CN111275124A - Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system - Google Patents
Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system Download PDFInfo
- Publication number
- CN111275124A CN111275124A CN202010084873.0A CN202010084873A CN111275124A CN 111275124 A CN111275124 A CN 111275124A CN 202010084873 A CN202010084873 A CN 202010084873A CN 111275124 A CN111275124 A CN 111275124A
- Authority
- CN
- China
- Prior art keywords
- color
- adsorbent
- transformer respirator
- sample
- color space
- 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.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 13
- 238000001514 detection method Methods 0.000 title claims abstract description 11
- 238000010276 construction Methods 0.000 title abstract description 3
- 238000013145 classification model Methods 0.000 title description 2
- 239000003086 colorant Substances 0.000 claims abstract description 65
- 239000003463 adsorbent Substances 0.000 claims description 176
- 238000012549 training Methods 0.000 claims description 47
- 238000004891 communication Methods 0.000 claims description 46
- 238000000034 method Methods 0.000 claims description 22
- 238000012706 support-vector machine Methods 0.000 claims description 11
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- IRERQBUNZFJFGC-UHFFFAOYSA-L azure blue Chemical compound [Na+].[Na+].[Na+].[Na+].[Na+].[Na+].[Na+].[Na+].[Al+3].[Al+3].[Al+3].[Al+3].[Al+3].[Al+3].[S-]S[S-].[O-][Si]([O-])([O-])[O-].[O-][Si]([O-])([O-])[O-].[O-][Si]([O-])([O-])[O-].[O-][Si]([O-])([O-])[O-].[O-][Si]([O-])([O-])[O-].[O-][Si]([O-])([O-])[O-] IRERQBUNZFJFGC-UHFFFAOYSA-L 0.000 claims description 4
- 235000013799 ultramarine blue Nutrition 0.000 claims description 4
- 239000002594 sorbent Substances 0.000 claims description 3
- VYPSYNLAJGMNEJ-UHFFFAOYSA-N Silicium dioxide Chemical compound O=[Si]=O VYPSYNLAJGMNEJ-UHFFFAOYSA-N 0.000 description 20
- 239000000741 silica gel Substances 0.000 description 20
- 229910002027 silica gel Inorganic materials 0.000 description 20
- 229960001866 silicon dioxide Drugs 0.000 description 20
- 238000012360 testing method Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 230000008878 coupling Effects 0.000 description 4
- 238000010168 coupling process Methods 0.000 description 4
- 238000005859 coupling reaction Methods 0.000 description 4
- 239000003921 oil Substances 0.000 description 4
- 239000007787 solid Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000029058 respiratory gaseous exchange Effects 0.000 description 3
- 238000010521 absorption reaction Methods 0.000 description 2
- 238000009413 insulation Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 229920002379 silicone rubber Polymers 0.000 description 2
- 239000006096 absorbing agent Substances 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 229920001296 polysiloxane Polymers 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/03—Recognition of patterns in medical or anatomical images
- G06V2201/034—Recognition of patterns in medical or anatomical images of medical instruments
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Spectrometry And Color Measurement (AREA)
Abstract
The invention discloses a transformer respirator detection method, a classifier identification model construction method, a computer readable medium storing a transformer respirator detection program, and a transformer respirator monitoring system, and belongs to the field of transformer detection. The detection of the available state of the transformer respirator is realized by associating at least 5 characteristics of hue, saturation and brightness in an HSV color space, blue concentration offset and red concentration offset in an YCbCr color space with the classification of non-failed colors and failed colors of the transformer respirator.
Description
Technical Field
The invention relates to the technical field of transformer respirator maintenance, in particular to a transformer respirator detection method, a method for constructing a classifier identification model, a computer readable medium with a transformer respirator detection program stored therein and a transformer respirator monitoring system.
Background
The respirator is also called a moisture absorber, is an auxiliary safety protection device of a main transformer, and is widely applied to power systems. South weather is wet and the frequency of silica gel adsorbent failure in respirators is high. Therefore, it is a routine maintenance work for the main transformer to replace the silica gel of the breather. The traditional silicon rubber tank of the respirator needs to be withdrawn from the main transformer before replacement to protect heavy gas, and the heavy gas protection is restarted after the replacement is finished. Because the time of replacing the silicon rubber tank of the respirator is long, the heavy gas relay and the respirator can not protect the main transformer in the period, if the transformer breaks down in the period, the protection is rejected, serious power grid risks are caused, and even the main transformer is damaged. Meanwhile, in the time of replacing the silica gel of the respirator, the surrounding humid gas directly enters the inside of the transformer without the respirator, so that transformer oil is affected with damp, the insulation property of the oil is reduced, and the transformer oil loses the insulation property along with the increase of the replacement times, so that the main transformer is scrapped or other power grid accidents occur.
Chinese patent document CN110148511A, published in 2019, 8, and 20, describes a main transformer respirator auxiliary device, which comprises a main transformer breathing tube and a main respirator silicone tank, wherein a three-way pipe with a flange plate is vertically installed at the tail end of the main transformer breathing tube; a main pipe butterfly valve is arranged at the joint of the main pipe of the three-way pipe with the flange and the main respirator silica gel tank; and a branch pipe butterfly valve is arranged at the joint of the branch pipe of the three-way pipe with the flange and the silica gel tank of the auxiliary respirator. This patent document also discloses a method for replacing a main transformer respirator auxiliary device, comprising: when the auxiliary silica gel tank fails, the branch pipe butterfly valve is closed, and the main pipe butterfly valve is opened; the silica gel tank of the auxiliary respirator is detached; replacing the failed silica gel in the auxiliary breathing silica gel tank with qualified silica gel; the replaced silica gel tank of the auxiliary respirator is filled back to the branch pipe butterfly valve; and the main pipe butterfly valve is opened. When the product and the method are used, the color change degree of the silica gel in the silica gel tank of the main respirator needs to be observed manually, and whether the silica gel tank of the main respirator needs to be stopped and the silica gel tank of the auxiliary respirator needs to be started is judged.
Chinese patent document CN107977959A, published in 5/1/2018, describes a respirator state identification method suitable for an electric power robot, which includes the following steps: (1-1) collecting and calibrating a respirator and storing related information, wherein the stored related information comprises a respirator position, a respirator image, respirator image characteristics and a respirator image characteristic vector; (1-2) acquiring an image of the respirator at the current moment in the scene image as an image to be detected; (1-3) extracting the features of the image to be detected in the step (1-2), calculating the feature vector of the image to be detected, and performing feature matching with the features of the respirator image and the feature vector of the respirator image in the step (1-1) to obtain a respirator sub-image; (1-4) performing image preprocessing calculation on the respirator subimages in the step (1-3) to obtain preprocessed respirator subimages; and (1-5) performing state analysis on the preprocessed image in the step (1-4) to obtain the current state of the respirator in the scene image.
Disclosure of Invention
One of the objectives of the present invention is to provide a transformer respirator detection method to detect the available status of a transformer respirator.
The invention also provides a method for constructing a classifier recognition model, so that the transformer respirator detection method can be applied to computer equipment.
It is still another object of the present invention to provide a computer readable medium storing a transformer respirator detection program for facilitating a computer device to detect a usage status of a transformer respirator.
The invention also provides a transformer respirator monitoring system, which is convenient for automatically detecting the available state of the transformer respirator.
In order to solve the technical problems, the following technical scheme can be selected according to the needs:
a transformer respirator detection method comprises the following steps:
a step of establishing a classifier identification model capable of classifying an available state of the adsorbent as non-failed or failed based on color characteristics [ H, S, V, Cb, Cr ] of the adsorbent, wherein H is a hue in an HSV color space, S is a saturation in an HSV color space, V is a brightness in an HSV color space, Cb is a blue concentration offset in a YCbCr color space, and Cr is a red concentration offset in the YCbCr color space;
acquiring color data of an adsorbent sample of a transformer respirator to be detected, wherein the color data comprises color characteristics [ H, S, V, Cb, Cr ] corresponding to an adsorbent of the adsorbent sample, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space, and the available state of the adsorbent sample can reflect the available state of all adsorbents of the transformer respirator to be detected;
a step of estimating a usable state of the adsorbent sample by identifying a color of each unit adsorbent in the adsorbent sample using the classifier identification model and determining a usable state of the unit adsorbent, and estimating at least two of a total number of unit adsorbents in an unexpired state of the adsorbent sample, a total number of unit adsorbents in a failed state of the adsorbent sample, and a total number of unit adsorbents in the adsorbent sample;
and determining the available state of the transformer respirator to be detected according to the available state of the adsorbent sample.
A method of constructing a classifier recognition model, comprising the steps of:
acquiring a training image sample, wherein pixel colors in the training image sample comprise a positive color and a negative color, the positive color belongs to a first color gamut, the negative color belongs to a second color gamut, and the intersection of the first color gamut and the second color gamut is empty;
classifying training image samples, wherein the training image samples with the marked pixel colors as positive example colors are positive example training samples, and the training image samples with the marked pixel colors as negative example colors are negative example training samples;
establishing an initial classifier identification model for associating characteristic components [ H, S, V, Cb and Cr ] of positive example colors with a first color gamut and associating characteristic components [ H, S, V, Cb and Cr ] of negative example colors with a second color gamut, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space;
and optimizing the initial classifier identification model to obtain an optimal classifier identification model.
Preferably, the positive training image sample adopts images with multiple positive colors formed in a natural state; the counterexample training image sample adopts images with various counterexample colors formed in a natural state.
Preferably, a Support Vector Machine (SVM) algorithm is adopted to perform classification training on the training image samples so as to obtain an optimal classifier identification model.
Preferably, the number of samples belonging to the first color gamut: the number of samples belonging to the second color gamut is 1: 1.
further, the union of the first color gamut and the second color gamut is equal to the color gamut of the CCS color circle, the union of the third color gamut and the fourth color gamut is equal to the second color gamut, and the intersection of the third color gamut and the fourth color gamut is empty; the first color gamut corresponds to a red color region between the purple and orange of the CCS hue circle, the third color gamut corresponds to a blue color region between the cyan and ultramarine blue of the CCS hue circle, and the number of samples belonging to the third color gamut in the training image sample is as follows: the number of samples belonging to the fourth color gamut is 3.5-4.5: 1.
a computer readable medium storing a transformer respirator detection program, the transformer respirator detection program comprising an input module, a color classification module and an output module, the input module being configured to obtain image data of an adsorbent sample of a transformer respirator to be detected, the available state of the adsorbent sample being capable of reflecting the available state of all adsorbents of the transformer respirator to be detected, the image data including color features [ H, S, V, Cb, Cr ] corresponding to each pixel point in an image of the adsorbent sample, wherein H is hue in an HSV color space, S is saturation in an HSV color space, V is brightness in the HSV color space, Cb is blue concentration offset in a YCr color space, and Cr is red concentration offset in the YCbCr color space;
the color classification module is used for classifying whether the color of each pixel point in the image of the adsorbent sample belongs to a first color gamut, wherein the first color gamut corresponds to a color region when the adsorbent is in a failure state, or the first color gamut corresponds to a color region when the adsorbent is not in a failure state;
the output module is used for outputting at least two of the total number of pixel point colors in the image of the adsorbent sample belonging to the first color gamut, the total number of pixel point colors not belonging to the first color gamut and the total number of pixel points in the image of the adsorbent sample.
The transformer respirator monitoring system comprises a lower computer, wherein the lower computer comprises a controller, a computer readable medium storing a transformer respirator detection program is arranged in the controller, or the controller is in communication connection with the computer readable medium storing the transformer respirator detection program.
Furthermore, the device further comprises an upper computer, the lower computer further comprises a communication module which is used for being in communication connection with the upper computer, and the communication module is in communication connection with the controller.
The color identification module is in communication connection with the controller and used for being installed at a corresponding position of the transformer respirator to be detected so as to identify color data of the adsorbent of the transformer respirator to be detected.
The device further comprises a color camera module in communication connection with the controller, wherein the color camera module is used for being installed at a corresponding position of the transformer respirator to be detected so as to acquire color image data of the adsorbent of the transformer respirator to be detected.
In the field of transformer maintenance, those skilled in the art have never considered using machine identification algorithms to identify the health of the sorbent within the breather. The specific reasons are as follows: (1) humidity transducer can record the internal humidity of jar of respirator, and the adsorbent in the respirator became invalid the back, and the internal humidity of respirator jar can reach certain level, can indirectly judge the adsorbent in the respirator like this and need change. However, under actual conditions, the time is required for the adsorbent to adsorb moisture, the humidity sensor senses the humidity instantly, and the humidity in the respirator tank measured by the humidity sensor cannot be directly used; at the moment, when the measurement result of the humidity sensor can judge that the adsorbent in the respirator needs to be replaced, the moisture absorption effect in the respirator is invalid, and the service life of the equipment is influenced; this can result in a waste of resources if the adsorbent in the respirator is replaced prematurely. (2) Theoretical researchers think that the color sensor can realize the discernment to the adsorbent colour of respirator, but, under actual conditions, the colour change of the silica gel that discolours in the respirator is progressive process, and the colour rank and colour block size constitute the judgement factor that judges whether adsorbent in the respirator needs to be changed jointly. In addition, when an image of the adsorbent of the respirator is captured, the color level of the adsorbent is greatly affected by the brightness of the ambient light during the capturing, and moreover, the image also has more noise. The single color sensor cannot identify the available state of the transformer respirator with high accuracy. When the machine identification algorithm is adopted, the adsorbent and oil in the respirator are not negatively influenced, and meanwhile, the identification pertinence is improved by aiming at the characteristic components [ H, S, V, Cb and Cr ] of the allochroic silica gel of the respirator. In selecting the training image sample, the invention also selects the image formed in the natural state to make it closer to the color state involved in the color gradient process of the adsorbent, thereby improving the accuracy of the machine recognition algorithm in recognizing the available state of the adsorbent in the respirator.
Compared with the prior art, the invention has the beneficial effects that:
(1) the RGB color space has the minimum discrimination on colors, the HSV color space can obviously distinguish the difference of hue and saturation in the colors, and the YCbCr color space can obviously distinguish red and blue. The invention adopts hue, saturation and brightness under HSV color space and blue concentration offset and red concentration offset of YCbCr color space as characteristics, can better identify the color of the adsorbent in the transformer respirator, and further know the available state of the adsorbent.
(2) In an image formed in a natural state, the transition of colors is relatively soft, and generally includes derivative colors of the same color at different brightness and gradation colors of similar colors. By adopting the image as a training image sample, the stability of the classifier is good, and the classifier is more suitable for identifying the image color in a real state.
(3) Compared with other classification algorithms, the Support Vector Machine (SVM) algorithm is more robust to noise abnormity and the like in data sets, and the obtained classifier identification model is good in identification effect and more stable.
(4) After the computer readable medium storing the transformer respirator detection program is connected with the controller, the controller can operate the transformer respirator detection program to identify the available state of the transformer respirator.
(5) When the controller passes through communication module and host computer communication, can know the available state of transformer respirator through the host computer is long-range, improved transformer respirator's state identification efficiency.
Detailed Description
The following examples are presented to illustrate the present invention and to assist those skilled in the art in understanding and practicing the present invention. Unless otherwise indicated, the following embodiments and technical terms therein should not be understood to depart from the background of the technical knowledge in the technical field.
The adsorbent is placed in the existing transformer respirator tank body and is doped with allochroic silica gel. When the color-changing silica gel changes from blue to light red, the adsorbent is wet and must be replaced and dried. Thus, when the color of the adsorbent is blue, the usable state of the adsorbent is not spent; when the adsorbent color is light red, the usable state of the adsorbent is spent. When the working condition of the transformer respirator is identified by machine vision, the state of the transformer respirator can be judged only by the color of the adsorbent in the transformer respirator, and other parts such as texture, contour, shape and the like are unchanged.
First part of the invention
A transformer respirator detection method comprises the following steps:
and establishing a classifier identification model. The method specifically comprises the following steps: the classifier identification model is capable of classifying the usable state of the adsorbent as non-failed or failed based on the color characteristics [ H, S, V, Cb, Cr ] of the adsorbent, where H is hue in HSV color space, S is saturation in HSV color space, V is brightness in HSV color space, Cb is blue concentration offset in YCbCr color space, and Cr is red concentration offset in YCbCr color space;
the method comprises the steps of obtaining color data of an adsorbent sample of the transformer respirator to be detected, wherein the color data comprises color characteristics [ H, S, V, Cb and Cr ] corresponding to an adsorbent of the adsorbent sample, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space. The available state of the adsorbent sample can reflect the available state of all adsorbents of the transformer respirator to be detected; the adsorbent sample can be all adsorbents in the transformer respirator to be detected or part of adsorbents in the transformer respirator to be detected; generally, the adsorbent on the inner wall of the transformer respirator to be detected can also be regarded as an adsorbent sample of the transformer respirator to be detected, so that the available state of the transformer respirator to be detected is judged by observing the color of the adsorbent through a window of the transformer respirator to be detected. In addition, the adsorbent sample may be a test paper capable of reflecting the moisture content of the adsorbent. The test paper preferably has the following characteristics: the test paper showed a blue color when the adsorbent was tested without water absorption. When the test paper is tested for a moist adsorbent (i.e., the adsorbent does not have an adsorbing effect), the test paper appears pink or red. The test paper can be arranged in an adsorbent tank body of the transformer respirator.
And measuring the available state of the adsorbent sample. The method specifically comprises the following steps: the color of each unit adsorbent in the adsorbent sample is identified using a classifier identification model and the availability status of that unit adsorbent is determined. Each unit adsorbent can be an adsorbent, a pile of adsorbents with the same size, a pixel point in an adsorbent image, and a color block with the same size in the adsorbent image. And calculating at least two of the total number of unitary adsorbents in an unexpired state of the adsorbent sample, the total number of unitary adsorbents in a failed state of the adsorbent sample, and the total number of unitary adsorbents in the adsorbent sample. Since the total number of the unit adsorbents in the non-failure state of the adsorbent sample + the total number of the unit adsorbents in the failure state of the adsorbent sample is the total number of the unit adsorbents in the adsorbent sample, two of the total number of the unit adsorbents in the non-failure state of the adsorbent sample, the total number of the unit adsorbents in the failure state of the adsorbent sample, and the total number of the unit adsorbents in the adsorbent sample are calculated, and the total number of the third one can be known.
And determining the available state of the transformer respirator to be detected according to the available state of the adsorbent sample. According to the current specification, when the failed adsorbent in the transformer respirator to be detected is: the non-failure adsorbent is more than or equal to 2: 1, the adsorbent in the transformer respirator to be detected needs to be replaced. Thus, when the total number of unitary adsorbents in a spent state for an adsorbent sample: the total number of the unit adsorbents in the non-failure state of the adsorbent sample is more than or equal to 2: the usable state of the adsorbent sample is the unusable state 1. Since the available state of the adsorbent sample can reflect the available states of all adsorbents of the transformer respirator to be detected, the available state of the transformer respirator to be detected is also corresponding to an unavailable state.
Second part of the invention
A method of constructing a classifier recognition model, comprising the steps of:
acquiring a training image sample, wherein pixel colors in the training image sample comprise a positive color and a negative color, the positive color belongs to a first color gamut, the negative color belongs to a second color gamut, and the intersection of the first color gamut and the second color gamut is empty; according to the color of the existing allochroic silicagel, generally, the first color gamut corresponds to a red color region between purple and orange of a CCS color phase ring, or the first color gamut corresponds to a blue color region between blue-green and ultramarine-blue of the CCS color phase ring. Classifying training image samples, wherein the training image samples with the marked pixel colors as positive example colors are positive example training samples, and the training image samples with the marked pixel colors as negative example colors are negative example training samples;
establishing an initial classifier identification model for associating characteristic components [ H, S, V, Cb and Cr ] of positive example colors with a first color gamut and associating characteristic components [ H, S, V, Cb and Cr ] of negative example colors with a second color gamut, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the bCr color space;
the initial classifier identification model is optimized to obtain an optimal classifier identification model.
Example 1: a method of constructing a classifier recognition model, comprising the steps of:
the method comprises the steps of obtaining a training image sample, wherein pixel colors in the training image sample comprise a positive color and a negative color, the positive color belongs to a first color gamut, the negative color belongs to a second color gamut, and an intersection of the first color gamut and the second color gamut is empty. Considering that the pure color colorimetric card is used as a training image sample, the difference between the pure color colorimetric card and the application environment of the classifier identification model is large, for example, when the pure color colorimetric card image is collected, the collection environment of each training image is not easy to be consistent, the color of the pure color colorimetric card can change color in different illumination environments, and uncontrollable factors are too many. In this embodiment, the training sample image is a real image. For example, a "blue picture" may be retrieved in a search engine, with the colors in the mostly blue non-solid color picture as counterexample colors. These primarily blue non-solid color pictures allow a small number of other color pixels to appear. A "red picture" may be retrieved in a search engine, with colors in a non-solid picture that are primarily red as the positive example colors. These primarily red non-solid color pictures allow a small number of other color pixels to appear. The trial image sample obtained in this way has diversity, and is closer to the actual color appearing in the adsorbent color gradient process of the respirator of the transformer substation.
Classifying training image samples, wherein the training image samples with the marked pixel colors as positive example colors are positive example training samples, and the training image samples with the marked pixel colors as negative example colors are negative example training samples;
and carrying out classification training on the classified training image samples by adopting a Support Vector Machine (SVM) algorithm so as to obtain an optimal classifier identification model. The characteristic components of the SVM algorithm are [ H, S, V, Cb, Cr ], wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space; the classifier identification model is used to associate the feature components [ H, S, V, Cb, Cr ] of the positive example color with a first color gamut and the feature components [ H, S, V, Cb, Cr ] of the negative example color with a second color gamut. The present invention uses libsvm calculation software developed by professor Lin chi-Jen, Taiwan university, and others, and determines optimal parameters using Cross _ validation and Grid _ search functions. Finally, the classifier adopting the Gaussian kernel function is obtained. The support vector machine SVM algorithm is adopted to classify the non-pure-color training image samples, the obtained classifier has better universality on color change, and the method is more suitable for identifying the color of the sample in the transformer respirator.
Preferably, the number of samples belonging to the first color gamut: the number of samples belonging to the second color gamut is 1: 1.
further, the union of the first color gamut and the second color gamut is equal to the color gamut of the CCS color circle, the union of the third color gamut and the fourth color gamut is equal to the second color gamut, and the intersection of the third color gamut and the fourth color gamut is empty; the first color gamut corresponds to a red color region between the purple and orange of the CCS hue circle, the third color gamut corresponds to a blue color region between the cyan and ultramarine blue of the CCS hue circle, and the number of samples belonging to the third color gamut in the training image sample is as follows: the number of samples belonging to the fourth color gamut is 3.5-4.5: 1.
in specific implementation, the inventor takes the number of samples belonging to the first color gamut as 1000, takes the number of samples belonging to the third color gamut as 800, and takes the number of samples belonging to the third color gamut as 200.
The method in the embodiment is applied to detecting the available states of 50 transformer respirators, the classification accuracy is in the range of 82% -88%, and the classification accuracy is concentrated at 85%. This result is considered acceptable considering the large amount of noise present in the sorbent sample image of the transformer respirator.
Third part of the invention
A computer readable medium storing a transformer respirator detection program, the transformer respirator detection program comprising an input module, a color classification module, and an output module.
The input module is used for acquiring image data of an adsorbent sample of the transformer respirator to be detected. The image data includes color features [ H, S, V, Cb, Cr ] corresponding to each pixel point in the image of the adsorbent sample, where the pixel point may be understood as a pixel point in a strict sense or as a unit color block. Wherein, H is hue under HSV color space, S is saturation under HSV color space, V is brightness under HSV color space, Cb is blue concentration offset under YCbCr color space, and Cr is red concentration offset under YCbCr color space. The available state of the adsorbent sample can reflect the available state of all adsorbents of the transformer respirator to be detected.
The color classification module is used for classifying whether the color of each pixel point in the image of the adsorbent sample belongs to a first color gamut, wherein the first color gamut corresponds to a color region when the adsorbent is in a failure state, or the first color gamut corresponds to a color region when the adsorbent is not in a failure state.
The output module is used for outputting at least two of the total number of pixel point colors in the image of the adsorbent sample, the total number of pixel point colors not belonging to the first color gamut and the total number of pixel points in the image of the adsorbent sample. The sum of the pixel colors belonging to the first color gamut in the image of the adsorbent sample and the sum of the pixel colors not belonging to the first color gamut are equal to the sum of the pixel colors in the image of the adsorbent sample, so that two of the total number of the pixel colors belonging to the first color gamut in the image of the adsorbent sample, the total number of the pixel colors not belonging to the first color gamut in the image of the adsorbent sample and the total number of the pixel colors belonging to the first color gamut in the image of the adsorbent sample are obtained, and the total number of the third pixel is known.
When the transformer respirator detection program is operated, the transformer respirator detection program executes the following steps:
the input module acquires image data of an adsorbent sample of the transformer respirator to be detected.
After the color classification module receives image data of an adsorbent sample of the transformer respirator to be detected, whether the color of each pixel point in the image of the adsorbent sample belongs to the first color gamut is classified. When the color of the pixel point in the image of the adsorbent sample belongs to the first color gamut, the total number of the pixel point colors in the image of the adsorbent sample belongs to the first color gamut is + 1; when the color of the pixel point in the image of the adsorbent sample does not belong to the first color gamut, the pixel point color does not belong to the total number +1 of the first color gamut. And after the color of each pixel point in the image of the adsorbent sample is classified, outputting the total number of the pixel point colors in the image of the adsorbent sample belonging to the first color gamut and the total number of the pixel point colors in the image of the adsorbent sample not belonging to the first color gamut.
The output module outputs the total number of pixel point colors in the image of the adsorbent sample belonging to the first color gamut and the total number of pixel point colors in the image of the adsorbent sample not belonging to the first color gamut.
According to the requirement, the output module can further output the ratio of the total number of the pixel point colors in the image of the adsorbent sample belonging to the first color gamut to the total number of the pixel point colors in the image of the adsorbent sample not belonging to the first color gamut.
According to the requirement, the output module can also output the available state of the transformer respirator to be detected according to the ratio of the total number of the pixel point colors in the image of the adsorbent sample belonging to the first color gamut to the total number of the pixel point colors in the image of the adsorbent sample not belonging to the first color gamut. For example, the first color gamut corresponds to the color region when the adsorbent has failed, and the color of the pixel in the image of the adsorbent sample belongs to the total number of the first color gamut: the total number of pixel point colors in the image of the adsorbent sample which do not belong to the first color gamut is more than or equal to 2: 1, the output module outputs data or a signal which codes information that the adsorbent in the transformer respirator to be detected needs to be replaced, and the signal can trigger an alarm to send out an alarm signal or automatically replace the transformer respirator. The first color gamut corresponds to a color region when the adsorbent has failed, and the color of a pixel point in the image of the adsorbent sample belongs to the total number of the first color gamut: the total number of pixel point colors in the image of the adsorbent sample which do not belong to the first color gamut is less than 2: 1, the output module outputs data or signals which are coded with information that the adsorbent in the transformer respirator to be detected does not need to be replaced temporarily.
Fourth aspect of the invention
The transformer respirator monitoring system comprises a lower computer, wherein the lower computer comprises a controller, a computer readable medium in which a transformer respirator detection program is stored is arranged in the controller, or the controller is in communication connection with the computer readable medium in which the transformer respirator detection program is stored. That is, the transformer respirator detection program may be stored in the controller, in a computer readable medium externally attached to the controller, or in a computer device with a communication function, which is in communication connection with the controller.
Example 2: a transformer respirator monitoring system comprises a lower computer and an upper computer, wherein the lower computer comprises a controller and a communication module for communication connection with the upper computer, a computer readable medium storing a transformer respirator detection program is arranged in the controller, or the controller is in communication connection with the computer readable medium storing the transformer respirator detection program; the communication module is in communication connection with the controller. The communication module can be a wired communication module, a wireless communication module, a mechanical coupling module, an optical coupling module, an acoustic coupling module and the like. In the prior art, the wired communication module comprises an RS485 communication module, an RS232 communication module, an RS422 communication module, an SPI communication module, an I2C communication module, a serial port communication module and the like, and also can realize wired communication through electrical signal communication, parallel port communication and the like.
Example 3: a transformer respirator monitoring system comprises a lower computer, wherein the lower computer comprises a controller and a color identification module. The controller is internally provided with a computer readable medium which stores a transformer respirator detection program, or the controller is in communication connection with the computer readable medium which stores the transformer respirator detection program. The color recognition module is in communication connection with the controller. The communication connection means that communication is formed between the connected devices through transmission interaction of signals. The specific implementation modes of the communication connection are as follows: electrical signal communication, electromagnetic wave communication, mechanical coupling communication, optical communication, voice communication, and the like. The color identification module is used for being installed at a corresponding position of the transformer respirator to be detected so as to identify color data of the adsorbent of the transformer respirator to be detected. The color recognition module may employ a color sensing chip including a TCS3200 type for corresponding connection with the controller.
Example 4: a transformer respirator monitoring system comprises a lower computer, wherein the lower computer comprises a controller and a color camera module. The controller is internally provided with a computer readable medium which stores a transformer respirator detection program, or the controller is in communication connection with the computer readable medium which stores the transformer respirator detection program. The color camera module is in communication connection with the controller and is used for being installed at a corresponding position of the transformer respirator to be detected so as to acquire color image data of the adsorbent of the transformer respirator to be detected.
The present invention is described in detail with reference to the examples. It should be understood that in practice it is not intended to be exhaustive of all possible embodiments, and the inventive concepts of the present invention are presented herein by way of illustration. Without departing from the inventive concept of the present invention and without any creative work, a person skilled in the art should, in all of the embodiments, make optional combinations of technical features and experimental changes of specific parameters, or make a routine replacement of the disclosed technical means by using the prior art in the technical field to form specific embodiments, which belong to the content implicitly disclosed by the present invention.
Claims (10)
1. The transformer respirator detection method is characterized by comprising the following steps:
a step of establishing a classifier identification model capable of classifying an available state of the adsorbent as non-failed or failed based on color characteristics [ H, S, V, Cb, Cr ] of the adsorbent, wherein H is a hue in an HSV color space, S is a saturation in an HSV color space, V is a brightness in an HSV color space, Cb is a blue concentration offset in a YCbCr color space, and Cr is a red concentration offset in the YCbCr color space;
acquiring color data of an adsorbent sample of a transformer respirator to be detected, wherein the color data comprises color characteristics [ H, S, V, Cb, Cr ] corresponding to an adsorbent of the adsorbent sample, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space, and the available state of the adsorbent sample can reflect the available state of all adsorbents of the transformer respirator to be detected;
a step of estimating a usable state of the adsorbent sample by identifying a color of each unit adsorbent in the adsorbent sample using the classifier identification model and determining a usable state of the unit adsorbent, and estimating at least two of a total number of unit adsorbents in an unexpired state of the adsorbent sample, a total number of unit adsorbents in a failed state of the adsorbent sample, and a total number of unit adsorbents in the adsorbent sample;
and determining the available state of the transformer respirator to be detected according to the available state of the adsorbent sample.
2. A method of constructing a classifier recognition model, comprising the steps of:
acquiring a training image sample, wherein pixel colors in the training image sample comprise a positive color and a negative color, the positive color belongs to a first color gamut, the negative color belongs to a second color gamut, and the intersection of the first color gamut and the second color gamut is empty;
classifying training image samples, wherein the training image samples with the marked pixel colors as positive example colors are positive example training samples, and the training image samples with the marked pixel colors as negative example colors are negative example training samples;
establishing an initial classifier identification model for associating characteristic components [ H, S, V, Cb and Cr ] of positive example colors with a first color gamut and associating characteristic components [ H, S, V, Cb and Cr ] of negative example colors with a second color gamut, wherein H is hue under an HSV color space, S is saturation under the HSV color space, V is brightness under the HSV color space, Cb is blue concentration offset under an YCbCr color space, and Cr is red concentration offset under the YCbCr color space;
and optimizing the initial classifier identification model to obtain an optimal classifier identification model.
3. The method of constructing a classifier identification model of claim 2 wherein the positive training image samples are images with multiple positive colors formed in a natural state; the counterexample training image sample adopts images with various counterexample colors formed in a natural state.
4. The method of claim 2, wherein a Support Vector Machine (SVM) algorithm is used to perform classification training on the classified training image samples to obtain an optimal classifier recognition model.
5. The method of constructing a classifier recognition model of claim 4 wherein the number of samples belonging to the first color gamut is: the number of samples belonging to the second color gamut is 1: 1.
6. the method of constructing a classifier identification model of claim 5 wherein the union of the first and second gamuts is equal to the gamut of the CCS hue circle, the union of the third and fourth gamuts is equal to the second gamut, and the intersection of the third and fourth gamuts is empty; the first color gamut corresponds to a red color region between the purple and orange of the CCS hue circle, the third color gamut corresponds to a blue color region between the cyan and ultramarine blue of the CCS hue circle, and the number of samples belonging to the third color gamut in the training image sample is as follows: the number of samples belonging to the fourth color gamut is 3.5-4.5: 1.
7. the computer-readable medium is characterized by comprising an input module, a color classification module and an output module, wherein the input module is used for acquiring image data of an adsorbent sample of a transformer respirator to be detected, the available state of the adsorbent sample can reflect the available state of all adsorbents of the transformer respirator to be detected, and the image data comprises color features [ H, S, V, Cb, Cr ] corresponding to each pixel point in an image of the adsorbent sample, wherein H is hue under HSV color space, S is saturation under HSV color space, V is brightness under HSV color space, Cb is blue concentration offset under YCbCr color space, and Cr is red concentration offset under YCbCr color space;
the color classification module is used for classifying whether the color of each pixel point in the image of the adsorbent sample belongs to a first color gamut, wherein the first color gamut corresponds to a color region when the adsorbent is in a failure state, or the first color gamut corresponds to a color region when the adsorbent is not in a failure state;
the output module is used for outputting at least two of the total number of pixel point colors in the image of the adsorbent sample belonging to the first color gamut, the total number of pixel point colors not belonging to the first color gamut and the total number of pixel points in the image of the adsorbent sample.
8. A transformer respirator monitoring system comprising a lower computer comprising a controller, wherein the controller has the computer readable medium having the transformer respirator detection program of claim 7 embedded therein, or is communicatively coupled to the computer readable medium having the transformer respirator detection program of claim 7.
9. The transformer respirator monitoring system of claim 8, further comprising an upper computer, wherein the lower computer further comprises a communication module for communicative connection with the upper computer, the communication module being communicatively connected with the controller.
10. The transformer respirator monitoring system of claim 8, further comprising a color identification module in communication with the controller, the color identification module being configured to be mounted at a location corresponding to a transformer respirator to be tested to identify color data of the sorbent of the transformer respirator to be tested; or the color camera module is in communication connection with the controller and used for being installed at a corresponding position of the transformer respirator to be detected so as to acquire color image data of the adsorbent of the transformer respirator to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010084873.0A CN111275124A (en) | 2020-02-10 | 2020-02-10 | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010084873.0A CN111275124A (en) | 2020-02-10 | 2020-02-10 | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111275124A true CN111275124A (en) | 2020-06-12 |
Family
ID=70999360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010084873.0A Pending CN111275124A (en) | 2020-02-10 | 2020-02-10 | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111275124A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112069886A (en) * | 2020-07-31 | 2020-12-11 | 许继集团有限公司 | Transformer substation respirator state intelligent identification method and system |
| CN114037887A (en) * | 2021-09-26 | 2022-02-11 | 佛山中科云图智能科技有限公司 | State detection method and device for transformer respirator, medium and electronic equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010182287A (en) * | 2008-07-17 | 2010-08-19 | Steven C Kays | Intelligent adaptive design |
| CN205301155U (en) * | 2015-12-14 | 2016-06-08 | 广东电网有限责任公司惠州供电局 | Novel main become response of discolouing of respirator silica gel device |
| CN106644963A (en) * | 2016-12-23 | 2017-05-10 | 贵州电网有限责任公司电力科学研究院 | Transformer respirator detection method based on color histogram statistic |
| CN107731474A (en) * | 2017-10-24 | 2018-02-23 | 周广新 | The Novel respirator and its automatic alarm prompt system of a kind of transformer |
| CN107832770A (en) * | 2017-11-08 | 2018-03-23 | 浙江国自机器人技术有限公司 | A kind of equipment routing inspection method, apparatus, system, storage medium and crusing robot |
| CN107977959A (en) * | 2017-11-21 | 2018-05-01 | 武汉中元华电科技股份有限公司 | A kind of respirator state identification method suitable for electric operating robot |
| CN110309843A (en) * | 2019-02-02 | 2019-10-08 | 国网浙江省电力有限公司湖州供电公司 | The automatic identifying method of multiclass component in a kind of power equipment image |
-
2020
- 2020-02-10 CN CN202010084873.0A patent/CN111275124A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010182287A (en) * | 2008-07-17 | 2010-08-19 | Steven C Kays | Intelligent adaptive design |
| CN205301155U (en) * | 2015-12-14 | 2016-06-08 | 广东电网有限责任公司惠州供电局 | Novel main become response of discolouing of respirator silica gel device |
| CN106644963A (en) * | 2016-12-23 | 2017-05-10 | 贵州电网有限责任公司电力科学研究院 | Transformer respirator detection method based on color histogram statistic |
| CN107731474A (en) * | 2017-10-24 | 2018-02-23 | 周广新 | The Novel respirator and its automatic alarm prompt system of a kind of transformer |
| CN107832770A (en) * | 2017-11-08 | 2018-03-23 | 浙江国自机器人技术有限公司 | A kind of equipment routing inspection method, apparatus, system, storage medium and crusing robot |
| CN107977959A (en) * | 2017-11-21 | 2018-05-01 | 武汉中元华电科技股份有限公司 | A kind of respirator state identification method suitable for electric operating robot |
| CN110309843A (en) * | 2019-02-02 | 2019-10-08 | 国网浙江省电力有限公司湖州供电公司 | The automatic identifying method of multiclass component in a kind of power equipment image |
Non-Patent Citations (1)
| Title |
|---|
| 张家齐: "重庆市兰家沱变电站视频监控技术研究及其应用", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112069886A (en) * | 2020-07-31 | 2020-12-11 | 许继集团有限公司 | Transformer substation respirator state intelligent identification method and system |
| CN114037887A (en) * | 2021-09-26 | 2022-02-11 | 佛山中科云图智能科技有限公司 | State detection method and device for transformer respirator, medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414334B (en) | Intelligent water quality identification method based on unmanned aerial vehicle inspection | |
| CN111275124A (en) | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system | |
| CN102297867A (en) | Detection system for assembly quality of wiring harness | |
| CN103324943A (en) | Identification method of complex device panel image multi-sub zone state | |
| CN101187684B (en) | Main board light-emitting diode detection device and method | |
| CN110346093A (en) | Oil-immersed sleeve pipe leakage of oil detection system and method based on intelligent patrol detection platform | |
| CN110738289A (en) | Multi-dimensional linkage comprehensive studying and judging device for electric power operation standardization and using method thereof | |
| CN111899211A (en) | Transformer oil leakage fault detection system and method based on images and smell | |
| CN111590639A (en) | Automatic light supplementing system and method for transformer substation inspection robot and inspection robot | |
| CN114445330A (en) | Method and system for detecting appearance defects of components | |
| CN109297978B (en) | Binocular imaging-based power line unmanned aerial vehicle inspection and defect intelligent diagnosis system | |
| CN115855840B (en) | Engine oil detection method and device | |
| CN107860559A (en) | Improve the intelligent AR testing agencies of product qualification rate | |
| CN105872340A (en) | Underwater camera | |
| WO2023109735A1 (en) | Image detection system and method for electrical connector sealing ring of battery pack, and battery swapping station | |
| CN109975312A (en) | A kind of cigarette bobbin paper quality detection device and method based on machine vision technique | |
| CN109635684A (en) | A kind of food traceability system | |
| CN215297341U (en) | Water body peculiar smell detection device | |
| CN206546417U (en) | A kind of GIS switch fault automatic recognition systems based on picture recognition | |
| CN116052041A (en) | Indicating lamp state identification method based on depth network | |
| CN104458736A (en) | Device for detecting pesticide residue and detection method | |
| CN209127817U (en) | A kind of elevator wire rope detector | |
| CN114910478A (en) | Visual interface instrument for chemical interface detection and detection method thereof | |
| CN114820616B (en) | Equipment state detection method and device for flashing mode indicator light | |
| CN107402110A (en) | The gas leak detection device of new energy car battery lid |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200612 |
|
| RJ01 | Rejection of invention patent application after publication |