CN116758057A

CN116758057A - Communication equipment defect detection method based on artificial intelligence

Info

Publication number: CN116758057A
Application number: CN202310999915.7A
Authority: CN
Inventors: 李潇; 田高铬; 潘熙元
Original assignee: Shandong Heming Electric Co ltd
Current assignee: Shandong Heming Electric Co ltd
Priority date: 2023-08-10
Filing date: 2023-08-10
Publication date: 2023-09-15

Abstract

The invention relates to the technical field of image segmentation, in particular to a communication equipment defect detection method based on artificial intelligence, which comprises the following steps: acquiring a target image corresponding to a parabolic antenna to be detected, and screening a target shadow area corresponding to a target feed source component from the target image; acquiring a target origin of a minimum gray value pixel point used for representing a target shadow area in the target shadow area; gray level change rule analysis processing is carried out on the pixel points on the straight line passing through the target origin and the shadow pixel points; determining suspected corrosion points based on the suspected corrosion indexes to generate a suspected corrosion area; performing edge regularity analysis treatment on each suspected corrosion area; and determining the suspected corrosion area with the edge irregularity index larger than the preset corrosion threshold as a target corrosion defect area. According to the invention, the accuracy of corrosion defect detection on the shadow area of the feed source component of the parabolic antenna is improved by performing image data processing on the target image.

Description

Communication equipment defect detection method based on artificial intelligence

Technical Field

The invention relates to the technical field of image segmentation, in particular to a communication equipment defect detection method based on artificial intelligence.

Background

The communication device may be a communication related device such as a parabolic antenna. The working environment of the parabolic antenna is outdoor, and the corrosion-resistant layer on the surface of the antenna is dropped due to foreign matter impact or wind, sun and the like, so that the antenna has corrosion defects, the gain performance of the antenna is reduced, loss or limited transmission distance of signals is caused when the signals are transmitted, and therefore the corrosion defects of the parabolic antenna are required to be detected. Under illumination, there is often a shadow of the parabolic reflector that makes up the parabolic antenna, and due to the effect of the shadow, it is often difficult to separate a corrosion defect region from the shadow region of the feed component. Currently, when dividing an image, the following methods are generally adopted: the image is segmented according to the difference of the gray values.

However, when dividing the corrosion defective area from the hatched area of the feed component according to the difference in gray values, there are often the following technical problems:

because the gray values in the corrosion area and the non-corrosion area in the shadow area of the feed source component are often not different, when the corrosion defect area is divided from the shadow area of the feed source component, if only the difference of the gray values is considered, misjudgment of the corrosion defect pixel point is likely to be caused, and therefore the accuracy of corrosion defect detection on the shadow area of the feed source component of the parabolic antenna 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 that the accuracy of corrosion defect detection on a shadow area of a feed source component of a parabolic antenna is low, the invention provides a communication equipment defect detection method based on artificial intelligence.

The invention provides a communication equipment defect detection method based on artificial intelligence, which comprises the following steps:

acquiring a target image corresponding to a parabolic antenna to be detected, and screening a target shadow area corresponding to a target feed source component from the target image;

acquiring a target origin of a minimum gray value pixel point used for representing the target shadow region in the target shadow region;

for each shadow pixel point in the target shadow region, carrying out gray level change rule analysis processing on the pixel points on a straight line passing through the target origin and the shadow pixel points to obtain suspected corrosion indexes corresponding to the shadow pixel points;

Shadow pixel points with suspected corrosion indexes larger than a preset suspected threshold value are determined to be suspected corrosion points, and a suspected corrosion area is generated according to each suspected corrosion point;

performing edge regularity analysis treatment on each suspected corrosion area to obtain an edge irregularity index corresponding to the suspected corrosion area;

and determining each suspected corrosion area with the edge irregularity index larger than the preset corrosion threshold value as a target corrosion defect area.

Optionally, the method further comprises:

screening a target parabolic area corresponding to a parabolic reflector of the parabolic antenna to be detected from the target image;

determining the ratio of the total area of all the target corrosion areas in the area of the target parabolic area as a target ratio;

and when the target duty ratio is larger than a preset duty ratio threshold, sending warning information of serious corrosion.

Optionally, the analyzing the gray level change rule of the pixel point on the straight line passing through the target origin and the shadow pixel point to obtain a suspected corrosion index corresponding to the shadow pixel point includes:

determining a straight line passing through the target origin and the shadow pixel point as a target straight line corresponding to the shadow pixel point;

Taking the pixel points intersecting the target straight line in a preset circular neighborhood corresponding to the shadow pixel points as reference pixel points to obtain a reference pixel point sequence corresponding to the shadow pixel points;

determining an average value of gray values corresponding to all reference pixel points in the reference pixel point sequence as a target gray index corresponding to the shadow pixel point;

determining a first suspected index corresponding to the shadow pixel point according to the target gray index, wherein the target gray index and the first suspected index are in negative correlation;

determining the difference value of gray values corresponding to every two adjacent reference pixel points in the reference pixel point sequence as the target gray difference between the two reference pixel points to obtain a target gray difference sequence corresponding to the shadow pixel points;

performing discrete condition analysis processing on the target gray level difference sequence to obtain a second suspected index corresponding to the shadow pixel point;

and determining a suspected corrosion index corresponding to the shadow pixel point according to the first suspected index and the second suspected index, wherein the first suspected index and the second suspected index are positively correlated with the suspected corrosion index.

Optionally, the performing discrete condition analysis processing on the target gray level difference sequence to obtain a second suspected indicator corresponding to the shadow pixel point includes:

the frequency of each target gray level difference in the target gray level difference sequence is used as the corresponding target frequency of each target gray level difference;

and determining target information entropy corresponding to the target gray level difference sequence according to target frequencies corresponding to various target gray level differences, and taking the target information entropy as a second suspected index corresponding to the shadow pixel point.

Optionally, the generating a suspected corrosion area according to each suspected corrosion point includes:

and carrying out region growth on all the obtained suspected corrosion points according to the positions corresponding to the suspected corrosion points, and determining each region obtained after the region growth as a suspected corrosion region.

Optionally, the performing edge regularity analysis on each suspected corrosion area to obtain an edge irregularity index corresponding to the suspected corrosion area includes:

determining the distance between each edge pixel point on the edge of the suspected corrosion area and the target origin as a reference distance to obtain a reference distance set corresponding to the suspected corrosion area;

Determining variances of all the reference distances in the reference distance set as first clutter indexes corresponding to the suspected corrosion areas;

screening a reference suspected region set corresponding to each suspected corrosion region from the suspected corrosion region set;

determining Euclidean distance between the Fourier descriptor corresponding to the suspected corrosion region and the Fourier descriptor corresponding to each reference suspected region in the reference suspected region set as a target distance, and obtaining a target distance set corresponding to the suspected corrosion region;

determining a second clutter index corresponding to the suspected corrosion area according to the target distance set, wherein the target distance and the second clutter index are positively correlated;

and determining an edge irregularity index corresponding to the suspected corrosion area according to the first clutter index and the second clutter index, wherein the first clutter index and the second clutter index are positively correlated with the edge irregularity index.

Optionally, the screening the target parabolic area corresponding to the parabolic reflector of the parabolic antenna to be detected from the target image includes:

carrying out Hough circle transformation detection on the target image to obtain a candidate circle set;

Screening out the largest candidate circle from the candidate circle set to be used as a target circle;

and determining the circular area where the target circle is located as a target parabolic area.

Optionally, the acquiring the target origin of the pixel point with the minimum gray value in the target shadow area, which is used for representing the target shadow area, includes:

performing corner detection on the target shadow area to obtain a first corner and a second corner;

and determining the midpoint of the connecting line of the first angular point and the second angular point as a target origin.

The invention has the following beneficial effects:

according to the communication equipment defect detection method based on artificial intelligence, through image data processing on the target image, the technical problem that the accuracy of corrosion defect detection on the shadow area of the feed source part of the parabolic antenna is low is solved, and the accuracy of corrosion defect detection on the shadow area of the feed source part of the parabolic antenna is improved. Firstly, a target shadow area corresponding to the target feed source component is screened out from the acquired target image, so that whether corrosion defects exist in the target shadow area can be conveniently and subsequently judged. Then, because the shadow of the feed source component on the parabolic reflector is in a fan-shaped radial shape, the shadow is deeper near the middle part of the target shadow area, and the corresponding gray value is smaller, so that the obtained target origin of the pixel point with the minimum gray value for representing the target shadow area in the target shadow area can represent the middle part of the target shadow area. If the feed source component shadow area has no corrosion defect, when the pixel point in the feed source component shadow area is closer to the middle part, the gray value corresponding to the pixel point is usually indicated to be smaller; when the pixel point in the shadow area of the feed source component is far from the middle part, the gray value corresponding to the pixel point is relatively larger, however, when corrosion defects exist in the shadow area of the feed source component, the rule is broken, so that gray change rule analysis processing is conducted on the pixel point on a straight line passing through the target origin and the shadow pixel point, and whether the shadow pixel point is the corrosion defect pixel point can be conveniently judged through judging gray change rule of the pixel point on the straight line passing through the target origin and the shadow pixel point. Then, since the edges of the corrosion areas are often irregular, that is, the regularity is often poor, the edge regularity analysis processing is performed on each suspected corrosion area, so that whether the suspected corrosion area is a corrosion area can be conveniently judged later. Finally, each suspected corrosion area with the irregular edge index larger than the preset corrosion threshold value is determined as a target corrosion defect area, so that the corrosion defect detection of the target shadow area is realized, and compared with the mode that only different gray values are considered, the method comprehensively considers a plurality of indexes related to the corrosion defect, such as a target origin, the suspected corrosion index, the irregular edge index and the like, so that the final judgment result is relatively more objective, and the accuracy of the corrosion defect detection of the target shadow area is 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 an artificial intelligence based communication device defect detection method of the present invention;

FIG. 2 is a schematic diagram of a process of region growing using a first region growing rule according to the present invention;

FIG. 3 is a schematic view of a first corner point and a second corner point of the present invention;

FIG. 4 is a schematic diagram of a reference pixel point according to the present invention;

fig. 5 is a schematic diagram of a process of region growing using a second region growing rule according to the present invention.

Wherein, the reference numerals include: a first region 201, a second region 202, a third region 203, a closed figure 301, a first dot a, a second dot B, a pixel 401, a straight line 402, a circular region 403, a fourth region 501, a fifth region 502, and a sixth region 503.

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.

acquiring a target origin of a minimum gray value pixel point used for representing a target shadow region in the target shadow region;

for each shadow pixel point in the target shadow region, carrying out gray level change rule analysis processing on the pixel points on the straight line passing through the target origin and the shadow pixel points to obtain suspected corrosion indexes corresponding to the shadow pixel points;

The following detailed development of each step is performed:

referring to FIG. 1, a flow diagram of some embodiments of an artificial intelligence based communication device defect detection method in accordance with the present invention is shown. The communication equipment defect detection method based on artificial intelligence comprises the following steps:

step S1, a target image corresponding to a parabolic antenna to be detected is obtained, and a target shadow area corresponding to a target feed source component is screened out from the target image.

In some embodiments, a target image corresponding to the parabolic antenna to be detected may be obtained, and a target shadow area corresponding to the target feed source component may be screened from the target image.

The parabolic antenna to be detected can be a parabolic antenna to be subjected to corrosion defect detection. The parabolic antenna may be a planar antenna consisting of a parabolic reflector and a feed element (radiator) located at its focal point. The target image may be an image of the parabolic antenna to be detected. The target feed component may be a feed component that makes up the parabolic antenna to be tested. The target shadow area may be an area where a shadow of a target feed component formed on a parabolic reflector constituting the parabolic antenna to be detected is located.

The target shadow area corresponding to the target feed source component is screened from the acquired target image, so that whether corrosion defects exist in the target shadow area can be conveniently and subsequently judged.

As an example, this step may include the steps of:

first, obtaining an initial image corresponding to a parabolic antenna to be detected.

For example, an image of the parabolic antenna to be detected may be acquired as an initial image by a camera, such as a rotatable tele camera. The initial image may be an RGB (Red Green Blue) image, among others.

And secondly, performing image preprocessing on the initial image to obtain a target image.

Wherein image preprocessing may include, but is not limited to: denoising, graying, and image enhancement. The target image may be an initial image after image preprocessing.

For example, image preprocessing of the initial image to obtain the target image may include the sub-steps of:

and the first substep, carrying out graying on the initial image by adopting an average value method to obtain a gray image.

The grayscale image may be an initial image after graying.

And a second sub-step, adopting a Gaussian filter denoising algorithm to denoise the gray level image to obtain a target image.

The target image may be a gray-scale image after denoising.

And thirdly, screening out a target shadow area corresponding to the target feed source component from the target image.

For example, screening the target shadow region corresponding to the target feed component from the target image may include the sub-steps of:

and a first substep, carrying out Hough circle transformation detection on the target image, and taking the obtained circle as a candidate circle to obtain a candidate circle set.

The hough circle transformation detection is also called hough circle detection.

And a second sub-step of screening out the largest candidate circle from the candidate circle set as a target circle.

And a third substep, determining a circular area where the target circle is located as a target parabolic area.

It should be noted that the boundary of the parabolic reflector constituting the parabolic antenna to be detected is often the largest circle in the parabolic antenna to be detected. The target parabolic area may thus characterize the area where the parabolic reflectors constituting the parabolic antenna to be tested are located. The shadow of the target feed component tends to fall on the parabolic reflector that makes up the parabolic antenna to be tested. I.e., the shadows of the target feed components tend to fall on the target parabolic areas.

And a fourth sub-step, performing edge detection on the target parabolic area to obtain an edge area.

For example, the target parabolic area may be regarded as an image, and the image is subjected to edge detection, so that the obtained edge image is the edge area. Wherein the edge image is a binary image.

And a fifth sub-step, performing region growth on the edge region according to pixel values corresponding to all pixel points in the edge region, and taking the region obtained by the region growth as a first reference region to obtain a first reference region set, wherein the rule of region growth can be that adjacent pixel points with the same corresponding pixel values are divided into the same region.

For example, the region growing rule that the adjacent pixels with the same corresponding pixel value are divided into the same region is described as the first region growing rule, and the process of performing region growing by using the first region growing rule may be as shown in fig. 2, and the square filled with oblique lines may represent the pixel with the pixel value of 0. The unfilled squares may represent pixels with a pixel value of 1. The second region 202 and the third region 203 can be obtained by performing region growth on the first region 201 using the first region growth rule.

It should be noted that, according to the pixel values corresponding to the pixel points in the edge area, the edge area is subjected to area growth, and the area surrounded by each closed edge can be grown into an area, so that the first reference area can be the area surrounded by the closed edge. Since the boundary of the feed shadow is composed of closed edges, if the feed shadow exists in the target parabolic region, the region where the feed shadow exists tends to exist in the first reference region set.

And a sixth substep, determining each first reference area in the first reference area set to correspond to an area in the target image as a second reference area corresponding to each first reference area, and obtaining a second reference area set.

The second reference area corresponding to the first reference area may be an area where edge detection is not performed in the first reference area.

And a seventh substep, inputting a second reference area in the second reference area set into a feed source shadow recognition network which is trained in advance, judging whether the second reference area is a feed source shadow or not through the feed source shadow recognition network, and taking the second reference area which is judged to be the feed source shadow as a target shadow area.

Wherein the feed shadow recognition network may be used to recognize feed shadows. The feed shadow recognition network may be DNN (Deep Neural Networks, deep neural network). The feed shadows may be shadows of the feed components.

Optionally, the training process of the feed shadow recognition network may include the steps of:

the first step is to acquire a set of sample feed source shadow areas.

The sample feed source shadow areas in the sample feed source shadow area set can be image areas formed by feed source shadows with different sizes and different illumination intensities.

And secondly, constructing a feed source shadow recognition network.

For example, a DNN may be constructed and the constructed DNN used as a pre-training feed shadow recognition network.

Thirdly, training the feed source shadow recognition network according to the sample feed source shadow region set to obtain the trained feed source shadow recognition network.

For example, the sample feed source shadow region set may be used as a training set to train the feed source shadow recognition network, so as to obtain a trained feed source shadow recognition network.

And S2, acquiring a target origin of a minimum gray value pixel point used for representing the target shadow region in the target shadow region.

In some embodiments, a target origin of a minimum gray value pixel point in the target shadow region for characterizing the target shadow region may be obtained.

As an example, this step may include the steps of:

and the first step, carrying out corner detection on the target shadow area to obtain a first corner and a second corner.

For example, a FAST-12 (Features from Accelerated Segment Test, corner detection) algorithm may be used to detect two corners from within the target shadow region, and use the two corners as the first corner and the second corner, respectively. Parameters of the FAST algorithm can be adjusted according to actual conditions, so that two corner points can be detected.

It should be noted that, the first corner and the second corner are screened out from the target shadow area, so that the target origin can be conveniently determined later.

As shown in fig. 3, the region where the closed graph 301 is located may characterize the target shadow region. The first and second points a and B may represent a first and second corner point, respectively.

It should be noted that, since the parabolic reflector is shaped as a paraboloid, and the feed source component is shaped as a circle, the shadow of the feed source component on the parabolic reflector is in a fan-shaped radial shape (as shown in fig. 3), and the size of the shadow is changed along with the change of illumination, but the structural characteristics of the shadow are unchanged, and the whole structure still presents a fan-shaped radiation structure. Therefore, the first corner point and the second corner point are acquired, and the subsequent analysis of the gray level change of the shadow region of the target can be facilitated.

And secondly, determining the midpoint of the connecting line of the first angular point and the second angular point as a target origin.

The target origin may be a midpoint of a line connecting the first corner and the second corner.

For example, the first corner point and the second corner point may be connected, and a line segment obtained by the connection is used as a first line segment, and a midpoint of the first line segment is used as a target origin.

The shadow depths at the respective positions in the target shadow area tend to be different, and since the shadows of the feed source components on the parabolic reflector tend to be fan-shaped radial, the shadow tends to be deeper as they are closer to the middle portion of the target shadow area. The target origin may often approximately characterize the location of the deepest shadow within the shadow region of the target. If no corrosion defect exists in the target shadow area, when the pixel point in the target shadow area is closer to the target origin, the gray value corresponding to the pixel point is usually indicated to be smaller; when a pixel point in the target shadow area is farther from the target origin, the gray value corresponding to the pixel point is often indicated to be relatively larger. Secondly, if the pixel point with the smallest gray value is selected directly and randomly as the target origin, the pixel point with the smallest gray value is possibly caused to be located at the position close to the edge of the target shadow area due to the abnormal gray value at a certain position, so that the gray change rule is subsequently carried out by utilizing the pixel point on the straight line passing through the target origin and the shadow pixel point, the determined suspected corrosion index is lower in accuracy, and the shadow of the feed source component on the parabolic reflector is often in a fan-shaped radial shape, the structure is relatively fixed, so that the midpoint of the connecting line of the first angular point and the second angular point is determined as the target origin, and the obtained target origin can be relatively more capable of representing the middle part of the target shadow area.

And S3, for each shadow pixel point in the target shadow area, carrying out gray level change rule analysis processing on the pixel points on the straight line passing through the target origin and the shadow pixel points, and obtaining a suspected corrosion index corresponding to the shadow pixel points.

In some embodiments, for each shadow pixel point in the target shadow area, gray scale change rule analysis processing may be performed on a pixel point on a straight line passing through the target origin and the shadow pixel point, so as to obtain a suspected corrosion indicator corresponding to the shadow pixel point.

The shadow pixel points may be pixel points within a target shadow region. For example, a pixel within the target shadow region may be referred to as a shadow pixel.

It should be noted that, based on the target origin, gray distribution analysis is performed on each shadow pixel point in the target shadow area, so that accuracy of determining the suspected corrosion index corresponding to each shadow pixel point can be improved.

As an example, this step may include the steps of:

and a first step of determining a straight line passing through the target origin and the shadow pixel point as a target straight line corresponding to the shadow pixel point.

For example, for a certain shadow pixel, a straight line passing through the shadow pixel and the target origin may be regarded as a target straight line corresponding to the shadow pixel.

And secondly, taking the pixel points intersecting the target straight line in the preset circular neighborhood corresponding to the shadow pixel points as reference pixel points to obtain a reference pixel point sequence corresponding to the shadow pixel points.

The preset circular neighborhood may be a preset circular neighborhood. For example, the preset circular neighborhood may be a circular neighborhood with a radius of 3.

For example, for a certain shadow pixel, a pixel point intersecting with a target straight line corresponding to the shadow pixel in a preset circular neighboring area corresponding to the shadow pixel may be used as a reference pixel to obtain a reference pixel set corresponding to the shadow pixel. And sequencing the reference pixel point set corresponding to the shadow pixel point according to the distance between the reference pixel point and the target origin point and the sequence from near to far to obtain the reference pixel point sequence corresponding to the shadow pixel point. The shadow pixel point may be located at a center of a preset circular neighborhood corresponding to the shadow pixel point.

The pixel points intersecting the target straight line corresponding to the shadow pixel points in the preset circular neighboring area corresponding to the shadow pixel points are often the pixel points on a diameter in the preset circular neighboring area, which is in the same direction as the target straight line. Therefore, when the reference pixel point sequence is determined, the preset circular neighborhood is adopted, so that the number of the reference pixel points in the reference pixel point sequence corresponding to each shadow pixel point is approximately equal, and the subsequent processing can be facilitated.

As shown in fig. 4, the square may represent pixels, and the network-filled pixels are noted as first shadow pixels. The linefilled pixels may characterize a reference pixel of the first shadow pixel. The pixel with line filling comprises: network-filled pixels and diagonal-filled pixels. The pixel point 401 may characterize the target origin. Line 402 may represent a target line corresponding to a first shadow pixel. The circular region 403 may represent a preset circular neighborhood corresponding to the first shadow pixel. If no corrosion defect exists in the target shadow area, when the reference pixel point is closer to the target origin, the gray value corresponding to the reference pixel point is usually indicated to be smaller; when the reference pixel point is far from the target origin, the gray value corresponding to the reference pixel point is often indicated to be relatively larger.

And thirdly, determining the average value of gray values corresponding to all the reference pixel points in the reference pixel point sequence as a target gray index corresponding to the shadow pixel point.

For example, for a certain shadow pixel, the average value of the gray values corresponding to all the reference pixels in the reference pixel sequence corresponding to the shadow pixel may be used as the target gray index corresponding to the shadow pixel.

Fourth, according to the target gray index, determining a first suspected index corresponding to the shadow pixel point.

The target gray index may be negatively correlated with the first suspected index.

And fifthly, determining the difference value of the gray values corresponding to every two adjacent reference pixel points in the reference pixel point sequence as the target gray difference between the two reference pixel points to obtain the target gray difference sequence corresponding to the shadow pixel points.

For example, for any two adjacent reference pixels in the reference pixel sequence, a difference between a gray value corresponding to a next reference pixel and a gray value corresponding to a previous reference pixel in the two reference pixels may be used as the target gray difference between the two reference pixels. For example, the target gray scale difference between the first reference pixel point and the second reference pixel point in the reference pixel point sequence may be: the gray value corresponding to the second reference pixel point is subtracted from the gray value corresponding to the first reference pixel point.

And sixthly, performing discrete condition analysis processing on the target gray level difference sequence to obtain a second suspected index corresponding to the shadow pixel point.

For example, performing discrete condition analysis processing on the target gray level difference sequence to obtain the second suspected index corresponding to the shadow pixel point may include the following substeps:

a first sub-step of taking, as a target frequency corresponding to each target gradation difference, a frequency at which each target gradation difference in the target gradation difference sequence appears in the target gradation difference sequence.

For example, if a certain target gray scale difference sequence is {1,2,1,1,3}, 3 target gray scale differences are 1,2 and 3 in the target gray scale difference sequence, wherein the target frequency corresponding to the target gray scale difference 1 may be 3/5, the target frequency corresponding to the target gray scale difference 2 may be 1/5, and the target frequency corresponding to the target gray scale difference 3 may be 1/5.

And a second sub-step of determining target information entropy corresponding to the target gray level difference sequence according to target frequencies corresponding to various target gray level differences, and using the target information entropy as a second suspected index corresponding to the shadow pixel point.

The target information entropy corresponding to the target gray level difference sequence may be the information entropy of the target gray level difference sequence.

Seventh, determining a suspected corrosion index corresponding to the shadow pixel point according to the first suspected index and the second suspected index.

The first suspected index and the second suspected index may both be positively correlated with the suspected corrosion index.

For example, the formula for determining the suspected corrosion index corresponding to the shadow pixel point may be:

；

wherein, the liquid crystal display device comprises a liquid crystal display device,is a suspected corrosion index corresponding to the t-th shadow pixel point in the target shadow region. />Is the first suspected index corresponding to the t shadow pixel point. />Is corresponding to the t-th shadow pixel pointAnd the second suspected index is the target information entropy corresponding to the target gray difference sequence corresponding to the t-th shadow pixel point. />Is a natural constant +.>To the power. />Is the number of reference pixel points in the reference pixel point sequence corresponding to the t-th shadow pixel point.Is the gray value corresponding to the j-th reference pixel point in the reference pixel point sequence corresponding to the t-th shadow pixel point. />The target gray index corresponding to the t shadow pixel point is the average value of gray values corresponding to all reference pixel points in the reference pixel point sequence corresponding to the t shadow pixel point. />And->And has negative correlation. />Is the number of the target gray scale differences in the target gray scale difference sequence corresponding to the t-th shadow pixel point. />The target frequency corresponding to the a-th target gray scale difference in the target gray scale difference sequence corresponding to the t-th shadow pixel point; namely, the frequency of occurrence of the a-th target gray scale difference in the target gray scale difference sequence corresponding to the t-th shadow pixel point; i.e. the ratio of the first number to the second number. Wherein the first number is the t-th shadow image The number of the a-th target gray scale differences in the target gray scale difference sequence corresponding to the pixel points. The second number is the number of target gray differences in the target gray difference sequence corresponding to the t-th shadow pixel point. />Is based on b->Logarithmic (log). />Is->Normalized values. />Is a normalization function, and normalization can be achieved. b is a preset number greater than 1. For example, b may be 2./>And->All are in charge of>And shows positive correlation. t is the sequence number of the shadow pixel point in the target shadow region. j is the serial number of the reference pixel point in the reference pixel point sequence corresponding to the t-th shadow pixel point. a is the number of the target gray difference in the target gray difference sequence corresponding to the t-th shadow pixel point.

It should be noted that, compared to the non-corroded area in the shadow, the corroded area in the shadow tends to be darker in color, i.e., the gray value corresponding to the corroded area in the shadow tends to be lower. When (when)When the pixel is bigger, the gray value corresponding to the reference pixel of the reference pixel sequence corresponding to the t-th shadow pixel is usually lower, and the more likely that the t-th shadow pixel is in a corrosion areaAnd (5) pixel points. Since the target shadow area is in a fan-shaped radial shape, the gray value in the target shadow area is gradually changed, so that if no corrosion defect exists in the target shadow area, the target gray difference in the target gray difference sequence corresponding to the shadow pixel point is relatively stable, and the fluctuation degree is relatively small. If corrosion defects exist in the target shadow area, the target gray scale difference in the target gray scale difference sequence corresponding to the shadow pixel point is often disordered, and the fluctuation degree is often larger. When- >When the target gray level difference sequence is larger, the fluctuation degree of the target gray level difference in the target gray level difference sequence corresponding to the t-th shadow pixel point is larger, the target gray level difference in the target gray level difference sequence corresponding to the t-th shadow pixel point is more disordered, the more likely that corrosion defects exist to cause gray level fluctuation is indicated, and the more likely that the t-th shadow pixel point is a pixel point in a corrosion area is indicated. Thus, when->The larger the shading pixel point, the more likely it is to be a pixel point in the eroded area.

And S4, determining shadow pixels with suspected corrosion indexes larger than a preset suspected threshold value as suspected corrosion points, and generating a suspected corrosion area according to each suspected corrosion point.

In some embodiments, shadow pixels with suspected corrosion indexes greater than a preset suspected threshold may be determined as suspected corrosion points, and a suspected corrosion area may be generated according to each suspected corrosion point.

It should be noted that, based on the suspected corrosion index, the accuracy of the suspected corrosion area set division can be improved.

As an example, this step may include the steps of:

and determining the shadow pixel point as a suspected corrosion point when the suspected corrosion index corresponding to the shadow pixel point is larger than a preset suspected threshold value.

The preset suspected threshold may be a preset threshold. For example, the preset suspected threshold may be 0.5.

And secondly, carrying out region growth on all the obtained suspected corrosion points according to the positions corresponding to the suspected corrosion points, determining each region obtained after the region growth as a suspected corrosion region, and obtaining a plurality of suspected corrosion regions, wherein the rule of the region growth can be that the suspected corrosion points with the distance smaller than or equal to a preset distance threshold value are divided into the same region. The preset distance threshold may be a preset threshold. For example, the preset distance threshold may be 0.5.

For example, if the distance between two suspected corrosion points is less than or equal to a preset distance threshold, the two suspected corrosion points may be partitioned into the same suspected corrosion region.

For example, the region growing rule that the suspected corrosion points with the distance smaller than or equal to the preset distance threshold are divided into the same region is described as the second region growing rule, the process of performing region growing by using the second region growing rule can be as shown in fig. 5, and the square grid filled with the vertical lines can represent the suspected corrosion points. The fifth region 502 and the sixth region 503 can be obtained by performing region growth on the suspected corrosion points in the fourth region 501 using the second region growth rule.

The suspected corrosion region may be a region surrounded by suspected corrosion points.

And S5, carrying out edge regularity analysis processing on each suspected corrosion area to obtain an edge irregularity index corresponding to the suspected corrosion area.

In some embodiments, edge regularity analysis may be performed on each suspected corrosion area to obtain an edge irregularity index corresponding to the suspected corrosion area.

It should be noted that, performing edge regularity analysis processing on each suspected corrosion region in the suspected corrosion region set can improve accuracy of determining an edge irregularity index corresponding to each suspected corrosion region.

As an example, this step may include the steps of:

the first step, determining the distance between each edge pixel point on the edge of the suspected corrosion area and the target origin as a reference distance, and obtaining a reference distance set corresponding to the suspected corrosion area.

The reference distance set corresponding to the suspected corrosion area may include: and the distance between each edge pixel point on the edge of the suspected corrosion area and the target origin.

And a second step of determining the variance of all the reference distances in the reference distance set as a first clutter index corresponding to the suspected corrosion area.

For example, the variance of all the reference distances in the set of reference distances corresponding to the suspected corrosion region may be determined as the first clutter indicator corresponding to the suspected corrosion region.

And thirdly, screening out a reference suspected region set corresponding to each suspected corrosion region from the suspected corrosion region sets.

The reference suspected region in the reference suspected region set corresponding to the suspected corrosion region may be other suspected corrosion regions except for the suspected corrosion region. For example, the reference suspected region set corresponding to the first suspected region in the suspected region set may include: all of the suspected corrosion areas in the set of suspected corrosion areas except the first suspected corrosion area.

And fourthly, determining Euclidean distance between the Fourier descriptor corresponding to the suspected corrosion area and the Fourier descriptor corresponding to each reference suspected area in the reference suspected area set as a target distance to obtain a target distance set corresponding to the suspected corrosion area.

And fifthly, determining a second clutter index corresponding to the suspected corrosion area according to the target distance set.

Wherein the target distance may be positively correlated with the second clutter indicator.

And a sixth step of determining an edge irregularity index corresponding to the suspected corrosion region according to the first clutter index and the second clutter index.

The first clutter index and the second clutter index may both be positively correlated with the edge irregularity index.

For example, the formula corresponding to the edge irregularity index for determining the suspected corrosion region may be:

；

wherein, the liquid crystal display device comprises a liquid crystal display device,is an edge irregularity index corresponding to the ith suspected corrosion area in the suspected corrosion area set. />Is the first clutter index corresponding to the ith suspected corrosion area. />The target distance corresponding to the h reference suspected region in the reference suspected region set corresponding to the i suspected region is the Euclidean distance between the Fourier descriptor corresponding to the i suspected region and the Fourier descriptor corresponding to the h reference suspected region. />Is the number of reference suspected regions in the set of reference suspected regions corresponding to the ith suspected corrosion region. />Is a second clutter indicator corresponding to the ith suspected corrosion area. />And->And shows positive correlation. />And->All are in charge of>And shows positive correlation. />Is->Normalized values. />Is a normalization function, and normalization can be achieved. i is the number of the suspected corrosion region in the set of suspected corrosion regions. h is the serial number of the reference suspected region in the reference suspected region set corresponding to the ith suspected corrosion region.

It should be noted that the edges of the corrosion defective region tend to be relatively irregular. When (when)The larger the distribution of the reference distances in the reference distance set corresponding to the ith suspected corrosion area, the more irregular the edge of the ith suspected corrosion area, the more disordered the edge of the ith suspected corrosion area, and the more likely the ith suspected corrosion area is a corrosion defect area. When->The greater the outline of the ith suspected corrosion area, the more dissimilar the outline of each reference suspected area in the reference suspected area set, the more consistent the outline of the ith suspected corrosion area to the outline of the corrosion defect area, and the more likely the ith suspected corrosion area is the corrosion defect area. Thus, when->The larger the i-th suspected corrosion region, the more likely it is to be a corrosion defect region. Second, the general corrosion tends not to be completely concentrated together, tends to be relatively dispersed in a region, and is suspected when corrosion is presentThe number of suspected corrosion areas in the set of corrosion-like areas is often more than one, so in general, the number of suspected corrosion areas in the set of suspected corrosion areas is often 0 or more than 1. If only one suspected corrosion area in the suspected corrosion area set occurs in the detection process, the suspected corrosion area can be independently determined, for example, whether the suspected corrosion area is a corrosion defect area can be manually determined, or +. >As an index of edge irregularities corresponding to the suspected corroded area.

And S6, determining each suspected corrosion area with the edge irregularity index larger than the preset corrosion threshold value as a target corrosion defect area.

In some embodiments, each suspected corrosion region having an edge irregularity index greater than a preset corrosion threshold may be determined as a target corrosion defect region.

It should be noted that, according to the edge irregularity index, the accuracy of the subsequent judgment of whether each suspected corrosion area is a corrosion defect area can be improved.

As an example, when the edge irregularity index corresponding to the suspected corrosion region is greater than the preset corrosion threshold, the suspected corrosion region may be determined as the target corrosion defect region.

The preset corrosion threshold may be a preset threshold. For example, the preset corrosion threshold may be 0.7.

Optionally, the artificial intelligence based communication device defect detection method may further include the steps of:

and a first step of screening out a target parabolic area corresponding to a parabolic reflector of the parabolic antenna to be detected from the target image.

For example, the screening of the target parabolic area corresponding to the parabolic reflector of the parabolic antenna to be detected from the target image may comprise the following sub-steps:

And a first substep, carrying out Hough circle transformation detection on the target image to obtain a candidate circle set.

And a second sub-step of screening out the largest candidate circle from the candidate circle set to be used as a target circle.

And a third sub-step of determining a circular area where the target circle is located as a target parabolic area.

And secondly, determining the ratio of the total area of all the target corrosion areas in the area of the target parabolic area as a target ratio.

For example, the formula for determining the target duty cycle may be:

；

wherein, the liquid crystal display device comprises a liquid crystal display device,is the target duty cycle. sof is the total area of all target corrosion areas, i.e. the total number of pixels in all target corrosion areas. som is the area of the target parabolic area, i.e., the total number of pixels in the target parabolic area.

When the following is performedThe larger the area, the larger the area ratio of the corrosion defect in the target shadow area, and the more serious the corrosion condition in the target shadow area.

And thirdly, when the target duty ratio is larger than a preset duty ratio threshold value, sending warning information of serious corrosion.

The preset duty ratio threshold may be a preset threshold. For example, the preset duty cycle threshold may be 2%. The warning information may be indicative of more severe corrosion in the shadow area of the target. For example, the warning information may be "there is serious corrosion in the target shadow area (shadow area of the target feed source component), and repair is required.

Alternatively, the sof may also be the total area of all areas of the target parabolic area where corrosion defects occur, i.e., the target area and all target corrosion areasThe total area of the domain, wherein the target area may be the total area of the target parabolic area where corrosion defects occur in areas other than the target shadow area. When sof represents the total area of all the areas where corrosion defects occur in the target parabolic area,corrosion conditions in the target parabolic region may be characterized. The warning information sent at this time may be indicative of more severe corrosion in the target parabolic area. For example, the warning information may be "there is a serious corrosion in the target parabolic area (parabolic reflector), and the maintenance is required. Because the area except the target shadow area in the target parabolic area is often a non-shadow area, the corrosion in the non-shadow area is often obvious, and the corrosion can be often identified by means of manual visual inspection or a neural network, and the details are not repeated.

In conclusion, the target shadow area corresponding to the target feed source component is screened out from the acquired target image, so that whether corrosion defects exist in the target shadow area or not can be conveniently judged. And then, the first corner point and the second corner point are screened out from the target shadow area, so that the subsequent determination of the target origin point can be facilitated. Then, the shadow depths at the respective positions within the target shadow area tend to be different, and since the shadows of the feed components on the parabolic reflector tend to be fan-shaped radial, the shadow tends to be deeper closer to the middle portion of the target shadow area. The target origin may often approximately characterize the location of the deepest shadow within the shadow region of the target. If no corrosion defect exists in the target shadow area, when the pixel point in the target shadow area is closer to the target origin, the gray value corresponding to the pixel point is usually indicated to be smaller; when a pixel point in the target shadow area is farther from the target origin, the gray value corresponding to the pixel point is often indicated to be relatively larger. And continuing to perform gray distribution analysis processing on each shadow pixel point in the target shadow region based on the target origin, so that the accuracy of determining the suspected corrosion index corresponding to each shadow pixel point can be improved. Then, based on the suspected corrosion index, the accuracy of the suspected corrosion area set division can be improved. And then, carrying out edge regularity analysis processing on each suspected corrosion area in the suspected corrosion area set, so that the accuracy of determining the edge irregularity index corresponding to each suspected corrosion area can be improved. Finally, according to the edge irregularity index, the accuracy of the subsequent judgment of whether each suspected corrosion area is a corrosion defect area can be improved. Second, the severity of corrosion of the parabolic reflector can be determined based on the total area of all target corrosion areas.

The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will 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 spirit of the invention and are intended to be included within the scope of the invention.

Claims

1. The communication equipment defect detection method based on artificial intelligence is characterized by comprising the following steps of:

2. The method for detecting defects of a communication device based on artificial intelligence of claim 1, further comprising:

3. The method for detecting defects of communication equipment based on artificial intelligence according to claim 1, wherein the step of analyzing gray level change rule of pixels on a line passing through the target origin and the shadow pixels to obtain suspected corrosion indexes corresponding to the shadow pixels comprises the steps of:

4. The method for detecting defects of communication equipment based on artificial intelligence according to claim 3, wherein the performing discrete condition analysis processing on the target gray level difference sequence to obtain a second suspected index corresponding to the shadow pixel point comprises:

5. The method for detecting defects of a communication device based on artificial intelligence according to claim 1, wherein the generating suspected corrosion areas according to each suspected corrosion point comprises:

6. The method for detecting defects of communication equipment based on artificial intelligence according to claim 1, wherein the performing edge regularity analysis processing on each suspected corrosion area to obtain an edge irregularity index corresponding to the suspected corrosion area comprises:

7. The method for detecting defects of communication equipment based on artificial intelligence according to claim 2, wherein the step of screening out a target parabolic area corresponding to a parabolic reflector of a parabolic antenna to be detected from the target image comprises the steps of:

8. The method for detecting defects of a communication device based on artificial intelligence according to claim 1, wherein the obtaining a target origin of a minimum gray value pixel point in the target shadow area for characterizing the target shadow area comprises: