CN114463299A - Infrared image detection method for wall bushing - Google Patents
Infrared image detection method for wall bushing Download PDFInfo
- Publication number
- CN114463299A CN114463299A CN202210090989.4A CN202210090989A CN114463299A CN 114463299 A CN114463299 A CN 114463299A CN 202210090989 A CN202210090989 A CN 202210090989A CN 114463299 A CN114463299 A CN 114463299A
- Authority
- CN
- China
- Prior art keywords
- wall bushing
- infrared image
- model
- data
- detection method
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Radiation Pyrometers (AREA)
Abstract
The invention discloses an infrared image detection method of a wall bushing, which comprises the steps of collecting temperature visual data of an infrared image of the wall bushing, sorting and labeling to prepare a data set; dividing the data set into a training set, a verification set and a test set, introducing a deep learning model which is built, trained and feature-fused by a keras framework to obtain a target detection model of the through-wall bushing equipment, and processing an infrared image of the through-wall bushing to be detected by adopting the target detection model to obtain a detection result of the equipment position and the equipment category of the through-wall bushing. The invention has the characteristics of accurate positioning and high detection accuracy.
Description
Technical Field
The invention relates to a detection method, in particular to an infrared image detection method of a wall bushing.
Background
The infrared temperature measurement technology is to utilize a thermal infrared imager to detect infrared radiation emitted by an object, and obtain a visual temperature distribution map of the equipment from an invisible radiation image through a series of conversion processing, so as to realize the detection of abnormal temperature points or abnormal temperature distribution of the power transformation equipment. In recent years, infrared thermometry diagnosis technology has become mature and widely used in various industries. When the technology is used for transformer substation inspection, large-area rapid scanning is generally carried out on equipment to be inspected firstly, the overall heating condition of the equipment is monitored, and then suspected points found in rapid inspection are accurately detected, so that the equipment faults are accurately inspected. The infrared thermal imager detection has the advantages of non-contact uninterrupted power detection of power equipment and the like, is widely used in each power unit, and is applied and popularized along with intelligent equipment such as intelligent inspection and fixed-point monitoring of a transformer substation, so that the conventional method for detecting the power equipment mainly takes infrared images as a research basis, and the problem of abnormal heating of the power equipment is researched by accurately positioning the position of the power equipment. However, due to the fact that the infrared picture contains various pseudo colors and the shooting environment is complex, the problems that the interference of the shot equipment is serious, the types of training data are few and the like are caused, and therefore the existing method has the problems that the positioning of the detection of the power equipment is not accurate enough, the accuracy of subsequent feature extraction is low, and the detection accuracy is low.
Disclosure of Invention
The invention aims to provide an infrared image detection method of a wall bushing. The invention has the characteristics of accurate positioning and high detection accuracy.
The technical scheme of the invention is as follows: an infrared image detection method of a wall bushing comprises the following steps:
s1, collecting data: collecting temperature visual data of an infrared image of the wall bushing, sorting and marking to prepare a data set;
s2, dividing the data set into a training set, a verification set and a test set;
s3, building a Faster R-CNN detection network model: adopting a keras framework, selecting a MobilenetV2 as a main feature extraction network, adopting a training set on ImageNet for pre-training, fusing the last three features with different sizes, enhancing feature extraction, calculating loss by using smooth _ l1, optimizing network parameters by Adam, simultaneously adopting an Early stopping technology to prevent overfitting of the model, adopting a verification set to verify the effectiveness of the model, adopting a test set to test, and evaluating the result of the model to obtain a target detection model of the wall bushing equipment;
s4, detection: and processing the infrared image of the wall bushing to be detected by adopting a target detection model, inputting unknown temperature visual data into the target detection model, and obtaining a detection result of the equipment position and the equipment category of the wall bushing.
In the foregoing method for detecting an infrared image of a wall bushing, in step S1, the data of the wall bushing is categorized into a short and fat wall bushing and a long and thin wall bushing according to the appearance.
In the infrared image detection method for the wall bushing, 5000 pieces of temperature visual data are collected at will for each category, and the temperature visual data are subjected to standardization processing to obtain the temperature visual data with the input model size of 640 × 480 × 1.
In the foregoing infrared image detection method for a wall bushing, in step S1, the labels are to label the category of the two category data and the coordinates of the target frame, and to train the fast R-CNN detection network model.
In the foregoing method for detecting an infrared image of a wall bushing, in step S2, the data set is divided into a training set, a verification set, and a test set according to a data ratio of 5:1: 4.
In the method for detecting the infrared image of the through-wall bushing, in the model training process of step S3, the feature extraction is specifically enhanced by fusing the last three effective features of different sizes of MobilenetV2 with an FPN network.
In the foregoing method for detecting an infrared image of a wall bushing, in step S3, the pre-training is performed for 300 iterations, the previous 100 iterations are performed, the size of Batch size is 32, and the initial learning rate is set to 1 e-3; after 200 iterations, the Batch size was taken to 16 and the initial learning rate was set to 1 e-4.
In the foregoing infrared image detection method for a wall bushing, in step S4, the mapp of the target detection model is 98.3%, and the average IoU is 86.2%.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, a plurality of pieces of temperature visual data are collected from the infrared image and used as detection data, the through-wall bushing is classified and feature fused, so that the mAP detected by the model reaches 98.3%, the Avg _ IoU reaches 86.2%, the positioning and detection effectiveness of the model is high, and the result accuracy of the position detection of the through-wall bushing equipment is improved.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are not to be construed as limiting the invention.
Example (b):
as shown in fig. 1, an infrared image detection method for a wall bushing includes the following steps:
s1, collecting data: collecting temperature visual data of an infrared image of the wall bushing, sorting and marking to prepare a data set; the arrangement is classified into two categories, namely a short and fat wall bushing and a long and thin wall bushing according to appearance forms. The length of the dwarf wall bushing is not more than the width; the length of the slender wall bushing is larger than the width. 5000 pieces of temperature visual data are randomly collected in each category, and the temperature visual data are subjected to standardization processing to obtain temperature visual data with the size of an input model being 640 x 480 x 1; and marking the category of the two category data and the coordinates of the target box for training the Faster R-CNN detection network model.
And S2, dividing the data set into a training set, a verification set and a test set according to the data ratio of 5:1: 4.
S3, building a Faster R-CNN detection network model: selecting a keyword frame, selecting a MobilenetV2 as a trunk feature extraction network, and pre-training on ImageNet by adopting a training set, wherein the pre-training is performed for 300 iterations, the previous 100 iterations are performed, the size of Batch size is 32, and the initial learning rate is set to be 1 e-3; after 200 times of iteration, the size of the Batch size is 16, and the initial learning rate is set to be 1 e-4; and fusing the last three effective features of the MobilenetV2 with different sizes by using an FPN network, enhancing feature extraction, calculating loss by using smooth _ l1, optimizing network parameters by Adam, preventing overfitting of the model by using Early stopping technology, verifying the effectiveness of the model by using a verification set, testing by using a test set, and evaluating a model result to obtain a target detection model of the through-wall casing equipment.
S4, detection: and processing the infrared image of the wall bushing to be detected by adopting a target detection model, inputting unknown temperature visual data into the target detection model, and obtaining a detection result of the equipment position and the equipment category of the wall bushing. The mAP of the target detection model is 98.3%, and the average IoU is 86.2%. And detecting the temperature difference between the equipment and the ambient temperature.
Detection contrast test of the target detection model:
taking 3,000 pieces of test data, and carrying out detection by adopting different methods and models, wherein the detection results are shown in table 1.
TABLE 1 comparison of wall bushing test results
Therefore, the target detection model and the detection method improve the accuracy of the positioning detection result of the wall bushing.
Claims (8)
1. An infrared image detection method of a wall bushing is characterized in that: the method comprises the following steps:
s1, collecting data: collecting temperature visual data of an infrared image of the wall bushing, sorting and marking to prepare a data set;
s2, dividing the data set into a training set, a verification set and a test set;
s3, building a Faster R-CNN detection network model: adopting a keras framework, selecting a MobilenetV2 as a main feature extraction network, adopting a training set on ImageNet for pre-training, fusing the last three features with different sizes, enhancing feature extraction, calculating loss by using smooth _ l1, optimizing network parameters by Adam, simultaneously adopting an Early stopping technology to prevent overfitting of the model, adopting a verification set to verify the effectiveness of the model, adopting a test set to test, and evaluating the result of the model to obtain a target detection model of the wall bushing equipment;
s4, detection: and processing the infrared image of the wall bushing to be detected by adopting a target detection model, inputting unknown temperature visual data into the target detection model, and obtaining a detection result of the equipment position and the equipment category of the wall bushing.
2. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: in step S1, the data of the wall bushing are sorted into two categories, namely a short and fat wall bushing and a long and thin wall bushing, according to the appearance.
3. The infrared image detection method of the through-wall bushing according to claim 2, characterized in that: 5000 pieces of temperature visual data are collected randomly in each category, and the temperature visual data are subjected to standardization processing to obtain the temperature visual data with the input model size of 640 x 480 x 1.
4. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: in the step S1, the labels are to label the categories of the two category data and the coordinates of the target box, and to train the Faster R-CNN detection network model.
5. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: the step S2 is to divide the data set into a training set, a verification set and a test set according to a data ratio of 5:1: 4.
6. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: in the model training process of step S3, the enhanced feature extraction specifically includes fusing the last three effective features of different sizes of MobilenetV2 by using an FPN network, and enhancing feature extraction.
7. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: in the step S3, the pre-training is performed for 300 iterations, the value of the Batch size is 32 for the first 100 iterations, and the initial learning rate is set to 1 e-3; after 200 iterations, the Batch size was taken to 16 and the initial learning rate was set to 1 e-4.
8. The infrared image detection method of the through-wall bushing according to claim 1, characterized in that: in step S4, the mapp of the target detection model is 98.3%, and the average IoU is 86.2%.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210090989.4A CN114463299A (en) | 2022-01-26 | 2022-01-26 | Infrared image detection method for wall bushing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210090989.4A CN114463299A (en) | 2022-01-26 | 2022-01-26 | Infrared image detection method for wall bushing |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114463299A true CN114463299A (en) | 2022-05-10 |
Family
ID=81411421
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210090989.4A Pending CN114463299A (en) | 2022-01-26 | 2022-01-26 | Infrared image detection method for wall bushing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114463299A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115587975A (en) * | 2022-09-28 | 2023-01-10 | 国网湖北省电力有限公司超高压公司 | Oil-filled equipment casing dirt deposition defect diagnosis system, method and medium |
CN116383612A (en) * | 2023-06-07 | 2023-07-04 | 浙江天铂云科光电股份有限公司 | Detection complement method for power equipment component frame based on temperature data |
-
2022
- 2022-01-26 CN CN202210090989.4A patent/CN114463299A/en active Pending
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115587975A (en) * | 2022-09-28 | 2023-01-10 | 国网湖北省电力有限公司超高压公司 | Oil-filled equipment casing dirt deposition defect diagnosis system, method and medium |
CN116383612A (en) * | 2023-06-07 | 2023-07-04 | 浙江天铂云科光电股份有限公司 | Detection complement method for power equipment component frame based on temperature data |
CN116383612B (en) * | 2023-06-07 | 2023-09-01 | 浙江天铂云科光电股份有限公司 | Detection complement method for power equipment component frame based on temperature data |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
CN114463299A (en) | Infrared image detection method for wall bushing | |
CN109544533B (en) | Metal plate defect detection and measurement method based on deep learning | |
CN112200178B (en) | Transformer substation insulator infrared image detection method based on artificial intelligence | |
CN110992349A (en) | Underground pipeline abnormity automatic positioning and identification method based on deep learning | |
CN110648305B (en) | Industrial image detection method, system and computer readable recording medium | |
CN107562696B (en) | On-line detection and control method for tire product quality | |
CN111624229A (en) | Intelligent charged equipment fault diagnosis method based on infrared imaging | |
CN105954412B (en) | For the sensor array optimization method of hickory nut freshness detection | |
CN111860593B (en) | Fan blade fault detection method based on generation countermeasure network | |
CN111008641B (en) | Power transmission line tower external force damage detection method based on convolutional neural network | |
CN112465124B (en) | Twin depth space-time neural network model acquisition/fault diagnosis method and device | |
CN106950472A (en) | A kind of isolator detecting method being imaged based on infrared and ultraviolet | |
CN106951863B (en) | Method for detecting change of infrared image of substation equipment based on random forest | |
TWI683262B (en) | Industrial image inspection method and system and computer readable recording medium | |
CN107833211A (en) | Zero resistance insulator automatic testing method and device based on infrared image | |
CN113077416A (en) | Welding spot welding defect detection method and system based on image processing | |
Zheng et al. | Intelligent diagnosis method of power equipment faults based on single‐stage infrared image target detection | |
CN110909774B (en) | Power transformation equipment heating defect reason distinguishing method based on Bayesian classification | |
CN111626104A (en) | Cable hidden danger point detection method and device based on unmanned aerial vehicle infrared thermal imagery | |
CN112927222B (en) | Method for realizing multi-type photovoltaic array hot spot detection based on hybrid improved Faster R-CNN | |
CN112508946B (en) | Cable tunnel anomaly detection method based on antagonistic neural network | |
CN110095204B (en) | Ethylene cracking furnace tube temperature diagnostic system | |
CN113538411A (en) | Insulator defect detection method and device | |
CN111488785A (en) | Rotary kiln working condition detection method based on in-kiln 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 |