray level image, namely the gray level change of the pixel point on each suspected local dark area interval has a regular change trend, further suspected degree measurement can be carried out on each suspected local dark area interval according to the corresponding gray level trend change.
For all edge pixel points in the suspected local dark area, the closer the position of the edge pixel point is to the center of the local dark area, the smaller the corresponding adaptive filter window is needed to avoid the influence of surrounding pixel points on edge information; in contrast, the closer to the inner boundary of the locally dark region, the larger the corresponding adaptive neighborhood window is required to smooth abnormal pixels such as noise as much as possible. The suspected degree of the dark area can represent the possibility that each suspected local dark area interval is a real local dark area interval, so that in each suspected local dark area interval of the gray value fluctuation curve, according to the position of each pixel point and the suspected degree of the dark area, the size of a more accurate self-adaptive filter window of each pixel point in each suspected local dark area interval is obtained, namely the purpose of calculating the self-adaptive filter window of the pixel point in the local dark area according to the method, so that the effect of filtering and enhancing the gray image of the punched mesh is better by combining the more accurate self-adaptive filter window and the preset priori filter window of the pixel point outside the suspected local dark area interval, the punched mesh enhanced image with better effect is obtained, and the accuracy of the subsequent drum-shaped health detection of the filter screen according to the punched mesh enhanced image is higher.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for intelligent health detection of a drum filter based on machine learning according to an embodiment of the present invention;
FIG. 2 is a gray scale image of a punched mesh prior to reinforcement provided by one embodiment of the present invention;
FIG. 3 is a view of a punched mesh reinforcement image reinforced in combination with an adaptive filter window according to one embodiment of the present invention;
fig. 4 is a gray scale image of a punched mesh sheet reinforced by a conventional bilateral filtering method according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of a drum filter intelligent health detection method based on machine learning according to the invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the intelligent health detection method of the drum filter screen based on machine learning.
Referring to fig. 1, a flowchart of a method for intelligent health detection of a drum filter based on machine learning according to an embodiment of the invention is shown, the method includes:
Step S1: and acquiring a gray value fluctuation curve of each row of pixel points in the gray image of the punching net sheet of the drum-shaped filter screen equipment.
The embodiment of the invention aims to provide a machine learning-based intelligent health detection method for a drum filter screen, which is used for calculating an adaptive filter window of a pixel point in a suspected local dark region interval according to gray value distribution conditions in a gray image of a punched mesh, so that the effect of filtering and enhancing the gray image of the punched mesh by combining the adaptive filter window is better, and the accuracy of health detection for the drum filter screen according to the enhanced image of the punched mesh is improved.
For the punched mesh, a large number of similar square block-shaped structures exist in the corresponding punched mesh gray level image, and a certain difference exists in gray level distribution between the local dark area and the normal area, so that the gray level change of each row can be analyzed to better reflect the change in gray level distribution. Therefore, the embodiment of the invention obtains the gray value fluctuation curve of each row of pixel points in the gray image of the punching net sheet of the drum-shaped filter screen equipment. It should be noted that, the square block structure presented by the punched mesh can present a certain gray level distribution rule on both rows and columns, so that an implementer can analyze the gray level fluctuation curve of each column of pixel points in the punched mesh gray level image, and further description is omitted here.
In the embodiment of the invention, a CCD camera is used for shooting an initial image of a punching net piece of drum-shaped filter screen equipment; and in consideration of the subsequent requirement of analysis based on gray level change trend, the initial image of the punched mesh obtained through shooting is grayed, so that the gray level image of the punched mesh required by the embodiment of the invention is obtained. It should be noted that, the practitioner may also collect the initial image of the punched mesh through other image collecting devices outside the CCD camera, which will not be further described herein.
Step S2: obtaining all suspected local dark area intervals in the gray value fluctuation curve according to the slope value change condition and the gray value change condition in the gray value fluctuation curve; and obtaining the suspected degree of the dark area in each suspected local dark area interval according to the gray value change trend of each pixel point in each suspected local dark area interval.
The reason that the effect of filtering and enhancing the punched mesh image is poor is that the bilateral filtering method adopts the filtering window with the same size for all pixel points, the corresponding area of the part of edges on the punched mesh, which generate diffraction effect, on the image is darker, the corresponding edge textures are not obvious, when the filtering window with the same size as the normal edge is adopted for filtering and enhancing the darker area, the corresponding edge textures are excessively processed, and the edge information is filtered, so that the accuracy of the obtained punched mesh enhanced image is lower. Therefore, if the darker area is analyzed, an adaptive filtering window capable of removing noise and retaining edge information is obtained, so that the subsequent filtering enhancement effect is better; it is therefore an object of embodiments of the present invention to calculate an adaptive filter window for pixel points in a locally dark region.
To calculate the adaptive filter window of the pixel points in the local dark area, the local dark area needs to be acquired first. For the punched mesh, a large number of similar square block-shaped structures exist in the corresponding punched mesh gray image, so that the gray fluctuation of each row of pixel points in the corresponding punched mesh gray image has obvious periodic characteristics under the condition of normal noise interference, and the local dark area is darker than the normal area, so that larger gray fluctuation appears at the junction of the local dark area and the normal area, larger gray change appears on a gray value fluctuation curve, namely the corresponding periodic characteristics change at the place, and the gray value of the corresponding highlight edge of the local dark area is reduced to a certain extent due to the influence of diffraction effect, so that the gray value of the corresponding local dark area is smaller than that of the normal area; therefore, the method and the device have the characteristics that all suspected local dark area sections in the gray value fluctuation curve are obtained according to the slope value change condition and the gray value change condition in the gray value fluctuation curve.
Preferably, the method for acquiring the suspected local dark region interval comprises the following steps:
In the gray value fluctuation curve, a curve segment between each extreme point and the next adjacent extreme point is used as an initial sub-curve segment of each extreme point; and taking the average value of the absolute values of the tangential slopes of all curve data points on each initial sub-curve segment as the reference slope value of each initial sub-curve segment. Performing cluster analysis on the reference slope values of all the initial sub-curve segments through a k-means clustering algorithm to obtain two initial sub-curve segment clusters; and taking all initial sub-curve segments in the initial sub-curve segment cluster with the largest mean value of the reference slope values of all corresponding initial sub-curve segments as reference sub-curve segments. It should be noted that, in the embodiment of the present invention, the method for obtaining the extreme point is newton method, and further description is omitted.
Firstly, the gray value corresponding to the edge texture in the punched mesh gray image is usually larger, so that the curve segment corresponding to the part of extreme points with larger corresponding gray value represents the edge texture, and because a large number of similar square block shape structures exist in the punched mesh gray image, the distribution of corresponding edge pixel points is more regular, and larger gray change usually exists at the positions of the edge pixel points, and further, more accurate reference sub-curve segments with larger gray change are screened out through clustering analysis, so that the subsequent analysis is convenient. It should be noted that, for the reference slope value of each initial sub-curve segment, besides the average value of the absolute values of the tangential slopes of all curve data points on each initial sub-curve segment, the implementer may also use the absolute value of the slope of the connecting line between two extreme points on each initial sub-curve segment as the reference slope value, which is not further described herein.
Further, larger gray level fluctuation appears at the junction of the local dark region and the normal region, and larger gray level change appears on a gray level fluctuation curve, so that a reference sub-curve section of the junction of the local dark region and the normal region can be screened out according to the junction characteristic of the local dark region and the normal region, and the required local dark region can be further obtained according to the reference sub-curve section of the junction. According to the embodiment of the invention, all the reference sub-curve segments are arranged in the sequence of the gray value fluctuation curve to obtain a sequence of the reference sub-curve segments; in the sequence of reference sub-curve segments, according to the slope distribution difference and the amplitude distribution condition on the gray value fluctuation curve between each reference sub-curve segment and the adjacent next reference curve segment, obtaining the distance characteristic value of each reference sub-curve segment; and further screening out the reference sub-curve section of the boundary between the local dark area and the normal area according to the distance characteristic value.
Preferably, each reference sub-curve segment on the gray value fluctuation curve is taken as the first one in turnA reference sub-curve segment, the/>, on the gray value fluctuation curveThe calculation formula of the distance characteristic values of the reference sub-curve segments comprises the following steps:
Wherein, Is the/>, on the gray value fluctuation curveDistance characteristic values of the reference sub-curve segments; /(I)Is the/>, on the gray value fluctuation curveOrdinate value of maximum point of each reference sub-curve segment, and the/>Minimum values between the ordinate values of the maximum points of the respective reference sub-curve segments; /(I)The mean value of the ordinate values of the maximum value points of all the reference sub-curve segments on the gray value fluctuation curve; /(I)Is the/>, on the gray value fluctuation curveReference slope value and the/>, of each reference sub-curve segmentDifferences between the reference slope values of the reference sub-curve segments; /(I)Is an exponential function with a natural constant as a base; /(I)In order to preset the first adjustment parameter, the preset first adjustment parameter is greater than 0 and is used for preventing the denominator from being 0, and the preset first adjustment parameter is set to be 0.1 according to the embodiment of the invention, and an implementer can adjust the preset first adjustment parameter according to the specific implementation environment.
The function of the distance characteristic value is to further screen the reference sub-curve section at the junction of the local dark area and the normal area, and the larger the corresponding distance characteristic value is, the more likely the corresponding distance characteristic value is the reference sub-curve section at the junction of the local dark area and the normal area. For the local dark region, the ordinate value of the corresponding maximum value point, namely the gray value of the edge pixel point in the region reflected by the corresponding reference sub-curve segment, and the gray value of the edge pixel point of the local dark region is smaller than that of the edge pixel point of the normal region; the reference sub-curve segment at the junction of the local dark region and the normal region and the adjacent reference sub-curve segment necessarily have the reference sub-curve segment at the local dark region, and the gray value of the edge pixel point of the local dark region is smaller than that of the edge pixel point in the normal region, soThe smaller the time, the more likely the reference sub-curve segment is in the local dark region, namely the more likely the reference sub-curve segment is at the junction of the local dark region and the normal region, namely the larger the distance characteristic value is; therefore, the molecular target is subjected to negative correlation mapping to obtain a distance characteristic value; the calculation robustness is ensured when the mean value of the ordinate values of all the maximum points is subtracted from the ordinate value of the maximum point, and the practitioner can select only the ordinate value of the maximum point as a molecule according to the specific implementation environment, so that no further description is given here.
Further, in the local dark region and the normal region, the reference slope values between adjacent reference sub-curve segments are generally similar due to the existence of periodic gray scale fluctuation, and the corresponding reference slope values are obviously changed due to brightness change at the junction of the local dark region and the normal region, so that the corresponding reference slope values are correspondingThe larger the reference sub-curve section is, the more likely the reference sub-curve section is at the junction of the local dark area and the normal area, namely the larger the distance characteristic value is; thus will/>And performing negative correlation mapping after dividing the matrix to obtain a distance characteristic value.
Because the larger the distance characteristic value is, the more likely the distance characteristic value belongs to the reference sub-curve segment at the junction of the local dark region and the normal region, the embodiment of the invention performs cluster analysis on the distance characteristic values of all the reference sub-curve segments through a k-means clustering algorithm to obtain two reference sub-curve segment clusters; taking maximum value points of all reference sub-curve segments in the reference sub-curve segment cluster with the maximum mean value of the distance characteristic values of all corresponding reference sub-curve segments as reference interval points; the embodiment of the invention divides the gray value fluctuation curve into at least two reference division sections by taking the reference interval points as intervals. Namely, taking a maximum value point of a reference sub-curve section with a larger distance characteristic value, which belongs to the junction of a local dark area and a normal area, as a reference interval point, and dividing an interval; because the reference interval point is the maximum value point of the reference sub-curve section at the junction of the local dark region and the normal region, the corresponding reference dividing interval is provided with the local dark region interval and the normal region interval at the same time; because the gray value of the pixel point in the local dark area is smaller than that of the normal area due to the influence of the diffraction effect, the embodiment of the invention further takes the gray value average value of all the pixel points in each reference division interval as the gray reference value of each reference division interval. Carrying out cluster analysis on gray scale reference values of all reference partition regions through a k-means clustering algorithm to obtain at least two reference partition region clusters; and taking all the reference dividing intervals in the reference dividing interval cluster with the minimum mean value of the gray reference values of all the corresponding reference dividing intervals as suspected local dark area intervals.
Further, considering the feature that the darker the center of the local dark area is on the punched mesh gray level image, namely, the gray level change of the pixel point on each suspected local dark area has a regular change trend, further suspected degree can be measured for each suspected local dark area according to the corresponding gray level trend change.
Preferably, the method for acquiring the suspected degree of the dark area comprises the following steps:
in the gray value fluctuation curves, a curve corresponding to a suspected local dark area interval is used as a suspected dark area curve; taking the fitting curve of the suspected dark area curve as a first fitting curve of a suspected local dark area interval; taking the fitting curve of all maximum points of the suspected dark area curve as a second fitting curve of the suspected local dark area interval; taking the fitting curve of all minimum points of the suspected dark area curve as a third fitting curve of the suspected local dark area interval;
Sequentially taking the first fitting curve, the second fitting curve and the third fitting curve as target fitting curves;
First, for a locally dark region due to diffraction effects, its corresponding overall gray scale characteristics, i.e. darkness distribution, are: the closer to the boundary of the local dark area, the higher the corresponding gray value, the closer to the center of the local dark area, and the lower the corresponding gray value; and periodic gray scale fluctuations also exist in locally dark areas; therefore, the gray value changes corresponding to the first fit curve, the second fit curve and the third fit curve all show symmetrical characteristics, namely the characteristics that the gray value is changed from high to low to high, and the corresponding slope distribution is generally symmetrical; accordingly, since the corresponding gray scale is changed from high to low to high, the number of extreme points of the corresponding fitting curve is usually small, and the corresponding fitting curve fluctuates more stably. According to the embodiment of the invention, the reference suspected degree of the target fitting curve is obtained according to the symmetrical trend and the fluctuation stability of the slope distribution of the target fitting curve.
Preferably, the method for acquiring the reference suspected degree includes:
And taking a negative correlation mapping value of the product between the number of extreme points in the target fitting curve and the absolute value of the accumulated value of the tangential slope values of all curve data points as the reference suspected degree of the target fitting curve. The number of extreme points of the fitting curve corresponding to the real local dark area is small; thus, when the number of extreme points is larger, the local dark area is more likely to be avoided; the slope distribution of the local dark region is generally symmetrical, so the accumulated value of the tangential slope values of the corresponding curve data points is generally close to 0, and thus the smaller the absolute value of the accumulated value of the tangential slope values of all curve data points, the more likely the local dark region is in the embodiment of the present invention; namely, the number of extreme points in the target fitting curve and the absolute value of the accumulated value of the tangential slope values of all curve data points are in negative correlation with the reference suspected degree; it should be noted that, when no extreme point exists in the target fitting curve, the gray level change characteristics which do not conform to the local dark area are considered, so that the parameters of the number of the corresponding extreme points are set to 3, the calculated reference suspected degree is more reasonable, and the implementer can also adjust according to the specific implementation environment.
In the embodiment of the invention, each suspected local dark region interval is taken as the firstThe suspected local dark region is the/>Target fitting curve/>, of each suspected local dark region intervalThe method for obtaining the reference suspected level of (2) is expressed as follows in terms of formula:
Wherein, For/>Target fitting curve/>, of each suspected local dark region intervalIs referred to as the suspected level,/>For/>Target fitting curve/>, of each suspected local dark region intervalExtreme point number of (2); /(I)For/>Target fitting curve/>, of each suspected local dark region intervalNumber of curve data points on; /(I)For/>Target fitting curve/>, of each suspected local dark region intervalCurve data points on/>Tangential slope values for the individual curve data points; /(I)Is an absolute value symbol; /(I)Is an exponential function with a base of natural constant.
And further combining the reference suspected degrees of the first fitting curve, the second fitting curve and the third fitting curve, and obtaining the suspected degree of the dark area of the suspected local dark area interval according to the reference suspected degrees among the first fitting curve, the second fitting curve and the third fitting curve.
Preferably, the method for obtaining the dark region suspected degree of the suspected local dark region interval according to the reference suspected degree among the first fitted curve, the second fitted curve and the third fitted curve comprises the following steps:
The standard deviation of the reference suspected degree among the first fitting curve, the second fitting curve and the third fitting curve is used as the suspected degree consistency of the suspected local dark region interval with the sum value of the preset second adjusting parameter; presetting a second adjusting parameter to be larger than 0, wherein the second adjusting parameter is used for preventing a denominator in subsequent calculation from being 0; in the embodiment of the invention, the preset second adjusting parameter is set to be 0.1, and an implementer can adjust the preset second adjusting parameter according to the specific implementation environment. Firstly, for a local dark area, due to the structural characteristics of a punched mesh, the corresponding gray level variation has periodical and regular gray level fluctuation, so that the symmetrical trend of slope distribution among the corresponding first fitting curve, the second fitting curve and the third fitting curve has higher consistency with fluctuation stability, and the probability that the corresponding suspected local dark area interval is a real local dark area is higher when the corresponding reference suspected degree is closer; calculating to obtain the suspected degree consistency through the standard deviation of the reference suspected degree among the first fitted curve, the second fitted curve and the third fitted curve; the smaller the correspondence of the suspected degree, the higher the final dark region is suspected.
Further, when the reference suspected degree of the target fitting curve is larger, the target fitting curve is more likely to be a local dark region; the larger the corresponding first, second and third fitted curves are, the greater the corresponding dark region is suspected; therefore, the embodiment of the invention takes the ratio of the average value of the reference suspected degrees among the first fitted curve, the second fitted curve and the third fitted curve to the consistency of the suspected degrees as the suspected degree of the dark region in the suspected local dark region interval.
In an embodiment of the invention, the firstThe method for obtaining the suspected degree of the dark area in each suspected local dark area interval is expressed as the following formula:
Wherein, For/>Dark region suspected level of each suspected local dark region interval,/>For/>The average value of reference suspected degrees among the first fitted curve, the second fitted curve and the third fitted curve corresponding to the suspected local dark region interval; /(I)For/>Standard deviations of reference suspected degrees among the first fitted curve, the second fitted curve and the third fitted curve corresponding to the suspected local dark region intervals; /(I)Presetting a second adjusting parameter; /(I)For/>The suspected degree consistency of each suspected locally dark region interval.
Step S3: in each suspected local dark region interval of the gray value fluctuation curve, according to the position of each pixel point and the suspected degree of the dark region, the size of the self-adaptive filter window of each pixel point in each suspected local dark region interval is obtained; and carrying out filtering enhancement on the gray level image of the punched mesh according to the self-adaptive filtering window and a preset priori filtering window of the pixel points outside the suspected local dark region interval to obtain a punched mesh enhanced image.
For all edge pixel points in the suspected local dark area, the closer the position of the edge pixel point is to the center of the local dark area, the smaller the corresponding adaptive filter window is needed to avoid the influence of surrounding pixel points on edge information; in contrast, the closer to the inner boundary of the locally dark region, the larger the corresponding adaptive neighborhood window is required to smooth abnormal pixels such as noise as much as possible. The suspected degree of the dark area can represent the possibility that each suspected local dark area interval is a real local dark area interval, so that in each suspected local dark area interval of the gray value fluctuation curve, the size of the self-adaptive filter window of each pixel point in each suspected local dark area interval is obtained according to the position of each pixel point and the suspected degree of the dark area, namely the purpose of calculating the self-adaptive filter window of the pixel point in the local dark area.
Preferably, the calculation formula of the size of the adaptive filter window includes:
Wherein, For/>The first/>, corresponding to each suspected local dark region intervalThe size of the adaptive filter window of each pixel point; /(I)The size of a priori filtering window is preset; /(I)The size of the preset minimum filter window is smaller than that of the preset priori filter window; /(I)For/>Sequentially from left to right in each suspected locally dark region interval/>Index values of the individual pixels; /(I)For/>The number of pixels in each suspected local dark region interval; /(I)For/>A dark region suspected level for each suspected local dark region interval; /(I)A minimum value of the suspected degree of the dark region in all the suspected local dark region intervals; /(I)Is a normalization function; /(I)Is an exponential function with a natural constant as a base; /(I)Is an absolute value symbol; /(I)To round the symbol up. In the embodiment of the invention, the size of the preset priori filtering window is set to be 9, namely a filtering window with the size of 9 multiplied by 9, namely the filtering window of each pixel point in the bilateral filtering algorithm is 9 multiplied by 9 under the condition that the size of the adaptive filtering window is not calculated; the size of the preset minimum filter window is set to be 3, namely, a filter window with the size of 3 multiplied by 3; the practitioner can select the preset priori filtering window and the preset minimum filtering window according to the specific implementation environment, and the details are not described here.
In a calculation formula of the size of the self-adaptive filter window, firstly, according to the characteristics of edge pixel points in a suspected local dark area, the closer the corresponding pixel points are to the center of the local dark area, the smaller the size of the corresponding self-adaptive filter window is required; thus (2)Can characterize the/>Index values of the centremost pixel points in the suspected local dark region sections; thus in/>Index value and/>, of each pixel point in sequence from left to right in each suspected local dark region intervalThe smaller the difference is, the closer the indication is to the center of the corresponding suspected local dark region interval, namely the smaller the corresponding adaptive filter window is required; considering that the suspected degrees of the dark areas in different suspected local dark area intervals are different, the greater the suspected degrees of the corresponding dark areas are, the greater the possibility that the corresponding suspected local dark area intervals belong to real local dark areas is, and the smaller the filter window of the real local dark areas is needed to prevent the effective textures from being filtered; thus by the pair of dark region suspected degrees between suspected local dark regions/>And weighting is carried out, so that the adaptive filter window calculated later is more accurate. Further pair/>And carrying out negative correlation mapping so that the calculated size of the adaptive filter window is between the size of the preset minimum filter window and the size of the preset priori filter window. It should be noted that, according to the specific implementation environment, the implementer may also calculate the index value of each pixel point in the order from right to left, which has no influence on the calculation of the degree of approach to the center, and will not be further described herein.
After the self-adaptive filter window size of the pixel points in the suspected local dark region is obtained, further carrying out filter enhancement on the punched mesh gray level image. Preferably, the method for filtering and enhancing the gray level image of the punched mesh is a bilateral filtering method.
FIG. 2 illustrates a pre-reinforcement punched mesh gray scale image provided by one embodiment of the present invention; FIG. 3 illustrates a punched mesh reinforcement image reinforced in combination with an adaptive filter window, according to one embodiment of the present invention; fig. 4 shows a punched mesh gray scale image reinforced by a conventional bilateral filtering method according to an embodiment of the present invention. Firstly, before enhancement, partial edge textures are not obvious due to the influence of diffraction effect, and the enhancement image of the punched mesh corresponding to the image enhancement method of the embodiment of the invention is obtained after enhancement, so that the edge textures influenced by the diffraction effect are reserved while noise influence is eliminated; in fig. 4, in an image obtained by adopting a filtering window with the same size for all pixel points by a traditional bilateral filtering method, part of edge information is not clear enough; therefore, the image enhancement effect corresponding to the embodiment of the invention is better.
Step S4: and carrying out health detection on the drum filter screen according to the punched mesh reinforcing image.
After the punched mesh reinforcing image is obtained, the health detection of the drum filter screen is further carried out according to the punched mesh reinforcing image. Preferably, the method for detecting the health of the drum filter according to the punched mesh enhanced image comprises the following steps:
Performing edge detection on the punched mesh reinforced image by an edge detection method to obtain a punched mesh edge image; template matching is carried out on the edge image of the punching net sheet and the edge image of the healthy punching net sheet, and health detection of the drum filter screen is carried out according to the template matching result; the higher the matching degree in the corresponding template matching result, the higher the health degree of the drum-shaped equipment is. Preferably, the edge detection method adopts Canny edge detection, an implementer can also adopt other edge detection methods according to specific implementation environments, and the purpose of acquiring an edge image through the edge detection is to facilitate subsequent template matching. It should be noted that, the practitioner can also perform the health detection of the drum filter screen according to the reinforcing image of the punched mesh by other methods according to the specific implementation environment; for example, the punched mesh reinforcing image is input into a trained convolutional neural network, and the health degree of the drum filter screen is output; therefore, the health detection of the drum filter is performed according to the health degree of the drum filter, and further description is omitted herein.
In summary, according to the method, firstly, a suspected local dark region interval is screened out according to the fluctuation condition of a gray scale fluctuation curve corresponding to each row of pixel points in a punched mesh gray scale image; combining the gray value change rule in the real local dark area, and obtaining the suspected degree of the dark area according to the gray change trend of the pixel points in each suspected local dark area interval; combining the suspected degree of the dark area and the position of each pixel point in the suspected local dark area interval to obtain an adaptive filter window; the method has the advantages that the punching net reinforcing image with better effect is obtained by combining the self-adaptive filter window, and the follow-up accuracy of the health detection of the drum filter screen according to the punching net reinforcing image is higher.
It should be noted that: the sequence of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments.
Claims (8)
1. The intelligent health detection method for the drum filter screen based on machine learning is characterized by comprising the following steps of:
Acquiring a gray value fluctuation curve of each row of pixel points in a gray image of a punching net piece of drum-shaped filter screen equipment;
Obtaining all suspected local dark area intervals in the gray value fluctuation curve according to the slope value change condition and the gray value change condition in the gray value fluctuation curve; obtaining the suspected degree of the dark area of each suspected local dark area interval according to the gray value change trend of each pixel point in each suspected local dark area interval;
In each suspected local dark region interval of the gray value fluctuation curve, according to the position of each pixel point and the suspected degree of the dark region, the size of an adaptive filter window of each pixel point in each suspected local dark region interval is obtained; according to the self-adaptive filter window and a preset priori filter window of the pixel points outside the suspected local dark region, carrying out filter enhancement on the gray level image of the punching net sheet to obtain an enhanced image of the punching net sheet;
performing drum filter screen health detection according to the punched mesh reinforcing image;
the method for acquiring the suspected local dark region interval comprises the following steps:
In the gray value fluctuation curve, a curve segment between each extreme point and the adjacent next extreme point is used as an initial sub-curve segment of each extreme point; taking the average value of the absolute values of the tangential slopes of all curve data points on each initial sub-curve segment as the reference slope value of each initial sub-curve segment;
performing cluster analysis on the reference slope values of all the initial sub-curve segments through a k-means clustering algorithm to obtain two initial sub-curve segment clusters; taking all initial sub-curve segments in the initial sub-curve segment cluster with the largest mean value of the reference slope values of all corresponding initial sub-curve segments as reference sub-curve segments;
Arranging all the reference sub-curve segments in the sequence of the gray value fluctuation curves to obtain a sequence of the reference sub-curve segments; in the sequence of reference sub-curve segments, according to the slope distribution difference and the amplitude distribution condition on the gray value fluctuation curve between each reference sub-curve segment and the adjacent next reference curve segment, obtaining the distance characteristic value of each reference sub-curve segment;
Performing cluster analysis on the distance characteristic values of all the reference sub-curve segments through a k-means clustering algorithm to obtain two reference sub-curve segment clusters; taking maximum value points of all reference sub-curve segments in the reference sub-curve segment cluster with the maximum mean value of the distance characteristic values of all corresponding reference sub-curve segments as reference interval points; dividing the gray value fluctuation curve into at least two reference division sections by taking reference interval points as intervals; taking the gray value average value of all pixel points in each reference division interval as the gray reference value of each reference division interval;
Carrying out cluster analysis on gray scale reference values of all reference partition regions through a k-means clustering algorithm to obtain at least two reference partition region clusters; taking all reference partition sections in the reference partition section cluster with the minimum mean value of gray reference values of all corresponding reference partition sections as suspected local dark area sections;
The calculation formula of the size of the adaptive filter window comprises:
Wherein, For/>The first/>, corresponding to each suspected local dark region intervalThe size of the adaptive filter window of each pixel point; /(I)The size of the prior filtering window is preset; /(I)The size of the preset minimum filter window is smaller than that of the preset priori filter window; /(I)For/>Sequentially from left to right in each suspected locally dark region interval/>Index values of the individual pixels; /(I)For/>The number of pixels in each suspected local dark region interval; For/> A dark region suspected level for each suspected local dark region interval; /(I)A minimum value of the suspected degree of the dark region in all the suspected local dark region intervals; /(I)Is a normalization function; /(I)Is an exponential function with a natural constant as a base; /(I)Is an absolute value symbol; /(I)To round the symbol up.
2. The machine learning based intelligent health detection method of a drum screen according to claim 1, wherein the method for obtaining the suspected degree of the dark area comprises:
in the gray value fluctuation curve, a curve corresponding to the suspected local dark region interval is used as a suspected dark region curve; taking the fitting curve of the suspected dark area curve as a first fitting curve of a suspected local dark area interval; taking the fitting curve of all maximum points of the suspected dark area curve as a second fitting curve of the suspected local dark area interval; taking the fitting curve of all minimum points of the suspected dark area curve as a third fitting curve of the suspected local dark area interval;
Sequentially taking the first fitting curve, the second fitting curve and the third fitting curve as target fitting curves;
obtaining a reference suspected degree of the target fitting curve according to the slope distribution symmetry trend and the fluctuation stability condition of the target fitting curve;
and obtaining the dark region suspected degree of the suspected local dark region interval according to the reference suspected degree among the first fitted curve, the second fitted curve and the third fitted curve.
3. The machine learning based intelligent health detection method of a drum filter according to claim 1, wherein the method for filtering and enhancing the gray level image of the punched mesh is a bilateral filtering method.
4. The machine learning based intelligent health detection method of a drum screen of claim 1, wherein the method for performing drum screen health detection based on the punched mesh reinforcement image comprises:
Performing edge detection on the punched mesh reinforced image by an edge detection method to obtain a punched mesh edge image; and performing template matching on the edge image of the punching net sheet and the edge image of the healthy punching net sheet, and performing drum filter screen health detection according to a template matching result.
5. The machine learning based intelligent health detection method of a drum filter according to claim 1, wherein the calculation formula of the distance characteristic value comprises:
Wherein, For the/>, on the gray value fluctuation curveDistance characteristic values of the reference sub-curve segments; /(I)For the/>, on the gray value fluctuation curveOrdinate value and the/>, of the maximum point of each reference sub-curve segmentMinimum values between the ordinate values of the maximum points of the respective reference sub-curve segments; /(I)The mean value of the ordinate values of the maximum points of all the reference sub-curve segments on the gray value fluctuation curve is taken as the mean value; /(I)For the/>, on the gray value fluctuation curveReference slope value and the/>, of each reference sub-curve segmentDifferences between the reference slope values of the reference sub-curve segments; /(I)Is an exponential function with a natural constant as a base; /(I)And the preset first adjusting parameter is larger than 0.
6. The machine learning based intelligent health detection method of a drum filter according to claim 2, wherein the method for obtaining the reference plausibility level comprises:
And taking a negative correlation mapping value of the product between the number of extreme points in the target fitting curve and the absolute value of the accumulated value of the tangential slope values of all curve data points as the reference suspected degree of the target fitting curve.
7. The machine learning based drum screen intelligent health detection method of claim 2, wherein the method for obtaining the dark region plausibility level of the plausible local dark region interval according to the reference plausibility level between the first fitted curve, the second fitted curve and the third fitted curve comprises:
taking the standard deviation of the reference suspected degree among the first fitting curve, the second fitting curve and the third fitting curve and the sum value of a preset second adjusting parameter as the suspected degree consistency of a suspected local dark area interval, wherein the preset second adjusting parameter is larger than 0;
And taking the ratio of the average value of the reference suspected degrees among the first fitted curve, the second fitted curve and the third fitted curve to the consistency of the suspected degrees as the suspected degree of the dark area in the suspected local dark area interval.
8. The machine learning based intelligent health detection method of a drum screen of claim 4, wherein the edge detection method employs Canny edge detection.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410323094.XA CN117952958B (en) | 2024-03-21 | 2024-03-21 | Drum filter screen intelligent health detection method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410323094.XA CN117952958B (en) | 2024-03-21 | 2024-03-21 | Drum filter screen intelligent health detection method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117952958A CN117952958A (en) | 2024-04-30 |
| CN117952958B true CN117952958B (en) | 2024-06-07 |
Family
ID=90794589
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410323094.XA Active CN117952958B (en) | 2024-03-21 | 2024-03-21 | Drum filter screen intelligent health detection method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117952958B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1489557A2 (en) * | 2003-06-10 | 2004-12-22 | hema electronic GmbH | Method for adaptive edge detection |
| CN109767439A (en) * | 2019-01-10 | 2019-05-17 | 中国科学院上海技术物理研究所 | A kind of multiple dimensioned difference of self-adapting window and the object detection method of bilateral filtering |
| CN114842017A (en) * | 2022-07-06 | 2022-08-02 | 济宁市海富电子科技有限公司 | HDMI cable surface quality detection method and system |
| CN116453029A (en) * | 2023-06-16 | 2023-07-18 | 济南东庆软件技术有限公司 | Building fire environment detection method based on image data |
| CN116580024A (en) * | 2023-07-12 | 2023-08-11 | 山东荣信集团有限公司 | Coke quality detection method based on image processing |
| CN117094914A (en) * | 2023-10-18 | 2023-11-21 | 广东申创光电科技有限公司 | Smart city road monitoring system based on computer vision |
| CN117574061A (en) * | 2024-01-16 | 2024-02-20 | 暨南大学 | PM2.5 and ozone pollution cooperative prevention and control prediction method and system |
-
2024
- 2024-03-21 CN CN202410323094.XA patent/CN117952958B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1489557A2 (en) * | 2003-06-10 | 2004-12-22 | hema electronic GmbH | Method for adaptive edge detection |
| CN109767439A (en) * | 2019-01-10 | 2019-05-17 | 中国科学院上海技术物理研究所 | A kind of multiple dimensioned difference of self-adapting window and the object detection method of bilateral filtering |
| CN114842017A (en) * | 2022-07-06 | 2022-08-02 | 济宁市海富电子科技有限公司 | HDMI cable surface quality detection method and system |
| CN116453029A (en) * | 2023-06-16 | 2023-07-18 | 济南东庆软件技术有限公司 | Building fire environment detection method based on image data |
| CN116580024A (en) * | 2023-07-12 | 2023-08-11 | 山东荣信集团有限公司 | Coke quality detection method based on image processing |
| CN117094914A (en) * | 2023-10-18 | 2023-11-21 | 广东申创光电科技有限公司 | Smart city road monitoring system based on computer vision |
| CN117574061A (en) * | 2024-01-16 | 2024-02-20 | 暨南大学 | PM2.5 and ozone pollution cooperative prevention and control prediction method and system |
Non-Patent Citations (3)
| Title |
|---|
| 一种快速的自适应图像增强算法;王小明;黄昶;周晨辰;刘锦高;;计算机***应用;20100815(第08期);全文 * |
| 基于加权引导滤波的快速自适应图像去雾;代维佳;马国亮;雷帮军;雷柏超;;信息通信;20190415(第04期);全文 * |
| 结构保持的双边滤波极化SAR图像降噪;杨学志;叶铭;吴克伟;郎文辉;郑鑫;李国强;;电子与信息学报;20150215(第02期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117952958A (en) | 2024-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115661147B (en) | Metering detection data identification method based on machine vision | |
| CN116091499B (en) | Abnormal paint production identification system | |
| CN116188462B (en) | Noble metal quality detection method and system based on visual identification | |
| CN115661669B (en) | Method and system for monitoring illegal farmland occupancy based on video monitoring | |
| CN109211918B (en) | Fabric bow weft detection method based on weft trend | |
| CN116977358B (en) | Visual auxiliary detection method for corrugated paper production quality | |
| CN109961432A (en) | A kind of detection method and system of filter cloth breakage | |
| CN117665935B (en) | Monitoring data processing method for broken rock mass supporting construction process | |
| CN115797473B (en) | Concrete forming evaluation method for civil engineering | |
| CN115018835B (en) | Automobile starter gear detection method | |
| CN116805312B (en) | Knitted fabric quality detection method based on image processing | |
| CN116883412B (en) | Graphene far infrared electric heating equipment fault detection method | |
| CN116757972A (en) | Fabric defect detection method capable of resisting influence of shadow noise | |
| CN115049671A (en) | Cloth surface defect detection method and system based on computer vision | |
| CN117576104B (en) | Visual detection method for health state of ultrafiltration membrane in purification process | |
| CN117876971B (en) | Building construction safety monitoring and early warning method based on machine vision | |
| CN116703251A (en) | Rubber ring production quality detection method based on artificial intelligence | |
| CN117952958B (en) | Drum filter screen intelligent health detection method based on machine learning | |
| CN116188510B (en) | Enterprise emission data acquisition system based on multiple sensors | |
| CN115841731B (en) | Infrared monitoring park fire early warning method | |
| CN116740579A (en) | Intelligent collection method for territorial space planning data | |
| CN115862006A (en) | Method for detecting bran stars in flour milling process | |
| JP2022184321A (en) | Smoke detection device | |
| CN117495862B (en) | Denture wearability detection device | |
| CN117911956B (en) | Dynamic monitoring method and system for processing environment of food processing equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |