CN113627378A - Line bolt defect identification method based on combination of Phash algorithm and deep learning - Google Patents
Line bolt defect identification method based on combination of Phash algorithm and deep learning Download PDFInfo
- Publication number
- CN113627378A CN113627378A CN202110954367.7A CN202110954367A CN113627378A CN 113627378 A CN113627378 A CN 113627378A CN 202110954367 A CN202110954367 A CN 202110954367A CN 113627378 A CN113627378 A CN 113627378A
- Authority
- CN
- China
- Prior art keywords
- bolt
- image
- detected
- sample set
- picture
- 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.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 34
- 238000013135 deep learning Methods 0.000 title claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000001514 detection method Methods 0.000 claims abstract description 17
- 230000005540 biological transmission Effects 0.000 claims abstract description 13
- 238000009432 framing Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 39
- 238000012545 processing Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000013519 translation Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 230000009471 action Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000003190 augmentative effect Effects 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a line bolt defect identification method based on a Phash algorithm combined with deep learning, which comprises the following steps: step S1: collecting bolt images of the power transmission line, framing the bolt positions, and segmenting and extracting the framed images to serve as an initial sample set; step S2: judging the bolt missing condition in the initial sample set; step S3: expanding the sample size of the picture without missing the bolts in the initial sample set to obtain an expanded sample set; step S4: leading the extended sample set as training data into a Faster R-CNN network model for training; step S5: the method has the advantages that the problems of low accuracy and easiness in being influenced by external light in the existing bolt loss detection method are solved, a more universal bolt defect detection technology is realized, and the method has higher accuracy and applicability, so that the acquisition of data to be detected is not limited by weather, position and equipment factors.
Description
Technical Field
The invention relates to the technical field of defect detection of power transmission lines, in particular to a line bolt defect identification method based on a Phash algorithm and deep learning.
Background
The transmission line is an important part in a national power system, generally exists in various complex environments, is inevitably subjected to different-degree loss after being influenced by external environmental factors such as wind, sunlight and the like, and is essential to overhaul the line in order to provide a more stable power supply environment for residents. The manual maintenance is not only expensive, but also lacks the guarantee to maintainer's safety, and some extreme geographical positions also can cause the hindrance to the manual maintenance of transmission line, and unmanned aerial vehicle's appearance provides more simple and convenient, safe method in the aspect of the line inspection is repaiied undoubtedly.
Although defect repair of the unmanned aerial vehicle in the power transmission line has been widely applied, there are still difficulties in performing defect analysis on a large amount of picture data acquired by the unmanned aerial vehicle. Traditional unmanned aerial vehicle patrols and examines image processing mode, has shortcomings such as inefficiency, degree of accuracy are not high. The bolt defect is a common device defect in a power transmission line, and the conventional bolt defect detection scheme is that a large number of data set features are extracted through deep learning, and then whether a bolt is defective or not is judged by using a detection identification network. The method has the advantages of large influence from the outside, narrow application range, low detection accuracy rate under the condition of bolt loss, and incapability of detecting the bolt model without the template.
Disclosure of Invention
The invention aims to provide a line bolt defect identification method based on a Phash algorithm combined with deep learning. The method aims to solve the problems that the defect detection method in the prior art is greatly influenced by the outside, has a narrow application range, has low detection accuracy under the condition of bolt loss and cannot detect a bolt model without a template.
In order to achieve the purpose, the invention is realized by the following technical scheme:
a line bolt defect identification method based on a Phash algorithm and deep learning comprises the following steps:
step S1: collecting bolt images of the power transmission line, framing bolt positions on the bolt images, and segmenting and extracting the images of the framed bolt positions to serve as an initial sample set;
step S2: judging the bolt missing condition in the initial sample set by a Phash algorithm, and selecting a similarity threshold;
step S3: taking the picture without missing the bolts in the initial sample set as a basic sample set, and expanding the basic sample set by sample size to obtain an expanded sample set;
step S4: establishing a Faster R-CNN network model for identifying bolt defects through deep learning, taking the extended sample set as training data, training the Faster R-CNN network model, and placing the trained Faster R-CNN network model on a CPU (Central processing Unit);
step S5: shooting the picture of the power transmission line by using an unmanned aerial vehicle and taking the picture as an image to be detected, judging the bolt missing condition in the image to be detected by using a Phash algorithm, inputting the image to be detected to the fast R-CNN network model after training is finished, and judging the bolt missing condition in the image to be detected by using the fast R-CNN network model.
Preferably, in the step S2, the bolt missing condition in the initial sample set is determined, and if it is determined that there is bolt missing in one picture in the initial sample set, the picture is marked and defined as bolt missing; if it is determined that there is no missing bolt in another picture in the initial sample set, the step S3 is executed.
Preferably, in step S2, the specific step of determining that the bolt is missing in the initial sample set includes:
step S201: graying any picture to be detected in the initial sample set to obtain a grayscale image, performing DCT (discrete cosine transformation) change on the grayscale image to obtain a 64 x 64 DCT coefficient matrix, intercepting a DCT matrix gary (i, j) of the upper left corner 16 x 16 in the 64 x 64 DCT coefficient matrix, and calculating an average value Av of the DCT matrix gary (i, j), wherein a calculation formula is as follows:
step S202: establishing a zero matrix zero (i, j) of 16 x 16, and comparing each numerical value in the DCT matrix gary (i, j) with the average value Av of the matrix;
if a certain value in the DCT matrix gary (i, j) is greater than the average value Av, changing the value in the corresponding zero matrix zero (i, j) to 1; if a value in the DCT matrix gary (i, j) is not greater than the average value Av, changing the value in the corresponding zero matrix zero (i, j) to 0, where the formula of the zero matrix zero (i, j) is:
the final zero matrix zero (i, j) outputs a binary image of the picture to be detected, wherein the binary image is 16 × 16, and if the numerical value is 1, the picture to be detected is displayed in white; a value of 0 indicates black.
Step S203: sequentially performing the processing in the step S201 and the step S202 on the pictures confirmed as the pictures with no missing bolt to obtain binary images of the pictures with no missing bolt, comparing the binary images of the pictures with the binary images of the pictures to be detected, calculating the number of different values to obtain hamming distances, and indirectly obtaining the similarity of the images according to the hamming distances, wherein the similarity formula of the images is as follows:
in the formula: d represents a Hamming distance; p represents the image similarity of the Phash algorithm.
Preferably, in step S203, the image similarity P is calculated to obtain a distance distribution mean value d and a standard deviation e, and then the similarity threshold is selected in an interval [ d + e, d-e ], and the bolt missing condition is determined according to the image similarity P and the similarity threshold.
Preferably, if the value of the similarity P of the image of the picture to be detected is greater than the similarity threshold, the picture to be detected is a bolt which is not missing; and if the value of the similarity P of the image of the picture to be detected is not more than the similarity threshold value, the picture to be detected is bolt missing.
Preferably, in the step S3, the base sample set is augmented by one or any combination of random flipping, random scaling, random translation and random picture enhancement, so as to obtain the augmented sample set.
Preferably, in step S4, the FasterR-CNN network model is trained, and a single GPU is required to be used for training, so that the training feature map can obtain the boxed portion through the RPN network and the detection network.
Preferably, the CPU is configured to identify the image to be detected.
Preferably, in step S5, if the image to be detected has a bolt defect, obtaining a position coordinate of the bolt by using the Faster R-CNN network model, and defining the position coordinate as the bolt defect; and if the image to be detected has no bolt defect, obtaining the position coordinate of the bolt by using the Faster R-CNN network model, and defining the position coordinate as the bolt defect.
Preferably, in step S5, the position coordinates of the bolt obtained by the Faster R-CNN network model are converted to obtain the coordinates of the position of the landmark bolt in the image to be detected, and the coordinates of the position of the landmark bolt in the image to be detected are filtered to obtain the accurate coordinates of the position of the landmark bolt in the image to be detected.
Compared with the prior art, the invention has the following beneficial effects:
the method solves the problems of low accuracy and easy influence of external light in the existing bolt loss detection method, can determine whether the bolt is lost or not by using a Phash algorithm, and then judges whether the bolt has defects or not under the condition that the bolt is not lost or not by using a deep learning model, namely, a more universal bolt defect detection technology is realized.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings used in the description will be briefly introduced, and it is obvious that the drawings in the following description are an embodiment of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts according to the drawings:
fig. 1 is a schematic flow chart of a line bolt defect identification method based on a Phash algorithm combined with deep learning according to an embodiment of the present invention;
fig. 2 is a schematic diagram of an RPN network according to an embodiment of the present invention.
Detailed Description
The method for identifying the defect of the line bolt based on the Phash algorithm and the deep learning is further described in detail with reference to fig. 1 and fig. 2 and the specific implementation manner. The advantages and features of the present invention will become more apparent from the following description. It is to be noted that the drawings are in a very simplified form and are all used in a non-precise scale for the purpose of facilitating and distinctly aiding in the description of the embodiments of the present invention. To make the objects, features and advantages of the present invention comprehensible, reference is made to the accompanying drawings. It should be understood that the structures, ratios, sizes, and the like shown in the drawings and described in the specification are only used for matching with the disclosure of the specification, so as to be understood and read by those skilled in the art, and are not used to limit the implementation conditions of the present invention, so that the present invention has no technical significance, and any structural modification, ratio relationship change or size adjustment should still fall within the scope of the present invention without affecting the efficacy and the achievable purpose of the present invention.
In view of the defects of the existing power transmission line defect detection method, the embodiment provides a line bolt defect identification method based on the combination of the Phash algorithm and the deep learning, which comprises the following steps:
a line bolt defect identification method based on a Phash algorithm and deep learning comprises the following steps:
step S1: collecting bolt images of the power transmission line, framing bolt positions on the bolt images, and segmenting and extracting the images of the framed bolt positions to serve as an initial sample set;
step S2: judging the bolt missing condition in the initial sample set by a Phash algorithm, and selecting a similarity threshold;
judging the bolt missing condition in an initial sample set, and if judging that a picture in the initial sample set has bolt missing, marking the picture and defining the picture as bolt missing; if it is determined that there is no missing bolt in another picture in the initial sample set, the step S3 is executed.
The concrete steps for judging the missing condition of the bolts in the initial sample set comprise:
step S201: graying any picture to be detected in the initial sample set to obtain a grayscale image, performing DCT (discrete cosine transformation) change on the grayscale image to obtain a 64 x 64 DCT coefficient matrix, intercepting a DCT matrix gary (i, j) of the upper left corner 16 x 16 in the 64 x 64 DCT coefficient matrix, and calculating an average value Av of the DCT matrix gary (i, j), wherein a calculation formula is as follows:
step S202: establishing a zero matrix zero (i, j) of 16 x 16, and comparing each numerical value in the DCT matrix gary (i, j) with the average value Av of the matrix;
if a certain value in the DCT matrix gary (i, j) is greater than the average value Av, changing the value in the corresponding zero matrix zero (i, j) to 1; if a value in the DCT matrix gary (i, j) is not greater than the average value Av, changing the value in the corresponding zero matrix zero (i, j) to 0, where the formula of the zero matrix zero (i, j) is:
the final zero matrix zero (i, j) outputs a binary image of the picture to be detected, wherein the binary image is 16 × 16, and if the numerical value is 1, the picture to be detected is displayed in white; a value of 0 indicates black.
Step S203: after the pictures confirmed as the pictures without missing bolts are processed in the steps S201 and S202, obtaining binary images of the pictures without missing bolts, comparing the binary images of the pictures without missing bolts with the binary images of the pictures to be detected, calculating the number of different values to obtain hamming distances, and indirectly obtaining the similarity of the images according to the hamming distances, wherein the similarity formula of the images is as follows:
in formula (3): d represents a Hamming distance; p represents the image similarity of the Phash algorithm.
And calculating the value of the image similarity P to obtain a distance distribution mean value d and a standard deviation epsilon, selecting the similarity threshold value in an interval [ d + epsilon, d-epsilon ], and judging the bolt missing condition according to the value of the image similarity P and the similarity threshold value.
If the value of the similarity P of the image of the picture to be detected is greater than the similarity threshold value, the picture to be detected is a bolt which is not lost; and if the value of the similarity P of the image of the picture to be detected is not more than the similarity threshold value, the picture to be detected is bolt missing.
Step S3: and taking the picture without missing the bolts in the initial sample set as a basic sample set, and expanding the basic sample set by sample size to obtain an expanded sample set.
And expanding the sample size of the basic sample set by adopting one or any combination of random overturning, random scaling, random translation and random picture enhancement.
In the embodiment, in the manual labeling process, the number of the obtained final available samples is small, so that a FasterR-CNN model with high accuracy cannot be supported. In the process of picture collection, the situations of different picture sizes, different bolt colors in different time periods and the like can occur, so that the manual marked image is expanded by adopting a random scaling and random color changing mode, the quantity of training data can be increased, and a Faster R-CNN network model with higher accuracy is obtained.
Step S4: establishing a Faster R-CNN network model for identifying bolt defects through deep learning, taking the extended sample set as training data, training the Faster R-CNN network model, and placing the trained Faster R-CNN network model on a CPU (Central processing Unit); and the CPU is used for identifying the image to be detected.
Refer to fig. 2 for a diagram of the RPN network operation process. And identifying the bolt defects in the image to be detected by using different characteristic layers in the Faster R-CNN network model by using an RPN algorithm. The RPN algorithm has higher detection precision for smaller objects, so that the small bolts in the image to be detected can be accurately detected by the RPN algorithm. Wherein the size ratio of the candidate objects is [1:0.5,1:1,0.5:1 ].
And training the Faster R-CNN network model by adopting a single GPU (graphics processing Unit), so that a training characteristic diagram can obtain a boxed part through an RPN (resilient packet network) and a detection network.
Step S5: shooting the picture of the power transmission line by using an unmanned aerial vehicle and taking the picture as an image to be detected, judging the bolt missing condition in the image to be detected by using a Phash algorithm, inputting the image to be detected to the fast R-CNN network model after training is finished, and judging the bolt missing condition in the image to be detected by using the fast R-CNN network model.
If the image to be detected has bolt defects, obtaining position coordinates of the bolt by using the Faster R-CNN network model, and defining the position coordinates as the defects of the bolt; and if the image to be detected has no bolt defect, obtaining the position coordinate of the bolt by using the Faster R-CNN network model, and defining the position coordinate as the bolt defect.
And converting the position coordinates of the bolt obtained by the Faster R-CNN network model to obtain the coordinates of the position of the marked bolt in the image to be detected, and filtering the coordinates of the position of the marked bolt in the image to be detected to obtain the accurate coordinates of the position of the marked bolt in the image to be detected.
Because the coordinates obtained by the Faster R-CNN network model are put into the original graph to be tested, the picture patterns are not matched, and the aspect ratio is different, the coordinates need to be converted. In the process, the coordinate value beyond the range of the detected image needs to be removed, and finally the accurate coordinate of the position of the marked bolt in the image to be detected is obtained.
In summary, the present embodiment is based on the existing well-developed deep learning technology, and mainly solves the problems of low accuracy and being easily affected by external light in the existing bolt loss detection method. Whether the bolt is missing or not can be determined by using a Phash algorithm, and then whether the bolt has defects or not is judged by a deep learning model under the condition that the bolt is not missing, so that a more universal bolt defect detection technology is realized. The method provided by the embodiment has higher accuracy and applicability, so that the acquisition of the data to be detected is not limited by the influence of weather, position and equipment factors.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
It should be noted that the apparatuses and methods disclosed in the embodiments herein can be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments herein. In this regard, each block in the flowchart or block diagrams may represent a module, a program, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, the functional modules in the embodiments herein may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.
Claims (10)
1. A line bolt defect identification method based on a Phash algorithm and deep learning is characterized by comprising the following steps:
step S1: collecting bolt images of the power transmission line, framing bolt positions on the bolt images, and segmenting and extracting the images of the framed bolt positions to serve as an initial sample set;
step S2: judging the bolt missing condition in the initial sample set by a Phash algorithm, and selecting a similarity threshold;
step S3: taking the picture without missing the bolts in the initial sample set as a basic sample set, and expanding the basic sample set by sample size to obtain an expanded sample set;
step S4: establishing a Faster R-CNN network model for identifying bolt defects through deep learning, taking the extended sample set as training data, training the Faster R-CNN network model, and placing the trained Faster R-CNN network model on a CPU (Central processing Unit);
step S5: shooting the picture of the power transmission line by using an unmanned aerial vehicle and taking the picture as an image to be detected, judging the bolt missing condition in the image to be detected by using a Phash algorithm, inputting the image to be detected to the fast R-CNN network model after training is finished, and judging the bolt missing condition in the image to be detected by using the fast R-CNN network model.
2. The method for identifying the defect of the line bolt based on the Phash algorithm and the deep learning according to claim 1, wherein in the step S2, the situation of the bolt missing in the initial sample set is determined, and if it is determined that a picture in the initial sample set has the bolt missing, the picture is marked and defined as the bolt missing; if it is determined that there is no missing bolt in another picture in the initial sample set, the step S3 is executed.
3. The method for identifying the defect of the line bolt based on the Phash algorithm combined with the deep learning of claim 1, wherein in the step S2, the specific step of judging the missing situation of the bolt in the initial sample set comprises:
step S201: graying any picture to be detected in the initial sample set to obtain a grayscale image, performing DCT (discrete cosine transformation) change on the grayscale image to obtain a 64 x 64 DCT coefficient matrix, intercepting a DCT matrix gary (i, j) of the upper left corner 16 x 16 in the 64 x 64 DCT coefficient matrix, and calculating an average value Av of the DCT matrix gary (i, j), wherein a calculation formula is as follows:
step S202: establishing a zero matrix zero (i, j) of 16 x 16, and comparing each numerical value in the DCT matrix gary (i, j) with the average value Av of the matrix;
if one value in the DCT matrix gary (i, j) is greater than the average value Av, changing the value in the corresponding zero matrix zero (i, j) to 1; if one value in the DCT matrix gary (i, j) is not more than the average value Av, the value in the corresponding zero matrix zero (i, j) is changed to 0,
the zero matrix zero (i, j) is expressed by the formula:
the zero matrix zero (i, j) outputs a 16 × 16 binary image of the picture to be detected, wherein if the numerical value is 1, white is displayed; if the numerical value is 0, black is displayed;
step S203: sequentially performing the processing in the step S201 and the step S202 on the pictures confirmed as the pictures with no missing bolt to obtain binary images of the pictures with no missing bolt, comparing the binary images of the pictures with the binary images of the pictures to be detected, calculating the number of different values to obtain hamming distances, and indirectly obtaining the similarity of the images according to the hamming distances, wherein the similarity formula of the images is as follows:
in the formula: d represents a Hamming distance; p represents the image similarity of the Phash algorithm.
4. The method for identifying the defect of the line bolt based on the Phash algorithm and the deep learning according to claim 3, wherein in the step S203, the value of the image similarity P is calculated to obtain a distance distribution mean value d and a standard deviation e thereof, and then the similarity threshold is selected in an interval [ d + e, d-e ], and the bolt missing condition is determined according to the value of the image similarity P of the image to be detected and the similarity threshold.
5. The line bolt defect identification method based on Phash algorithm combined with deep learning of claim 4, wherein in the step S203, if the value of the similarity P of the image of the picture to be tested is greater than the similarity threshold, the picture to be tested is a picture without bolt missing; and if the value of the similarity P of the image of the picture to be detected is not more than the similarity threshold value, the picture to be detected is bolt missing.
6. The method for identifying the defect of the line bolt based on the Phash algorithm and the deep learning according to claim 1, wherein in the step S3, the extended sample set is obtained by expanding the sample size of the basic sample set by using one or any combination of random inversion, random scaling, random translation and random picture enhancement.
7. The method for line bolt defect recognition based on Phash algorithm combined with deep learning according to claim 1, wherein in step S4, the FasterR-CNN network model is trained, which requires a single GPU for training, so that the training feature map can obtain the boxed part through the RPN network and the detection network.
8. The line bolt defect identification method based on Phash algorithm combined with deep learning according to claim 1, wherein the CPU is used for identifying the image to be detected.
9. The line bolt defect identification method based on Phash algorithm combined with deep learning according to claim 1, wherein in step S5, if the image to be detected has a bolt defect, the fasterrr-CNN network model is used to obtain the position coordinates of the bolt, and the position coordinates are defined as the bolt defect; and if the image to be detected has no bolt defect, obtaining the position coordinate of the bolt by using the FasterR-CNN network model, and defining the position coordinate as the bolt defect.
10. The line bolt defect identification method based on Phash algorithm combined with deep learning according to claim 9, wherein in step S5, the position coordinates of the bolt obtained by the faster r-CNN network model are converted to obtain the coordinates of the position of the labeled bolt in the image to be detected, and the coordinates of the position of the labeled bolt in the image to be detected are filtered to obtain the accurate coordinates of the position of the labeled bolt in the image to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110954367.7A CN113627378B (en) | 2021-08-19 | 2021-08-19 | Line bolt defect identification method based on Phash algorithm and deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110954367.7A CN113627378B (en) | 2021-08-19 | 2021-08-19 | Line bolt defect identification method based on Phash algorithm and deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113627378A true CN113627378A (en) | 2021-11-09 |
| CN113627378B CN113627378B (en) | 2024-07-02 |
Family
ID=78386658
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110954367.7A Active CN113627378B (en) | 2021-08-19 | 2021-08-19 | Line bolt defect identification method based on Phash algorithm and deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113627378B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108242054A (en) * | 2018-01-09 | 2018-07-03 | 北京百度网讯科技有限公司 | A kind of steel plate defect detection method, device, equipment and server |
| CN110232687A (en) * | 2019-06-21 | 2019-09-13 | 华北电力大学（保定） | The detection method of cotter bolt defect in a kind of electric inspection process image |
| CN110458807A (en) * | 2019-07-09 | 2019-11-15 | 常州大学 | A kind of railroad track defect Machine Vision Inspecting System |
| CN111797890A (en) * | 2020-05-18 | 2020-10-20 | 中国电力科学研究院有限公司 | Method and system for detecting defects of power transmission line equipment |
| CN112581445A (en) * | 2020-12-15 | 2021-03-30 | 中国电力科学研究院有限公司 | Detection method and device for bolt of power transmission line, storage medium and electronic equipment |
-
2021
- 2021-08-19 CN CN202110954367.7A patent/CN113627378B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108242054A (en) * | 2018-01-09 | 2018-07-03 | 北京百度网讯科技有限公司 | A kind of steel plate defect detection method, device, equipment and server |
| CN110232687A (en) * | 2019-06-21 | 2019-09-13 | 华北电力大学（保定） | The detection method of cotter bolt defect in a kind of electric inspection process image |
| CN110458807A (en) * | 2019-07-09 | 2019-11-15 | 常州大学 | A kind of railroad track defect Machine Vision Inspecting System |
| CN111797890A (en) * | 2020-05-18 | 2020-10-20 | 中国电力科学研究院有限公司 | Method and system for detecting defects of power transmission line equipment |
| CN112581445A (en) * | 2020-12-15 | 2021-03-30 | 中国电力科学研究院有限公司 | Detection method and device for bolt of power transmission line, storage medium and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 吴涛: ""基于深度学习的铁路接触网开口销状态检测方法研究"", 《CNKI硕士论文》, pages 1 - 63 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113627378B (en) | 2024-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11551341B2 (en) | Method and device for automatically drawing structural cracks and precisely measuring widths thereof | |
| CN107543828B (en) | Workpiece surface defect detection method and system | |
| CN106529537A (en) | Digital meter reading image recognition method | |
| CN111709909A (en) | General printing defect detection method based on deep learning and model thereof | |
| CN105574533B (en) | A kind of image characteristic extracting method and device | |
| CN114549835A (en) | Pointer instrument correction identification method and device based on deep learning | |
| CN112907614A (en) | Yoov 5-segnet insulator string contour extraction method based on depth feature fusion | |
| CN113592839B (en) | Distribution network line typical defect diagnosis method and system based on improved fast RCNN | |
| CN111598942A (en) | Method and system for automatically positioning electric power facility instrument | |
| CN113469178B (en) | Power meter identification method based on deep learning | |
| CN112699824B (en) | Method and device for detecting constant of electric energy meter and storage medium | |
| CN111914767A (en) | Scattered-pollution enterprise detection method and system based on multi-source remote sensing data | |
| CN112525925A (en) | Keyboard detection method, system, electronic equipment and medium | |
| CN114266881A (en) | Pointer type instrument automatic reading method based on improved semantic segmentation network | |
| CN113554667A (en) | Three-dimensional displacement detection method and device based on image recognition | |
| CN113065556B (en) | Digital instrument positioning and digital identifying method and device and computer equipment | |
| CN108992033B (en) | Grading device, equipment and storage medium for vision test | |
| CN114428110A (en) | Method and system for detecting defects of fluorescent magnetic powder inspection image of bearing ring | |
| CN111310651B (en) | Water surface detection method based on polarization camera and RGB-D sensor | |
| CN115578695B (en) | Water gauge water level machine vision detection method and device with free shooting visual angle | |
| CN113627378A (en) | Line bolt defect identification method based on combination of Phash algorithm and deep learning | |
| CN104268536B (en) | A kind of image method for detecting human face | |
| CN115588196A (en) | Pointer type instrument reading method and device based on machine vision | |
| CN111626175B (en) | Shaft type identification method based on deep convolutional neural network | |
| CN109360289B (en) | Power meter detection method fusing inspection robot positioning information |
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 |