CN113252207B - Electrical equipment metal surface temperature difference detection method and system - Google Patents
Electrical equipment metal surface temperature difference detection method and system Download PDFInfo
- Publication number
- CN113252207B CN113252207B CN202110798228.XA CN202110798228A CN113252207B CN 113252207 B CN113252207 B CN 113252207B CN 202110798228 A CN202110798228 A CN 202110798228A CN 113252207 B CN113252207 B CN 113252207B
- Authority
- CN
- China
- Prior art keywords
- temperature difference
- electrical equipment
- image
- metal surface
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01K—MEASURING TEMPERATURE; MEASURING QUANTITY OF HEAT; THERMALLY-SENSITIVE ELEMENTS NOT OTHERWISE PROVIDED FOR
- G01K11/00—Measuring temperature based upon physical or chemical changes not covered by groups G01K3/00, G01K5/00, G01K7/00 or G01K9/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Radiation Pyrometers (AREA)
Abstract
The invention discloses a method and a system for detecting the temperature difference of the metal surface of electrical equipment, which establish a visible light image dataset of the metal surface of the electrical equipment made of metal materials commonly used by the electrical equipment under the condition of reference temperature and according to the preset temperature rise step temperature rise condition, extract RGB pixel characteristics of pixels corresponding to different temperature areas in the image, combine the RGB pixel characteristics into a temperature difference characteristic dataset, compress and reduce the dimension of the temperature difference characteristic dataset to obtain a low-dimensional temperature difference characteristic dataset, input the low-dimensional temperature difference characteristic dataset into a temperature difference identification model based on a machine learning algorithm for training to obtain a trained temperature difference identification model, use the trained temperature difference identification model to detect the temperature difference of different areas of the same electrical equipment, solve the problems that the existing method for detecting the temperature rise fault of the electrical equipment has high cost, the measurement result is easily interfered by other infrared radiation sources and the measurement result needs to be artificially corrected, leading to technical problems of limited application.
Description
Technical Field
The invention relates to the technical field of electrical equipment safety detection, in particular to a method and a system for detecting temperature difference of a metal surface of electrical equipment.
Background
In an electric power system, most defects of power transmission, transformation and distribution equipment are accompanied by the phenomenon that the operating temperature of some areas of the electrical equipment is increased. The electric equipment in operation usually needs to carry out frequent temperature measurement operation, especially the temperature difference between different phases, and if there is great temperature difference between any two phases, there is a fault in the higher temperature phase probably, therefore, the temperature difference between different phases is the inspection item that inspection personnel paid particular attention to.
At present, the temperature difference detection mode of the electrical equipment is to use a thermal infrared imager to detect the temperature of the electrical equipment, but the thermal infrared imager is expensive and low in resolution, and the measurement result is easily interfered by other infrared radiation sources, so that inspection personnel with abundant operation and maintenance experience are required to correct the measurement result, and the application of the thermal infrared imager in the temperature rise fault detection of the electrical equipment is limited.
Disclosure of Invention
The invention provides a method and a system for detecting the temperature difference of the metal surface of electrical equipment, which are used for solving the technical problems that the existing method for detecting the temperature rise fault of the electrical equipment by using a thermal infrared imager has high cost, the measurement result is easily interfered by other infrared radiation sources, and the application is limited because the measurement result needs to be artificially corrected.
In view of the above, the first aspect of the present invention provides a method for detecting a temperature difference on a metal surface of an electrical device, including:
acquiring a visible light image dataset of the metal surface of the electrical equipment, wherein the visible light image dataset of the metal surface of the electrical equipment comprises: the method comprises the steps that a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment, which is increased by preset temperature rise step length on the basis of the reference temperature, are sequentially obtained;
performing RGB pixel feature extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray scale frequency distribution features, G component image gray scale frequency distribution features and B component image gray scale frequency distribution features in each image;
combining R component image gray frequency distribution characteristics, G component image gray frequency distribution characteristics and B component image gray frequency distribution characteristics of the first surface image and the second surface image respectively to obtain a temperature difference characteristic data set formed by temperature difference characteristics of each second surface image and the first surface image respectively, wherein the temperature difference characteristics carry temperature difference labels;
compressing and dimensionality reduction is carried out on temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data;
training a temperature difference recognition model constructed based on a machine learning algorithm according to a low-dimensional temperature difference characteristic data set to obtain a trained temperature difference recognition model;
the method comprises the steps of collecting a metal surface image of the electrical equipment to be detected in real time, extracting low-dimensional temperature difference characteristics from the metal surface image of the electrical equipment to be detected, inputting the low-dimensional temperature difference characteristics of the metal surface image of the electrical equipment to be detected into a trained temperature difference identification model, and obtaining a temperature difference detection result output by the trained temperature difference identification model.
Optionally, the electrical device metal surface visible light image dataset comprises: under different light source brightness, a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment with a preset temperature rise step length are sequentially increased on the basis of the reference temperature.
Optionally, the reference temperature is any one of values from 19 to 25 ℃.
Optionally, the preset temperature rise step is 1 ℃.
Optionally, the electrical device metal surface visible light image dataset includes a brass electrical device surface image dataset, an iron electrical device surface image dataset, and an aluminum alloy electrical device surface image dataset.
Optionally, the machine learning algorithm is any one of a K-nearest neighbor algorithm, a regression tree algorithm, a random forest regression algorithm, and a gradient boosting regression tree algorithm.
Optionally, the machine learning algorithm is a random forest regression algorithm.
Optionally, an automatic encoder is used to compress and reduce the dimension of the temperature difference characteristic data in the temperature difference characteristic data set.
Optionally, the number of the second surface images at each temperature rise step is the same, and is not less than 20.
The second aspect of the present invention provides a temperature difference detecting system for a metal surface of an electrical device, comprising: .
The device comprises a surface image data set acquisition module, a temperature rise step length acquisition module and a temperature rise step length acquisition module, wherein the surface image data set acquisition module is used for acquiring a visible light image data set of the metal surface of the electrical equipment, and the visible light image data set of the metal surface of the electrical equipment comprises a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment, which is increased by a preset temperature rise step length on the basis of the reference temperature;
the characteristic extraction module is used for performing RGB pixel characteristic extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray frequency distribution characteristics, G component image gray frequency distribution characteristics and B component image gray frequency distribution characteristics in each image;
the temperature difference characteristic data set construction module is used for respectively combining the R component image gray frequency distribution characteristic, the G component image gray frequency distribution characteristic and the B component image gray frequency distribution characteristic of the first surface image and the second surface image to obtain a temperature difference characteristic data set formed by the temperature difference characteristics of each second surface image and the first surface image, wherein the temperature difference characteristics carry temperature difference labels;
the dimension reduction module is used for compressing and reducing dimensions of the temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data;
the model training module is used for training a temperature difference recognition model constructed based on a machine learning algorithm according to the low-dimensional temperature difference characteristic data set to obtain a trained temperature difference recognition model;
the temperature difference detection module is used for acquiring the metal surface image of the electrical equipment to be detected in real time, extracting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected, and inputting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected into the trained temperature difference identification model to obtain the temperature difference detection result output by the trained temperature difference identification model.
According to the technical scheme, the embodiment of the invention has the following advantages:
the invention provides a method for detecting the temperature difference of the metal surface of electrical equipment, which utilizes the characteristic that the reflection characteristics of the metal surface to light at different temperatures are different, so that the illumination received by a camera imaging element is different, and finally the gray level of a digital image pixel output by a camera imaging system is changed along with the temperature of the metal surface, establishes a metal surface visible light image data set of the electrical equipment made of metal materials commonly used by the electrical equipment under the reference temperature condition and the preset temperature rise step temperature rise condition, extracts the RGB pixel characteristics of the corresponding pixels in different temperature areas in the image, combines the RGB pixel characteristics into a temperature difference characteristic data set, then compresses and reduces the dimension of the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set, trains through a temperature difference identification model of a machine learning algorithm to obtain a trained temperature difference identification model, and finally performs temperature difference detection on the electrical equipment by using the trained temperature difference identification model, the method does not need to rely on an infrared thermal imaging technology, does not need to manually correct the detection result, and solves the technical problems that the existing method for detecting the temperature rise fault of the electrical equipment by using the infrared thermal imager has high cost, the measurement result is easily interfered by other infrared radiation sources, and the measurement result needs to be manually corrected, so that the application is limited.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments described in the present invention, and it is obvious for those skilled in the art to obtain other drawings according to these drawings.
Fig. 1 is a schematic flow chart of a method for detecting a temperature difference on a metal surface of an electrical device according to an embodiment of the present invention;
FIG. 2 is a 19.0 ℃ copper plate image;
FIG. 3 is a 99.0 ℃ copper plate image;
FIG. 4 shows the RGB grayscale frequency distribution of a 19.0 ℃ copper plate image;
FIG. 5 shows RGB grayscale frequency distribution of a 99.0 deg.C copper plate image;
FIG. 6 is a schematic diagram of a temperature difference data construction process;
FIG. 7 is a schematic diagram of an autoencoder;
FIG. 8 is a schematic diagram showing the decrease of the loss function in the 300 rounds of training;
FIG. 9 is a diagram of the results of the high-dimensional raw data temperature difference identification part of four machine learning algorithms;
FIG. 10 is a diagram of the results of the low-dimensional raw data temperature difference identification part of four machine learning algorithms.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For easy understanding, referring to fig. 1, the present invention provides an embodiment of a method for detecting a temperature difference on a metal surface of an electrical device, including:
step 101, acquiring a visible light image data set of the metal surface of the electrical equipment, wherein the visible light image data set of the metal surface of the electrical equipment comprises: the method comprises the steps of sequentially increasing a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment with a preset temperature rise step length on the basis of the reference temperature.
In the embodiment of the invention, firstly, the visible light image data set of the metal surface of the electrical equipment is constructed, and the common material of the electrical equipment is brass, iron or aluminum alloy, so that the visible light image data set of the metal surface of the electrical equipment can comprise an image data set of the electrical equipment surface made of brass, an image data set of the electrical equipment surface made of iron and an image data set of the electrical equipment surface made of aluminum alloy. The visible light image data set of the metal surface of the electrical equipment comprises a first surface image of the metal surface of the electrical equipment, which is acquired by a shooting device at a reference temperature, wherein the normal operation temperature of the electrical equipment is generally 10-50 ℃. And then sequentially increasing a second surface image of the metal surface of the electrical equipment with a preset temperature rise step length on the basis of the reference temperature, namely increasing the temperature of the preset temperature rise step length per liter, and collecting a primary visible light image data set of the metal surface of the electrical equipment. The preset temperature rise step length is preferably 1 ℃, the data volume is increased rapidly due to the too small temperature rise step length, the data collection is difficult, and the temperature measurement is inaccurate due to the too large temperature rise step length, so that the optimal balance can be achieved by selecting the 1 ℃ temperature rise step length after a large number of experimental verifications. The temperature of the electrical equipment is raised from a reference temperature (which may be selected to be 19.0 ℃), and when the temperature of the electrical equipment measured by the thermocouple thermometer rises by 1.0 ℃ every time and stabilizes at the current value, a group of pictures (N =20) of the same number N are taken using a camera (which may be a high-performance digital camera or a camera phone). It should be noted that the metal surface of the electrical device in the present invention may be the surface of two different electrical devices, or may be different areas of the surface of the same electrical device, and when the metal surface is the surface of different electrical devices, it is necessary to ensure that the material of the surface of the electrical device is consistent.
In one embodiment, considering the anti-interference capability of the temperature difference identification model to the external illumination change, the visible light image data sets of the metal surface of the electrical equipment are included under different light source brightness, a first surface image of the metal surface of the electrical device at a reference temperature and a second surface image of the metal surface of the electrical device sequentially increasing a preset temperature rise step length on the basis of the reference temperature, namely, a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment with a preset temperature rise step length are sequentially increased on the basis of the reference temperature under different light source brightness, for example, a white light source capable of adjusting brightness is used for illumination, the brightness adjustment steps are four steps 1,2,3, and 4, and the brightness corresponding to each step is from low to high, so that a first surface image and a second surface image corresponding to the brightness of at least 2 steps of the four brightness steps need to be acquired. The light source can be selected as ambient light, and the visible light image data acquisition under the ambient light condition is closer to practical application.
102, performing RGB pixel feature extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray frequency distribution features, G component image gray frequency distribution features and B component image gray frequency distribution features in each image.
For the acquired surface image, areas corresponding to metal surfaces with different temperatures in the image can be cut out, the pixel size of the cut-out area can be selected to be 400 × 400, as shown in fig. 2 and 3, the picture cut-out areas corresponding to an a plate (copper material) with a temperature of 19.0 ℃ (reference temperature) and a B plate (copper material) with a temperature of 99.0 ℃ (temperature raised on the basis of reference temperature) in fig. 2 and 3 respectively, and image features are respectively extracted from pixels corresponding to the two areas. According to the theory of colorimetry, all colors can be synthesized with R, G, B three colors, which are called three primary colors. The visible light image is stored in the computer in the form of an array of m × n × 3, where "3" can be understood as three m × n two-dimensional grayscale images, representing an R component image, a G component image, and a B component image, respectively. Therefore, R, G, B gray scale distributions of all pixel points can be extracted from the visible light image, reflecting the color status, wherein the gray scale level of each primary color is in the range of (0,255). After the gray values of three primary colors of all pixel points m x n in the region are extracted, the number of pixels of a certain gray level of a certain primary color in the region can be obtained through calculation, and the ratio of the number of pixels of the certain primary color to the total number of pixel points in the region is the frequency of the certain gray level of the certain primary color, so that the gray frequency distribution of the three primary colors of the visible light image is obtained, as shown in fig. 4 and 5, fig. 4 and 5 are RGB gray distribution histograms corresponding to a panel a at 19.0 ℃ and a panel B at 99.0 ℃. When the temperature of the object changes, the reflectivity of the visible light changes along with the temperature change, so that the visible light image shot by the camera device also changes along with the temperature change, and the three-primary-color recovery frequency distribution contains temperature information. As can be seen from fig. 4 and 5, the RGB gray scale frequency distribution curves of the metal plates at different temperatures are changed.
103, combining the R component image gray frequency distribution characteristics, the G component image gray frequency distribution characteristics and the B component image gray frequency distribution characteristics of the first surface image and the second surface image respectively to obtain a temperature difference characteristic data set formed by the temperature difference characteristics of each second surface image and the first surface image, wherein the temperature difference characteristics carry temperature difference labels.
Feature vectors of an A plate image (temperature is recorded as T1) and a B plate image (temperature is recorded as T2) corresponding to the images in the figures 2 and 3 are combined to be used as features, and a temperature difference (T1-T2) is used as a label to form a temperature difference feature data set. Considering that the high-low order of T1 and T2 may not be decided in advance in actual operation, when constructing the temperature difference feature data set, the image features corresponding to T1 and T2 are set to be arranged in a random order in the temperature difference feature, as shown in fig. 6. Thus, the temperature difference recognition result may be positive and negative, and positive indicates that T1 is higher than T2, and negative indicates the opposite.
And step 104, compressing and dimensionality reduction is carried out on the temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data.
Because the original feature data of the first surface image (the a-plate image in fig. 2) and the second surface image (the B-plate image in fig. 3) each include 256 × 3= 768-dimensional data, and the temperature difference feature data combines the original feature data of both the first surface image and the second surface image, the original temperature difference features total 768 × 2= 1536-dimensional features, which have too large dimensions, seriously affect the calculation efficiency, consume large calculation resources and storage space, include more useless information and noise, and easily cause overfitting when the number of samples is small. Schematic diagram of the automatic encoder as shown in fig. 7, the encoder part first reduces the dimension of the input data to a code through a neural network, then decodes the code through a decoder to obtain an output which is as similar as possible to the original input data, and updates the network parameters by comparing the input and output data and minimizing the difference between them. The information is inevitably lost in the data dimension reduction process, but the main characteristics of the data can still be kept, which is the principle that the automatic encoder can filter out noise and useless information in the data. Here, use is made of:
1536 → ReLU → 1000 → ReLU → 700 → ReLU → 500 → ReLU → 300 → 500 → ReLU → 700 → ReLU → 1000 → ReLU → 1536 structure of the fully connected neural network constructs the automatic encoder, the ReLU is a nonlinear activation function, which can effectively prevent the gradient disappearance problem in the network parameter updating process. The loss function uses Mean Square Error (MSE). The number of training rounds is 300. The initial learning rate was set to 0.001, and then the learning rate was adjusted to 0.1 times the previous one every 100 rounds of training. And (3) adopting an Early Stopping (Early Stopping) strategy, and storing the model parameters which can enable the loss function to reach the minimum value in the training process. Fig. 8 shows the process of loss function reduction during 300 rounds of training.
And finally, storing the encoder part of the trained network, taking the original high-dimensional characteristics as the input of the encoder, wherein the output of the encoder is the compression characteristics of each piece of data, so that the low-dimensional temperature difference characteristics are obtained, and the dimension is 300. And each piece of original data is processed by an encoder, and the characteristics after dimensionality reduction are stored to form a low-dimensional temperature difference characteristic set for training a temperature difference identification model.
And 105, training a temperature difference recognition model constructed based on a machine learning algorithm according to the low-dimensional temperature difference characteristic data set to obtain the trained temperature difference recognition model.
And 106, acquiring a metal surface image of the electrical equipment to be detected in real time, extracting a low-dimensional temperature difference characteristic from the metal surface image of the electrical equipment to be detected, and inputting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected into the trained temperature difference identification model to obtain a temperature difference detection result output by the trained temperature difference identification model.
The temperature difference recognition model adopts a temperature difference recognition model based on a machine learning algorithm to compress and reduce the dimensions of original high-dimensional data to obtain low-dimensional temperature difference data, and low-dimensional temperature difference characteristic data sets corresponding to the visible light image data sets of the metal surface of the electrical equipment are respectively used for training the temperature difference recognition model to obtain a trained temperature difference recognition model. The method comprises the steps of performing temperature difference detection on the metal surface of the electrical equipment by using a trained temperature difference identification model, acquiring an image of the metal surface of the electrical equipment to be detected in real time in a mode of acquiring a visible light image dataset of the metal surface of the electrical equipment in the step 101, then processing the image according to the modes of the steps 102 to 105, extracting low-dimensional temperature difference characteristics, inputting the low-dimensional temperature difference characteristics of the image of the metal surface of the electrical equipment to be detected into the trained temperature difference identification model, and obtaining a temperature difference detection result output by the trained temperature difference identification model.
The method for detecting the temperature difference of the metal surface of the electrical equipment provided by the embodiment of the invention utilizes the characteristic that the reflection characteristics of the metal surface to light at different temperatures are different, so that the illumination received by a camera imaging element is different, and finally the gray level of a digital image pixel output by a camera imaging system is changed along with the temperature of the metal surface, establishes a data set of visible light images of the metal surface of the electrical equipment made of metal materials commonly used by the electrical equipment under the reference temperature condition and the preset temperature rise step temperature rise condition, extracts RGB pixel characteristics of pixels corresponding to different temperature areas in the images, combines the RGB pixel characteristics into a temperature difference characteristic data set, then compresses and reduces the dimension of the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set, trains through a temperature difference identification model of a machine learning algorithm to obtain a trained temperature difference identification model, and finally performs temperature difference detection on the electrical equipment by using the trained temperature difference identification model, the method does not need to rely on an infrared thermal imaging technology, does not need to manually correct the detection result, and solves the technical problems that the existing method for detecting the temperature rise fault of the electrical equipment by using the infrared thermal imager has high cost, the measurement result is easily interfered by other infrared radiation sources, and the measurement result needs to be manually corrected, so that the application is limited.
In one embodiment, the machine learning algorithm adopted in the present invention is any one of a K-Nearest Neighbor (kNN), a Regression Tree algorithm (RT), a Random Forest Regression algorithm (RFR), and a Gradient Boosting Regression Tree algorithm (GBRT). A random forest regression algorithm is preferred. The method comprises the steps of respectively taking an original temperature difference characteristic (namely high-dimensional temperature difference characteristic data) and a low-dimensional temperature difference characteristic as inputs, taking the temperature difference as a label, and modeling the mapping relation between the temperature difference characteristic and the temperature difference label by an algorithm through minimizing the difference between the output of a model and an actual label. The Mean Absolute Error (MAE) of the final model on the test set was used as the criterion for the model performance. Comparing the four algorithms, in which the high-dimensional original feature and the low-dimensional feature are used as input, the accuracy of identifying the temperature difference is shown in tables 1 and 2, and partial results of identifying the temperature difference by using the high-dimensional original feature and the low-dimensional feature (taking copper plate temperature difference identification as an example) are shown in fig. 9 and 10.
TABLE 1 prediction of temperature difference results from high dimensional temperature difference characterization data
TABLE 2 Low-dimensional temperature difference characteristic data prediction temperature difference results
As can be seen from the data in tables 1 and 2 and fig. 9 and 10:
1) the temperature difference between common metal surfaces with different temperatures can be effectively identified by using dual image characteristics (namely combining the characteristics of the first surface image and the second surface image);
2) the low-dimensional features after dimension reduction are compressed by using an automatic encoder, and compared with the original high-dimensional features, the low-dimensional features have higher identification precision;
3) of the four algorithms, the machine learning algorithm based on random forest regression has the highest prediction precision.
The invention also provides an embodiment of the electrical equipment metal surface temperature difference detection system, which comprises the following components:
the surface image data set acquisition module is used for acquiring a metal surface visible light image data set of the electrical equipment, and the metal surface visible light image data set of the electrical equipment comprises: the method comprises the steps that a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment, which is increased by preset temperature rise step length on the basis of the reference temperature, are sequentially obtained;
the characteristic extraction module is used for performing RGB pixel characteristic extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray frequency distribution characteristics, G component image gray frequency distribution characteristics and B component image gray frequency distribution characteristics in each image;
the temperature difference characteristic data set construction module is used for respectively combining the R component image gray frequency distribution characteristic, the G component image gray frequency distribution characteristic and the B component image gray frequency distribution characteristic of the first surface image and the second surface image to obtain a temperature difference characteristic data set formed by the temperature difference characteristics of each second surface image and the first surface image, wherein the temperature difference characteristics carry temperature difference labels;
the dimension reduction module is used for compressing and reducing dimensions of the temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data;
the model training module is used for training a temperature difference recognition model constructed based on a machine learning algorithm according to the low-dimensional temperature difference characteristic data set to obtain a trained temperature difference recognition model;
the temperature difference detection module is used for acquiring the metal surface image of the electrical equipment to be detected in real time, extracting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected, and inputting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected into the trained temperature difference identification model to obtain the temperature difference detection result output by the trained temperature difference identification model.
It should be noted that the embodiment of the system for detecting a temperature difference on a metal surface of an electrical device in the embodiment of the present invention is a system embodiment corresponding to the embodiment of the method for detecting a temperature difference on a metal surface of an electrical device, and the method in the embodiment of the method for detecting a temperature difference on a metal surface of an electrical device can achieve the same technical effects as the embodiment of the method for detecting a temperature difference on a metal surface of an electrical device, and will not be described herein again.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present 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 solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (8)
1. A method for detecting temperature difference of metal surface of electrical equipment is characterized by comprising the following steps:
acquiring a visible light image dataset of the metal surface of the electrical equipment, wherein the visible light image dataset of the metal surface of the electrical equipment comprises: the method comprises the steps that a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment, which is increased by preset temperature rise step length on the basis of the reference temperature, are sequentially obtained;
performing RGB pixel feature extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray scale frequency distribution features, G component image gray scale frequency distribution features and B component image gray scale frequency distribution features in each image;
combining R component image gray frequency distribution characteristics, G component image gray frequency distribution characteristics and B component image gray frequency distribution characteristics of the first surface image and the second surface image respectively to obtain a temperature difference characteristic data set formed by temperature difference characteristics of each second surface image and the first surface image respectively, wherein the temperature difference characteristics carry temperature difference labels;
compressing and dimensionality reduction is carried out on temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data;
training a temperature difference recognition model constructed based on a machine learning algorithm according to a low-dimensional temperature difference characteristic data set to obtain a trained temperature difference recognition model, wherein the machine learning algorithm is a random forest regression algorithm;
the method comprises the steps of collecting a metal surface image of the electrical equipment to be detected in real time, extracting low-dimensional temperature difference characteristics from the metal surface image of the electrical equipment to be detected, inputting the low-dimensional temperature difference characteristics of the metal surface image of the electrical equipment to be detected into a trained temperature difference identification model, and obtaining a temperature difference detection result output by the trained temperature difference identification model.
2. The method for detecting the temperature difference of the metal surface of the electrical equipment according to claim 1, wherein the visible light image data set of the metal surface of the electrical equipment comprises: under different light source brightness, a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment with a preset temperature rise step length are sequentially increased on the basis of the reference temperature.
3. The method for detecting the temperature difference of the metal surface of the electrical equipment according to claim 1, wherein the reference temperature is any one value of 19-25 ℃.
4. The method for detecting the temperature difference on the metal surface of the electrical equipment according to claim 1, wherein the preset temperature rise step is 1 ℃.
5. The method for detecting the temperature difference on the metal surface of the electrical equipment according to claim 1, wherein the visible light image dataset on the metal surface of the electrical equipment comprises a brass electrical equipment surface image dataset, an iron electrical equipment surface image dataset and an aluminum alloy electrical equipment surface image dataset.
6. The method for detecting the temperature difference on the metal surface of the electrical equipment according to claim 1, wherein an automatic encoder is adopted to compress and reduce the dimension of the temperature difference characteristic data in the temperature difference characteristic data set.
7. The method for detecting the temperature difference of the metal surface of the electrical equipment according to claim 1, wherein the number of the second surface images in each temperature rise step is the same, and is not less than 20.
8. The utility model provides an electrical equipment metal surface difference in temperature detecting system which characterized in that includes:
the device comprises a surface image data set acquisition module, a temperature rise step length acquisition module and a temperature rise step length acquisition module, wherein the surface image data set acquisition module is used for acquiring a visible light image data set of the metal surface of the electrical equipment, and the visible light image data set of the metal surface of the electrical equipment comprises a first surface image of the metal surface of the electrical equipment at a reference temperature and a second surface image of the metal surface of the electrical equipment, which is increased by a preset temperature rise step length on the basis of the reference temperature;
the characteristic extraction module is used for performing RGB pixel characteristic extraction on a first surface image and a second surface image in an electrical equipment metal surface visible light image data set to obtain R component image gray frequency distribution characteristics, G component image gray frequency distribution characteristics and B component image gray frequency distribution characteristics in each image;
the temperature difference characteristic data set construction module is used for respectively combining the R component image gray frequency distribution characteristic, the G component image gray frequency distribution characteristic and the B component image gray frequency distribution characteristic of the first surface image and the second surface image to obtain a temperature difference characteristic data set formed by the temperature difference characteristics of each second surface image and the first surface image, wherein the temperature difference characteristics carry temperature difference labels;
the dimension reduction module is used for compressing and reducing dimensions of the temperature difference characteristic data in the temperature difference characteristic data set to obtain a low-dimensional temperature difference characteristic data set formed by the low-dimensional temperature difference characteristic data;
the model training module is used for training a temperature difference recognition model constructed based on a machine learning algorithm according to the low-dimensional temperature difference characteristic data set to obtain the trained temperature difference recognition model, and the machine learning algorithm is a random forest regression algorithm;
the temperature difference detection module is used for acquiring the metal surface image of the electrical equipment to be detected in real time, extracting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected, and inputting the low-dimensional temperature difference characteristic of the metal surface image of the electrical equipment to be detected into the trained temperature difference identification model to obtain the temperature difference detection result output by the trained temperature difference identification model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110798228.XA CN113252207B (en) | 2021-07-15 | 2021-07-15 | Electrical equipment metal surface temperature difference detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110798228.XA CN113252207B (en) | 2021-07-15 | 2021-07-15 | Electrical equipment metal surface temperature difference detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113252207A CN113252207A (en) | 2021-08-13 |
| CN113252207B true CN113252207B (en) | 2021-11-02 |
Family
ID=77180363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110798228.XA Active CN113252207B (en) | 2021-07-15 | 2021-07-15 | Electrical equipment metal surface temperature difference detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113252207B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116543354A (en) * | 2023-06-27 | 2023-08-04 | 南方电网数字电网研究院有限公司 | Infrared image wire clip overheating identification method and device and computer equipment |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102007048970A1 (en) * | 2007-10-12 | 2009-04-23 | Siemens Ag | B0 field drift correction in a magnetic resonance tomographic temperature chart |
| CN107784648B (en) * | 2017-10-18 | 2021-08-24 | 广东正业科技股份有限公司 | Position compensation system and method of visual positioning device |
| CN108652627A (en) * | 2018-03-13 | 2018-10-16 | 安徽锐捷信息科技有限公司 | A kind of magnetic resonance temperature imaging method and device |
| WO2020124316A1 (en) * | 2018-12-17 | 2020-06-25 | SZ DJI Technology Co., Ltd. | Image processing method and apparatus |
| CN109767441B (en) * | 2019-01-15 | 2021-11-02 | 电子科技大学 | Automatic detection blind pixel marking method |
| CN110160661B (en) * | 2019-05-28 | 2020-08-18 | 华中科技大学 | Object surface temperature measuring method and device based on visible light photo |
| CN112053341A (en) * | 2020-09-01 | 2020-12-08 | 广东电网有限责任公司 | Surface temperature measurement method of alloy material based on light reflectivity and artificial intelligence |
| CN112633292A (en) * | 2020-09-01 | 2021-04-09 | 广东电网有限责任公司 | Method for measuring temperature of oxide layer on metal surface |
-
2021
- 2021-07-15 CN CN202110798228.XA patent/CN113252207B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113252207A (en) | 2021-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108875619B (en) | Video processing method and device, electronic equipment and computer readable storage medium | |
| CN107909562B (en) | Fast image fusion algorithm based on pixel level | |
| CN107123091B (en) | Near-infrared face image super-resolution reconstruction method based on deep learning | |
| JP2014515587A (en) | Learning image processing pipelines for digital imaging devices | |
| CN105282548B (en) | Ram ratio selects | |
| KR20050074590A (en) | Light source estimation apparatus, light source estimation method, and image-pickup apparatus, and image processing method | |
| CN109283439B (en) | Discharge state identification method based on three-primary-color chrominance information and machine learning | |
| US9519839B2 (en) | Illumination estimation using natural scene statistics | |
| JP2007183872A (en) | Dynamic camera color correction device and video retrieving device using the same | |
| CN113850324B (en) | Multispectral target detection method based on Yolov4 | |
| CN113252207B (en) | Electrical equipment metal surface temperature difference detection method and system | |
| Hussain et al. | Max-RGB based colour constancy using the sub-blocks of the image | |
| US8897378B2 (en) | Selective perceptual masking via scale separation in the spatial and temporal domains using intrinsic images for use in data compression | |
| CN110796592B (en) | Storage method of high dynamic range spectral image data | |
| Tominaga et al. | Scene illuminant estimation of multiple light sources | |
| CN117078608B (en) | Double-mask guide-based high-reflection leather surface defect detection method | |
| Wen | Image fusion based on improved IHS transform with weighted average | |
| Kanj | Flicker removal and color correction for high speed videos | |
| Heng et al. | MatSpectNet: Material Segmentation Network with Domain-Aware and Physically-Constrained Hyperspectral Reconstruction | |
| Rushdi et al. | Color de-rendering using coupled dictionary learning | |
| CN114240776B (en) | Demosaicing and compression fusion framework for MSFA hyperspectral image | |
| Makwana et al. | Efficient color transfer method based on colormap clustering for night vision applications | |
| Daly et al. | Spatiochromatic and temporal natural image statistics modelling: Applications from display analysis to neural networks | |
| CN116188797B (en) | Scene light source color estimation method capable of being effectively embedded into image signal processor | |
| CN112381784B (en) | Equipment detecting system based on multispectral image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |