CN115953774B - Alarm display digital recognition method based on machine vision - Google Patents
Alarm display digital recognition method based on machine vision Download PDFInfo
- Publication number
- CN115953774B CN115953774B CN202310214709.0A CN202310214709A CN115953774B CN 115953774 B CN115953774 B CN 115953774B CN 202310214709 A CN202310214709 A CN 202310214709A CN 115953774 B CN115953774 B CN 115953774B
- Authority
- CN
- China
- Prior art keywords
- target
- edge pixel
- pixel point
- cluster
- noise
- 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
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 40
- 230000008859 change Effects 0.000 claims abstract description 39
- 238000004458 analytical method Methods 0.000 claims abstract description 35
- 238000003708 edge detection Methods 0.000 claims abstract description 15
- 230000002776 aggregation Effects 0.000 claims abstract description 13
- 238000004220 aggregation Methods 0.000 claims abstract description 13
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 5
- 230000000875 corresponding effect Effects 0.000 description 187
- 238000001914 filtration Methods 0.000 description 10
- 239000007789 gas Substances 0.000 description 10
- 239000004973 liquid crystal related substance Substances 0.000 description 10
- 238000012544 monitoring process Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 7
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 description 6
- 230000001788 irregular Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 5
- 238000005065 mining Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 229910002092 carbon dioxide Inorganic materials 0.000 description 3
- 239000001569 carbon dioxide Substances 0.000 description 3
- 239000000758 substrate Substances 0.000 description 3
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 229910002091 carbon monoxide Inorganic materials 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013524 data verification Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000002834 transmittance Methods 0.000 description 1
Images
Abstract
The invention relates to the technical field of image data processing, in particular to a machine vision-based alarm display digital identification method, which comprises the following steps: acquiring a target surface image of a display panel of the alarm, and carrying out digital region identification processing on the target surface image; performing edge detection on the target digital region; clustering the target edge pixel points in the target edge pixel point set; performing noise aggregation influence analysis processing on the target cluster; carrying out noise distribution change analysis processing on the target cluster; carrying out noise size analysis processing on the target cluster; determining the influence degree of target noise corresponding to the target edge pixel points; and carrying out self-adaptive window denoising on the target surface image, and carrying out digital recognition on the target denoising image. The invention improves the accuracy of identifying the display number of the alarm by processing the image data of the target surface image, and is mainly applied to identifying the display number of the alarm.
Description
Technical Field
The invention relates to the technical field of image data processing, in particular to a machine vision-based alarm display digital identification method.
Background
Because the alarm usually carries out safety precaution through the display number on the alarm, the accurate identification of the display number is crucial. At present, when the display number of the alarm is identified, the following modes are generally adopted: and acquiring an image of the display number of the alarm, and identifying the display number of the alarm through the image. Because the environment that the alarm is located is often comparatively complicated, often lead to the image that gathers to contain more noise, often lead to the image quality of gathering not high, and then lead to the degree of accuracy that carries out discernment to the alarm demonstration number to be low, consequently, carry out the denoising to the image that gathers and be crucial.
At present, when denoising an image, the following methods are generally adopted: and filtering and denoising the image through a filtering window with a fixed size. However, when filtering and denoising an image using a filter window with a fixed size, there are often the following technical problems:
because the size of the filtering window is fixed, the denoising and the image detail protection cannot be simultaneously considered, the denoising effect of the image is low, and the accuracy of identifying the display number of the alarm is low.
Disclosure of Invention
The summary of the invention is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. The summary of the invention is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
In order to solve the technical problem of low accuracy in identifying the display numbers of the alarm, the invention provides a machine vision-based alarm display number identification method.
The invention provides a machine vision-based alarm display digital identification method, which comprises the following steps:
acquiring a target surface image of a display panel of an alarm, and carrying out digital region identification processing on the target surface image to obtain a target digital region;
performing edge detection on the target digital region to obtain a target edge pixel point set;
clustering the target edge pixel points in the target edge pixel point set to obtain a target cluster set;
performing noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster;
According to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters;
performing noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster;
determining a target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set;
and carrying out self-adaptive window denoising on the target surface image according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image, and carrying out digital recognition on the target denoising image.
Further, performing noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster, including:
And determining the initial noise influence degree corresponding to the target cluster according to the positions corresponding to the target edge pixel points in the target cluster, the positions corresponding to the cluster centers of the target cluster, the number of the target edge pixel points in the target cluster set and the number of the target edge pixel points in the target cluster.
Further, determining the initial noise influence degree corresponding to the target cluster according to the position corresponding to each target edge pixel point in the target cluster, the position corresponding to the cluster center of the target cluster, the number of target edge pixel points in the target cluster set, and the number of target edge pixel points in the target cluster, including:
determining the distance between each target edge pixel point in the target cluster and the clustering center of the target cluster according to the position corresponding to each target edge pixel point in the target cluster and the position corresponding to the clustering center of the target cluster, and taking the distance as the target distance corresponding to the target edge pixel point;
determining the average value of the target distances corresponding to all the target edge pixel points in the target cluster as the average value of the target distances corresponding to the target cluster;
Performing negative correlation on the target distance average value corresponding to the target cluster, and normalizing to obtain a distance index corresponding to the target cluster;
determining the ratio of the number of target edge pixel points in the target cluster to the number of target edge pixel points in the target cluster set as a number index corresponding to the target cluster;
and determining the product of the distance index and the quantity index corresponding to the target cluster as the initial noise influence degree corresponding to the target cluster.
Further, according to the initial noise influence degree corresponding to each target cluster in the target cluster set, performing noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters, including:
determining gradient amplitude values corresponding to all pixel points in a preset target sliding window corresponding to the target edge pixel points;
determining the average value of the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points as the target gradient average value corresponding to the target edge pixel points;
Screening out the minimum target gradient mean value from target gradient mean values corresponding to all target edge pixel points in the target cluster, and taking the minimum target gradient mean value as the minimum gradient mean value corresponding to the target cluster;
determining a difference value between a target gradient mean value corresponding to the target edge pixel point and a first gradient mean value as a target difference value corresponding to the target edge pixel point, wherein the first gradient mean value is a minimum gradient mean value corresponding to a target cluster in which the target edge pixel point is located;
and determining a product of a target difference value corresponding to the target edge pixel point and a first influence degree as a noise change index corresponding to the target edge pixel point, wherein the first influence degree is an initial noise influence degree corresponding to a target cluster where the target edge pixel point is located.
Further, performing noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel in the target cluster includes:
screening out the minimum gradient amplitude value from the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points, and taking the minimum gradient amplitude value as the minimum gradient amplitude value corresponding to the target edge pixel points;
For each pixel point in a target sliding window corresponding to the target edge pixel point, determining a difference value between the gradient amplitude corresponding to the pixel point and the minimum gradient amplitude corresponding to the target edge pixel point as a target gradient difference value corresponding to the pixel point;
and determining the average value of target gradient difference values corresponding to all pixel points in the target sliding window corresponding to the target edge pixel points as a noise size index corresponding to the target edge pixel points.
Further, determining, according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set, a target noise influence degree corresponding to the target edge pixel point includes:
and determining the product of the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set as the target noise influence degree corresponding to the target edge pixel point.
Further, according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set, performing adaptive window denoising on the target surface image to obtain a target denoising image, including:
Determining the size of a denoising window corresponding to each target edge pixel point in the target edge pixel point set according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set;
and denoising the target edge pixel points and a preset target neighborhood corresponding to the target edge pixel points according to the size of a denoising window corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image.
Further, determining, according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set, a size of a denoising window corresponding to the target edge pixel point includes:
determining the product of the target noise influence degree corresponding to the target edge pixel point and a preset target multiple as a first product corresponding to the target edge pixel point;
and carrying out upward rounding on the first product corresponding to the target edge pixel point to obtain the size of the denoising window corresponding to the target edge pixel point.
The invention has the following beneficial effects:
according to the machine vision-based alarm display digital identification method, the target surface image is subjected to image data processing, so that denoising identification of the target surface image is realized, the technical problem of low accuracy of identifying the alarm display digital is solved, and the accuracy of identifying the alarm display digital is improved. Firstly, acquiring a target surface image of an alarm display panel, and carrying out digital region identification processing on the target surface image to obtain a target digital region. The method can be convenient for subsequent analysis of noise conditions in the target digital region and can be convenient for subsequent accurate denoising of the target digital region. And then, carrying out edge detection on the target digital region to obtain a target edge pixel point set. Because noise often affects the distribution of edge pixels, edge detection results in a set of target edge pixels, which can facilitate subsequent analysis of noise contained in a target surface image. And then, clustering the target edge pixel points in the target edge pixel point set to obtain a target cluster set. The classification of the target edge pixel points can be realized, and the noise analysis of each target cluster can be conveniently carried out later. And continuing to analyze and process the noise aggregation influence on each target cluster in the target cluster set to obtain the initial noise influence degree corresponding to the target cluster. And the noise aggregation influence analysis processing is carried out on the target cluster, so that the accuracy of determining the initial noise influence degree can be improved. And then, according to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters. The initial noise influence degree and the noise distribution change are comprehensively considered, the accuracy of determining the noise change index can be improved, and follow-up accurate denoising can be facilitated. And then, carrying out noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster. The noise size of the target cluster is considered, so that follow-up accurate denoising can be facilitated. And determining the target noise influence degree corresponding to the target edge pixel point according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set. And the noise change index and the noise size index are comprehensively considered, so that the accuracy of determining the influence degree of the target noise is improved. And finally, carrying out self-adaptive window denoising on the target surface image according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image, and carrying out digital recognition on the target denoising image. According to the invention, the image data processing is carried out on the target surface image, so that the self-adaptive window denoising identification of the target surface image is realized, the denoising and the image detail protection can be realized, the technical problems of low image denoising effect and low accuracy of identifying the display number of the alarm are solved, and the image denoising effect and the accuracy of identifying the display number of the alarm are improved.
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 flow chart of a machine vision based alarm display digital identification method according to the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purpose, the following detailed description is given below of the specific implementation, structure, features and effects of the technical solution according to the present 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 invention provides a machine vision-based alarm display digital identification method, which comprises the following steps:
acquiring a target surface image of a display panel of the alarm, and carrying out digital region identification processing on the target surface image to obtain a target digital region;
edge detection is carried out on the target digital region, and a target edge pixel point set is obtained;
clustering the target edge pixel points in the target edge pixel point set to obtain a target cluster set;
performing noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster;
according to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining noise change indexes corresponding to each target edge pixel point in the target clusters;
performing noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster;
determining the target noise influence degree corresponding to the target edge pixel points according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set;
And carrying out self-adaptive window denoising on the target surface image according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image, and carrying out digital recognition on the target denoising image.
The following detailed development of each step is performed:
referring to FIG. 1, a flow chart of some embodiments of a machine vision based alarm display digital identification method in accordance with the present invention is shown. The machine vision-based alarm display digital identification method comprises the following steps:
step S1, obtaining a target surface image of a display panel of the alarm, and carrying out digital region identification processing on the target surface image to obtain a target digital region.
In some embodiments, a target surface image of the display panel of the alarm may be acquired, and the target surface image may be subjected to a digital region identification process to obtain a target digital region.
The alarm display panel can be a panel for displaying monitoring data on the target alarm. The target alarm may be a mining gas alarm. The monitoring data may be display data of an alarm to be monitored. For example, the monitoring data of the carbon dioxide alarm may be the carbon dioxide concentration. The target surface image may be an image of the surface of an alarm display panel. The target digital zone may be the zone in which the monitoring data is located.
It should be noted that mine gas alarms are commonly used in mine applications. The mine gas alarm mainly comprises two kinds, namely a fixed mine gas alarm and a portable mine gas alarm, and can monitor oxygen, gas, carbon dioxide and the like in a mine at any time. The image of the display panel of the mining gas alarm can be acquired, monitoring data in the image is identified, and when the monitoring data is not in a normal range, alarm prompt is carried out. For example, when the monitoring data is carbon monoxide concentration, if the identified carbon monoxide concentration is not within a normal range, an alarm prompt is made. Because the underground environment is complex, the collected image often contains more Gaussian white noise, the displayed monitoring data is often unclear, and the monitoring data cannot be accurately identified when being identified, so that the collected target surface image is subjected to denoising treatment subsequently.
As an example, this step may include the steps of:
first, a target surface image of an alarm display panel is acquired.
For example, a surface image of the alarm display panel may be acquired by a camera, and the surface image may be grayed to obtain a target surface image.
And secondly, carrying out digital region identification processing on the target surface image to obtain a target digital region.
For example, the target surface image is input into a digital recognition network which is trained in advance, a digital mask image is obtained through the digital recognition network, and the digital mask image is multiplied with the target surface image to obtain a target digital region.
The digital recognition network may be a semantic segmentation neural network.
Optionally, the training process of the digital identification network may comprise the following sub-steps:
a first sub-step of acquiring a sample image set and a sample mask image corresponding to each sample image in the sample image set.
The sample image set is a training set of the digital recognition network. The sample mask image corresponding to the sample image is a training label of the digital identification network. The sample mask image is a binary image.
For example, if the value range of the R value in the RGB values corresponding to the pixel points in the sample image is [230, 255], the pixel value corresponding to the pixel points is updated to 1, otherwise, the pixel value corresponding to the pixel points is updated to 0, and the updated sample image is used as the sample mask image corresponding to the sample image.
And a second sub-step, constructing a digital identification network.
For example, a semantic segmentation neural network can be constructed as a digital recognition network.
And a third sub-step of inputting the sample image set into a digital recognition network, and training the digital recognition network by using sample mask images corresponding to the sample images to obtain the trained digital recognition network.
The loss function of the digital recognition network training process is a cross entropy loss function.
It should be noted that, since each digital display of the alarm is displayed by using the brightness of seven transistors. The brightness of different transistors often represents a different number, for example, the number 8 when seven transistors are fully bright. Thus bright transistors, constitute the display data. In dark environments or environments with larger dust, the transistors of the mining gas alarm are generally red in color when they are bright because of the better transmittance of red. Since the pixel point is often red when the R value in the RGB values corresponding to the pixel point has a value range of [230, 255 ]. Thus, the target digital area can be identified through the digital identification network.
And S2, performing edge detection on the target digital region to obtain a target edge pixel point set.
In some embodiments, edge detection may be performed on the target digital region to obtain a target edge pixel point set.
The target edge pixel point in the target edge pixel point set may be an edge pixel point obtained by performing edge detection.
As an example, edge detection can be performed on the target digital region through a canny edge detection algorithm, so as to obtain a target edge pixel point set.
It should be noted that, because noise often affects the distribution of edge pixel points, for example, when noise is denser, the influence degree of noise is often greater, the detected transistor edges will be irregular, and the distribution of target edge pixel points will be irregular, so that the edge detection obtains a target edge pixel point set, which can facilitate the subsequent analysis of noise contained in the target surface image.
And S3, clustering the target edge pixel points in the target edge pixel point set to obtain a target cluster set.
In some embodiments, the target edge pixel points in the target edge pixel point set may be clustered to obtain a target cluster set.
As an example, density clustering may be performed on target edge pixels in the target edge pixel set to obtain a target cluster set.
It should be noted that, density clustering is performed on the target edge pixel points in the target edge pixel point set, so that classification of the target edge pixel points can be achieved, target cluster clusters with different densities can be obtained, and because the densities of the target edge pixel points in the same target cluster are the same, the noise densities of the target edge pixel points in the same target cluster are often the same, and therefore, subsequent noise aggregation analysis can be conveniently performed on each target cluster.
And S4, carrying out noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain the initial noise influence degree corresponding to the target cluster.
In some embodiments, noise aggregation influence analysis processing may be performed on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster set.
As an example, determining the initial noise impact level corresponding to the target cluster according to the position corresponding to each target edge pixel point in the target cluster, the position corresponding to the cluster center of the target cluster, the number of target edge pixel points in the target cluster set, and the number of target edge pixel points in the target cluster may include the following steps:
The first step, determining the distance between the target edge pixel point and the clustering center according to the position corresponding to each target edge pixel point in the target clustering cluster and the position corresponding to the clustering center of the target clustering cluster, and taking the distance as the target distance corresponding to the target edge pixel point.
And secondly, determining the average value of the target distances corresponding to the target edge pixel points in the target cluster as the average value of the target distances corresponding to the target cluster.
And thirdly, carrying out negative correlation on the target distance mean value corresponding to the target cluster, and normalizing to obtain the distance index corresponding to the target cluster.
And fourthly, determining the ratio of the number of the target edge pixel points in the target cluster to the number of the target edge pixel points in the target cluster set as a number index corresponding to the target cluster.
And fifthly, determining the product of the distance index and the quantity index corresponding to the target cluster as the initial noise influence degree corresponding to the target cluster.
As yet another example, the formula for determining the initial noise impact level corresponding to the target cluster may be:
Wherein, the liquid crystal display device comprises a liquid crystal display device,
is the initial noise influence degree corresponding to the t-th target cluster in the target cluster set. t is the sequence number of the target cluster in the target cluster set. e is a natural constant.
Is the number of target edge pixel points in the t-th target cluster in the target cluster set. N is the number of target edge pixel points in the target cluster set.
Is the abscissa included in the position corresponding to the i-th target edge pixel point in the t-th target cluster in the target cluster set. i is the sequence number of the target edge pixel point in the target cluster.
Is the ordinate included in the position corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set.
Is the abscissa included in the position corresponding to the cluster center of the t-th target cluster in the target cluster set.
Is the ordinate included in the position corresponding to the cluster center of the t-th target cluster in the target cluster set.
It should be noted that, when the influence degree of noise is greater, that is, the more noise is, the more irregular the edge obtained by canny edge detection is, the more the number of target edge pixel points in the target cluster is in the clustering. The noise distribution on different edges is different, the more the noise is, the closer the distance between the pixel points of the target edge is, and the influence range on the edge is The greater the degree, the greater the aggregative nature of the target edge pixels.
The average distance between the target edge pixel points in the t-th target cluster and the cluster center can be represented, and the smaller the average distance is, the greater the density of the target edge pixel points in the t-th target cluster is usually indicated.
Can realize the pair of
And (2) normalization of
The larger the target cluster is, the larger the density of the target edge pixel points in the t target cluster is, and the influence degree of noise on the t target cluster is larger.
It may be that the number of target edge pixels in the t-th target cluster is proportional to the total number of target edge pixels,
the larger the target cluster is, the larger the density of the target edge pixel points in the t target cluster is, and the influence degree of noise on the t target cluster is larger. Therefore, the greater the initial noise influence degree corresponding to the t-th target cluster, the greater the influence degree of noise on the t-th target cluster is often indicated.
And S5, according to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters.
In some embodiments, according to the initial noise influence degree corresponding to each target cluster in the target cluster set, noise distribution change analysis processing may be performed on the target clusters, so as to determine a noise change index corresponding to each target edge pixel point in the target cluster.
As an example, this step may include the steps of:
firstly, determining gradient amplitude values corresponding to all pixel points in a preset target sliding window corresponding to the target edge pixel points.
Wherein the target sliding window may be a 3 x 3 sliding window. The target edge pixel point may be located at a center position of the target sliding window corresponding to the target edge pixel point.
And secondly, determining the average value of the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points as the target gradient average value corresponding to the target edge pixel points.
And thirdly, screening out the minimum target gradient mean value from target gradient mean values corresponding to all target edge pixel points in the target cluster, and taking the minimum target gradient mean value as the minimum gradient mean value corresponding to the target cluster.
And fourthly, determining a difference value between the target gradient mean value corresponding to the target edge pixel point and the first gradient mean value as a target difference value corresponding to the target edge pixel point.
The first gradient mean value is a minimum gradient mean value corresponding to the target cluster where the target edge pixel point is located.
And fifthly, determining the product of the target difference value corresponding to the target edge pixel point and the first influence degree as a noise change index corresponding to the target edge pixel point.
The first influence degree is an initial noise influence degree corresponding to the target cluster where the target edge pixel point is located.
For example, the formula corresponding to the noise variation index corresponding to each target edge pixel point in the target cluster may be determined as:
wherein, the liquid crystal display device comprises a liquid crystal display device,
is a noise change index corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set. t is the sequence number of the target cluster in the target cluster set. i is the sequence number of the target edge pixel point in the target cluster.
Is the initial noise influence degree corresponding to the t-th target cluster in the target cluster set.
Is the target gradient mean value corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set.
Is the minimum gradient mean value corresponding to the t-th target cluster in the target cluster set.
Note that, when the t-th target cluster is not affected by noise,
often approximately 0. Due to the influence of noise, gradient change of target edge pixel points in the target cluster is severe. Thus, the first and second substrates are bonded together,
the larger the i-th target edge pixel point tends to be, the more affected by noise.
The larger the target edge pixel point in the t-th target cluster is, the more affected by noise is often explained. Thus, the first and second substrates are bonded together,
the larger the target edge pixel point in the t-th target cluster is, the more affected by noise is often explained.
And S6, carrying out noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster.
In some embodiments, noise size analysis may be performed on each target cluster in the set of target clusters, to determine a noise size indicator corresponding to each target edge pixel in the target cluster.
As an example, this step may include the steps of:
the first step, the minimum gradient amplitude value is selected from the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points and is used as the minimum gradient amplitude value corresponding to the target edge pixel points.
And a second step of determining, for each pixel point in the target sliding window corresponding to the target edge pixel point, a difference value between the gradient amplitude corresponding to the pixel point and the minimum gradient amplitude corresponding to the target edge pixel point as a target gradient difference value corresponding to the pixel point.
And thirdly, determining the average value of target gradient difference values corresponding to all pixel points in the target sliding window corresponding to the target edge pixel points as a noise size index corresponding to the target edge pixel points.
For example, the formula corresponding to the noise size indicator corresponding to each target edge pixel point in the target cluster may be determined as:
wherein, the liquid crystal display device comprises a liquid crystal display device,
is a noise size index corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set. t is the sequence number of the target cluster in the target cluster set. i is the sequence number of the target edge pixel point in the target cluster. R is the number of pixels within the target sliding window. r is the sequence number of the pixel point in the target sliding window.
Is the gradient amplitude corresponding to the r pixel point in the target sliding window corresponding to the i-th target edge pixel point in the t-th target cluster in the target cluster set. 
Is the minimum gradient amplitude corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set.
It should be noted that, when the noise in the area where the target edge pixel point is located is larger, the gradient change in the target sliding window corresponding to the target edge pixel point tends to be larger.
The larger the gradient change in the target sliding window corresponding to the ith target edge pixel point is, the larger the gradient change in the target sliding window is, thus
The larger the noise tends to be, the more noise is accounted for in the region where the target edge pixel point is located.
Step S7, determining the target noise influence degree corresponding to the target edge pixel points according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set.
In some embodiments, the target noise influence level corresponding to the target edge pixel point may be determined according to a noise variation index and a noise size index corresponding to each target edge pixel point in the target edge pixel point set.
As an example, a product of a noise variation index and a noise size index corresponding to each target edge pixel point in the set of target edge pixel points may be determined as a target noise influence degree corresponding to the target edge pixel point.
For example, the formula corresponding to the target noise influence degree corresponding to the target edge pixel point may be:
wherein, the liquid crystal display device comprises a liquid crystal display device,
is the target noise influence degree corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set.
Is a noise change index corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set.
Is a noise size index corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set. t is the sequence number of the target cluster in the target cluster set. i is the sequence number of the target edge pixel point in the target cluster.
When (when)
And
the larger the size of the product,
the larger the i-th target edge pixel point tends to be, the more affected by noise. Thus, comprehensively consider
And
can be improved
Accuracy of the determination.
And S8, performing adaptive window denoising on the target surface image according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image, and performing digital recognition on the target denoising image.
In some embodiments, the adaptive window denoising may be performed on the target surface image according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set, so as to obtain a target denoising image, and the digital recognition may be performed on the target denoising image.
As an example, this step may include the steps of:
the first step, determining the size of a denoising window corresponding to each target edge pixel point according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set.
The denoising window corresponding to the target edge pixel point may be a filtering window for filtering and denoising the target edge pixel point.
For example, according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set, determining the size of the denoising window corresponding to the target edge pixel point may include the following substeps:
and a first sub-step of determining the product of the target noise influence degree corresponding to the target edge pixel point and a preset target multiple as a first product corresponding to the target edge pixel point.
The target multiple may be a preset multiple. For example, the target multiple may be 10.
And a second sub-step of rounding up the first product corresponding to the target edge pixel point to obtain the size of the denoising window corresponding to the target edge pixel point.
For example, the formula corresponding to the size of the denoising window corresponding to the target edge pixel point may be:
Wherein, the liquid crystal display device comprises a liquid crystal display device,
is the size of the denoising window corresponding to the ith target edge pixel point in the nth target cluster in the target cluster set, that is, the denoising window corresponding to the ith target edge pixel point may be of a side length of
That is, the number of pixels in the denoising window corresponding to the ith target edge pixel is equal to
。
Is a target multiple. As an example of the presence of a metal such as,
may be 10.
Is the target noise influence degree corresponding to the ith target edge pixel point in the ith target cluster in the target cluster set. t is the sequence number of the target cluster in the target cluster set. i is the sequence number of the target edge pixel point in the target cluster.
Is a round-up function.
It should be noted that the number of the substrates,
the larger the i-th target edge pixel point is, the more affected by noise is often indicated, the larger the size of a required denoising window is often indicated for denoising the i-th target edge pixel point. Thus, a target multiple is set
Can make
Positively correlated with the size of the denoising window. Since the size of the denoising window tends not to be a fraction, it is upwardThe size of the denoising window obtained can be made integer.
And secondly, denoising the target edge pixel points and preset target neighborhoods corresponding to the target edge pixel points according to the size of a denoising window corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image.
Wherein the target neighborhood may be a preset neighborhood. For example, the target neighborhood may be a 5×5 neighborhood.
For example, filtering and denoising may be performed on each target edge pixel point in the target edge pixel point set and each pixel point in the target neighborhood corresponding to the target edge pixel point, where a filtering window for filtering and denoising the target edge pixel point is a denoising window corresponding to the target edge pixel point. And the filter window for filtering and denoising each pixel point in the target adjacent area corresponding to the target edge pixel point is a denoising window corresponding to the target edge pixel point.
And thirdly, carrying out digital recognition on the target denoising image.
For example, the target denoising image may be digitally identified by the oxford thresholding method.
For another example, the target denoising image can be digitally identified through a digital identification network.
Alternatively, OCR technology can be used to identify display data on a fixed mining gas alarm, record the data, and then transmit the data to a portable alarm for data verification, and can observe gas monitoring conditions under the coal mine in real time.
In summary, since noise often affects the distribution of edge pixels, for example, when the noise is denser, the influence degree of the noise is often greater, the detected transistor edges are irregular, and the distribution of target edge pixels is often irregular, so that the edge detection obtains a target edge pixel set, which can facilitate the subsequent analysis of noise contained in the target surface image. The density clustering is carried out on the target edge pixel points in the target edge pixel point set, the classification of the target edge pixel points can be realized, the target cluster clusters with different densities can be obtained, and as the densities of the target edge pixel points in the same target cluster are the same, the noise densities of the target edge pixel points in the same target cluster are always the same, so that the subsequent noise aggregation analysis of each target cluster can be facilitated. When the influence degree of noise is larger, that is, the more the noise is, the more irregular the edge obtained through canny edge detection is, so that the more the number of target edge pixel points in the target cluster is in clustering. And the distribution of noise on different edges is also different, the more the noise, the closer the distance between the target edge pixel points is, the greater the influence degree on the edges is, and the greater the aggregation of the target edge pixel points is. And the target cluster is subjected to noise distribution change and noise size analysis processing, so that follow-up accurate denoising can be facilitated. Therefore, the invention realizes the self-adaptive window denoising recognition of the target surface image by carrying out image data processing on the target surface image, can give consideration to denoising and protecting image details, solves the technical problems of low image denoising effect and low recognition accuracy of the display number of the alarm, and improves the image denoising effect and recognition accuracy of the display number of the alarm.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting thereof; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the scope of the embodiments of the present application, and are intended to be included within the scope of the present application.
Claims (6)
1. The machine vision-based alarm display digital recognition method is characterized by comprising the following steps of:
acquiring a target surface image of a display panel of an alarm, and carrying out digital region identification processing on the target surface image to obtain a target digital region;
performing edge detection on the target digital region to obtain a target edge pixel point set;
clustering the target edge pixel points in the target edge pixel point set to obtain a target cluster set;
performing noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster;
According to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters;
performing noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster;
determining a target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set according to the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set;
according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set, performing adaptive window denoising on the target surface image to obtain a target denoising image, and performing digital recognition on the target denoising image;
according to the initial noise influence degree corresponding to each target cluster in the target cluster set, carrying out noise distribution change analysis processing on the target clusters, and determining a noise change index corresponding to each target edge pixel point in the target clusters, wherein the noise change index comprises the following steps:
Determining gradient amplitude values corresponding to all pixel points in a preset target sliding window corresponding to the target edge pixel points;
determining the average value of the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points as the target gradient average value corresponding to the target edge pixel points;
screening out the minimum target gradient mean value from target gradient mean values corresponding to all target edge pixel points in the target cluster, and taking the minimum target gradient mean value as the minimum gradient mean value corresponding to the target cluster;
determining a difference value between a target gradient mean value corresponding to the target edge pixel point and a first gradient mean value as a target difference value corresponding to the target edge pixel point, wherein the first gradient mean value is a minimum gradient mean value corresponding to a target cluster in which the target edge pixel point is located;
determining a product of a target difference value corresponding to the target edge pixel point and a first influence degree as a noise change index corresponding to the target edge pixel point, wherein the first influence degree is an initial noise influence degree corresponding to a target cluster where the target edge pixel point is located;
performing noise size analysis processing on each target cluster in the target cluster set, and determining a noise size index corresponding to each target edge pixel point in the target cluster, including:
Screening out the minimum gradient amplitude value from the gradient amplitude values corresponding to all the pixel points in the target sliding window corresponding to the target edge pixel points, and taking the minimum gradient amplitude value as the minimum gradient amplitude value corresponding to the target edge pixel points;
for each pixel point in a target sliding window corresponding to the target edge pixel point, determining a difference value between the gradient amplitude corresponding to the pixel point and the minimum gradient amplitude corresponding to the target edge pixel point as a target gradient difference value corresponding to the pixel point;
and determining the average value of target gradient difference values corresponding to all pixel points in the target sliding window corresponding to the target edge pixel points as a noise size index corresponding to the target edge pixel points.
2. The machine vision-based alarm display digital identification method of claim 1, wherein performing noise aggregation influence analysis processing on each target cluster in the target cluster set to obtain an initial noise influence degree corresponding to the target cluster comprises:
and determining the initial noise influence degree corresponding to the target cluster according to the positions corresponding to the target edge pixel points in the target cluster, the positions corresponding to the cluster centers of the target cluster, the number of the target edge pixel points in the target cluster set and the number of the target edge pixel points in the target cluster.
3. The machine vision based alarm display digital identification method according to claim 2, wherein determining the initial noise impact level corresponding to the target cluster according to the position corresponding to each target edge pixel point in the target cluster, the position corresponding to the cluster center of the target cluster, the number of target edge pixel points in the target cluster set, and the number of target edge pixel points in the target cluster, comprises:
determining the distance between each target edge pixel point in the target cluster and the clustering center of the target cluster according to the position corresponding to each target edge pixel point in the target cluster and the position corresponding to the clustering center of the target cluster, and taking the distance as the target distance corresponding to the target edge pixel point;
determining the average value of the target distances corresponding to all the target edge pixel points in the target cluster as the average value of the target distances corresponding to the target cluster;
performing negative correlation on the target distance average value corresponding to the target cluster, and normalizing to obtain a distance index corresponding to the target cluster;
determining the ratio of the number of target edge pixel points in the target cluster to the number of target edge pixel points in the target cluster set as a number index corresponding to the target cluster;
And determining the product of the distance index and the quantity index corresponding to the target cluster as the initial noise influence degree corresponding to the target cluster.
4. The machine vision based alarm display digital identification method of claim 1, wherein determining the target noise impact level corresponding to each target edge pixel point in the set of target edge pixel points according to the noise variation index and the noise size index corresponding to the target edge pixel point comprises:
and determining the product of the noise change index and the noise size index corresponding to each target edge pixel point in the target edge pixel point set as the target noise influence degree corresponding to the target edge pixel point.
5. The machine vision-based alarm display digital identification method of claim 1, wherein the performing adaptive window denoising on the target surface image according to the target noise influence degree corresponding to each target edge pixel in the target edge pixel set to obtain a target denoised image comprises:
determining the size of a denoising window corresponding to each target edge pixel point in the target edge pixel point set according to the target noise influence degree corresponding to each target edge pixel point in the target edge pixel point set;
And denoising the target edge pixel points and a preset target neighborhood corresponding to the target edge pixel points according to the size of a denoising window corresponding to each target edge pixel point in the target edge pixel point set to obtain a target denoising image.
6. The machine vision based alarm display digital identification method of claim 5, wherein determining the size of the denoising window corresponding to each target edge pixel in the set of target edge pixels according to the target noise influence level corresponding to the target edge pixel comprises:
determining the product of the target noise influence degree corresponding to the target edge pixel point and a preset target multiple as a first product corresponding to the target edge pixel point;
and carrying out upward rounding on the first product corresponding to the target edge pixel point to obtain the size of the denoising window corresponding to the target edge pixel point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310214709.0A CN115953774B (en) | 2023-03-08 | 2023-03-08 | Alarm display digital recognition method based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310214709.0A CN115953774B (en) | 2023-03-08 | 2023-03-08 | Alarm display digital recognition method based on machine vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115953774A CN115953774A (en) | 2023-04-11 |
| CN115953774B true CN115953774B (en) | 2023-05-19 |
Family
ID=85896246
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310214709.0A Active CN115953774B (en) | 2023-03-08 | 2023-03-08 | Alarm display digital recognition method based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115953774B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116563296B (en) * | 2023-07-12 | 2023-09-05 | 吉林省禹语网络科技有限公司 | Identification method for abdomen CT image |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780263A (en) * | 2021-09-03 | 2021-12-10 | 华南师范大学 | Pressure alarm instrument positioning and reading identification method and device |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6807286B1 (en) * | 2000-04-13 | 2004-10-19 | Microsoft Corporation | Object recognition using binary image quantization and hough kernels |
| CN106257495A (en) * | 2015-06-19 | 2016-12-28 | 阿里巴巴集团控股有限公司 | A kind of digit recognition method and device |
| CN110598684B (en) * | 2019-07-19 | 2021-10-15 | 珠海格力电器股份有限公司 | Method, system, terminal device and storage medium for identifying telephone number in image |
| CN113723410A (en) * | 2020-05-21 | 2021-11-30 | 安徽小眯当家信息技术有限公司 | Digital tube digital identification method and device |
| CN115294158A (en) * | 2022-09-26 | 2022-11-04 | 江苏万森绿建装配式建筑有限公司 | Hot continuous rolling strip steel image segmentation method based on machine vision |
-
2023
- 2023-03-08 CN CN202310214709.0A patent/CN115953774B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113780263A (en) * | 2021-09-03 | 2021-12-10 | 华南师范大学 | Pressure alarm instrument positioning and reading identification method and device |
Non-Patent Citations (1)
| Title |
|---|
| 基于机器视觉技术的气体检测报警器检定装置的研究;郭波等;计量技术(第11期);49-51 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115953774A (en) | 2023-04-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110796046B (en) | Intelligent steel slag detection method and system based on convolutional neural network | |
| CN110334706B (en) | Image target identification method and device | |
| US20170091948A1 (en) | Method and system for automated analysis of cell images | |
| US8139817B2 (en) | Face image log creation | |
| CN108304816B (en) | Identity recognition method and device, storage medium and electronic equipment | |
| US9811885B2 (en) | Detecting glare in a frame of image data | |
| CN110033040B (en) | Flame identification method, system, medium and equipment | |
| CN111209876A (en) | Oil leakage defect detection method and system | |
| CN115953774B (en) | Alarm display digital recognition method based on machine vision | |
| CN110929635B (en) | False face video detection method and system based on face cross-over ratio under trust mechanism | |
| CN110728681B (en) | Mura defect detection method and device | |
| CN115063718B (en) | Fire detection method, device, equipment and storage medium | |
| CN104077776B (en) | A kind of visual background extracting method based on color space adaptive updates | |
| CN109460787A (en) | IDS Framework method for building up, device and data processing equipment | |
| CN109858570A (en) | Image classification method and system, computer equipment and medium | |
| CN115035113B (en) | Nondestructive testing method for production defects of mobile phone data line based on X-ray | |
| Rhee | Detection of spliced image forensics using texture analysis of median filter residual | |
| CN112419229A (en) | Display screen linear defect detection method and device and storage medium | |
| CN111199194A (en) | Automobile intelligent cabin instrument testing method based on machine vision and deep learning | |
| CN112862744A (en) | Intelligent detection method for internal defects of capacitor based on ultrasonic image | |
| CN112052730A (en) | 3D dynamic portrait recognition monitoring device and method | |
| CN117115147B (en) | Textile detection method and system based on machine vision | |
| US20160283805A1 (en) | Method and device for classifying an object in an image | |
| CN116363136B (en) | On-line screening method and system for automatic production of motor vehicle parts | |
| CN111259926A (en) | Meat freshness detection method and device, computing equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Digital Recognition Method for Alarm Display Based on Machine Vision Granted publication date: 20230519 Pledgee: Yutai County Branch of China Postal Savings Bank Co.,Ltd. Pledgor: JINING ANTAI MINE EQUIPMENT MANUFACTURING Co.,Ltd. Registration number: Y2024980012510 |