CN114612742A - Method and system for detecting defect of small target of power transmission line - Google Patents
Method and system for detecting defect of small target of power transmission line Download PDFInfo
- Publication number
- CN114612742A CN114612742A CN202210233188.9A CN202210233188A CN114612742A CN 114612742 A CN114612742 A CN 114612742A CN 202210233188 A CN202210233188 A CN 202210233188A CN 114612742 A CN114612742 A CN 114612742A
- Authority
- CN
- China
- Prior art keywords
- target
- power transmission
- transmission line
- stage
- image
- 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
- 230000007547 defect Effects 0.000 title claims abstract description 51
- 230000005540 biological transmission Effects 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000001514 detection method Methods 0.000 claims abstract description 66
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000007246 mechanism Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000001629 suppression Effects 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 description 15
- 238000013139 quantization Methods 0.000 description 5
- 239000003086 colorant Substances 0.000 description 4
- 230000000007 visual effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000013138 pruning Methods 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000002950 deficient Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 235000005770 birds nest Nutrition 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000012212 insulator Substances 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 235000005765 wild carrot Nutrition 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
- 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)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method and a system for detecting defects of small targets of a power transmission line, wherein the method comprises the following steps: acquiring and preprocessing image data of the power transmission line; inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image; and inputting the new image into the trained second-stage target detection model to obtain a target defect identification result. According to the method, a key area is positioned through a target detection model in a first stage, the key area is re-extracted according to a certain rule, and the position is re-distributed in proportion; since the pixel information of the original image is preserved, there is no loss of information, and a very good sample is provided for the second stage of detection.
Description
Technical Field
The invention relates to the technical field of defect detection of small targets of power transmission lines, in particular to a method and a system for detecting defects of small targets of power transmission lines.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
For a power transmission line scene, a large number of bolt connecting members exist on a transmission tower structure of the power transmission line scene, the same volume of small target bodies exist, the states of the bolt connecting members have great influence on the stability of the tower structure and the stability of a transmission line, and the defect detection of the bolt connecting members is one of important works for power fault inspection.
The existing detection on the defects of the power transmission line body is easily influenced by factors such as visual angle, background and shading, cannot be completely exposed in the visual field of a camera, and has the problems of high similarity of colors or appearances among hardware fittings and between the hardware fittings and other parts in the line, uneven illumination of outdoor environment, unstable light source and the like, so that the identification accuracy is low.
When a neural network algorithm is adopted for detection, a neural network with ideal performance generally comprises hundreds of millions of parameters, occupies a storage space of tens of thousands of megabytes, has high calculation energy consumption, and is difficult to be directly deployed on an edge calculation platform with a small power and an internal memory; once the neural network is designed in a light weight mode, the expression capacity of the network is weakened, and the accuracy of target identification is reduced.
A single neural network detection scheme is adopted, the detection omission ratio is limited by the size change of a target and the interference factors of an environmental background, and the omission ratio is high; if a neural network with higher complexity is adopted, the improvement on the detection capability of the small target is limited, and meanwhile, the defects of large calculation amount and poor real-time performance exist due to the higher complexity.
Disclosure of Invention
In order to solve the problems, the invention provides a method and a system for detecting the defects of the small targets of the power transmission line, which are used for improving the recognition rate of the defects of the common target bodies around the tower based on the fine-grained recognition algorithm of the deep convolutional neural network; the neural network is lightened, meanwhile, a target body defect database is expanded, the generalization capability of the lightweight network is improved, and the expression capability loss caused by network lightening is made up through a complete data set, so that the effective identification of the target body defect of the power transmission line is ensured.
In some embodiments, the following technical scheme is adopted:
a method for detecting defects of small targets of a power transmission line comprises the following steps:
acquiring and preprocessing image data of the power transmission line;
inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
and inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
In other embodiments, the following technical solutions are adopted:
a system for detecting defects of small targets of a power transmission line comprises:
the data acquisition module is used for acquiring and preprocessing the image data of the power transmission line;
the first-stage image processing module is used for inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
and the second-stage image processing module is used for inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
In other embodiments, the following technical solutions are adopted:
a terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is used for storing a plurality of instructions, and the instructions are suitable for being loaded by the processor and executing the method for detecting the small target defect of the power transmission line.
In other embodiments, the following technical solutions are adopted:
a computer readable storage medium, wherein a plurality of instructions are stored, and the instructions are suitable for being loaded by a processor of a terminal device and executing the method for detecting the small target defect of the power transmission line.
Compared with the prior art, the invention has the beneficial effects that:
(1) according to the method, a key area is positioned through a target detection model in a first stage, the key area is re-extracted according to a certain rule, and the position is re-distributed in proportion; since the pixel information of the original image is preserved, no information is lost, and a very good sample is provided for the detection of the second stage.
(2) The invention considers that small objects positioned on the boundary of the key area can be cut down or lose the boundary, so that edge pixels are expanded and detection is carried out again, thereby effectively ensuring that complete objects are detected.
(3) The invention carries out the NMS non-maximum value inhibition algorithm on the frame with higher coverage rate at the same position of the original image, thereby avoiding the situation of repeated detection and effectively preventing the occurrence of missed detection.
Additional features and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
FIG. 1 is a schematic process diagram of a method for detecting a defect of a small target of a power transmission line in an embodiment of the invention;
fig. 2(a) is a test picture, and fig. 2(b) - (c) are respectively a gradient visualization of the detection result after the attention mechanism is added and the detection result of the module without the attention mechanism.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
In the implementation process of the target detection task based on deep learning, the following problems are often faced: in the target classification task, generally, due to the difference of illumination conditions, shooting visual angles and distances in the image acquisition process, the non-rigid body deformation of an object and partial shielding of other objects, the apparent characteristics of an object example are greatly changed, great difficulty is brought to a visual recognition algorithm, and the performance of the algorithm needs to be enhanced. In order to utilize semantic information of a high-resolution picture to a greater extent and avoid invalid convolution and scaling operations performed by a neural network, the embodiment provides an algorithm for positioning a candidate region first and then detecting the candidate region.
Based on this, in one or more embodiments, a method for detecting a small target defect of a power transmission line is disclosed, and with reference to fig. 1, the method includes:
(1) acquiring and preprocessing image data of the power transmission line;
in the embodiment, a high-resolution picture under a power transmission line scene is obtained through a camera, and the resolution is 2K or more; carrying out spatial preprocessing such as image turning, clipping and the like and pixel preprocessing operations such as blurring, saturation changing and the like on the image, wherein the preprocessed image is changed into a low-resolution image, and the resolution is 416 x 416; the converted low-resolution image ensures that the small target defect area has higher area ratio and at least has one to two target detection objects.
In this embodiment, the small target detection object is a bolt, and the defect types include a defect of a bolt connection member, such as a defect of an inserted pin, an unreasonable or loose installation manner, abrasion and burning of a power transmission conductor, a bird nest on a bracket, a defect of an insulator, and the like.
(2) Inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
in this embodiment, key regions are first located, and each key region at least includes one target object.
The first-stage target detection model adopts a convolutional neural network model, and because a target main body often has some important regions in an image, the embodiment adds an attention mechanism in the neural network, learns the weight according to loss through the network, so that the effective characteristic diagram weight is amplified, and the model is trained in a mode that the ineffective or small-effect characteristic diagram weight is reduced, so as to achieve better performance.
In the embodiment, an attention module is additionally arranged between the original network basic structure and the bottleneck layer. The reason for selecting the position is that the bottleneck layer is subjected to up-sampling operation, more parameters are introduced, the model structure and the parameters are not suitable for introducing more, and the last layer of the basic structure has rich semantic information and smaller parameter quantity and is more suitable for introducing a new network structure. Each channel of the feature map obtained from the last layer of the basic structure is assigned a new weight, which can be learned along with the neural network back propagation process, so as to distinguish attention from background information. Visually, information such as different textures, colors, shapes and the like is distinguished by more direct weights, so that the precision is improved.
In this embodiment, the training process of the first-stage target detection model includes first performing data annotation, then performing model training, and finally obtaining a located key region.
The camera acquires image data, and labels a partial region of the original data, wherein the labeled data is a key region in the image and is used for roughly positioning a small target region, and the key region often contains more than one small target, so that the labeled data can be used as initial input data for second-stage detection.
For the data source in the first stage, a part of the data source comes from a data set provided by a power grid and is subjected to semi-manual labeling, because related technicians label defective parts, the colors and shapes of all pixel points of the picture are extracted firstly, the specific method uses an OpenCV function, and because the connectivity and the colors of the labeled area and the unlabeled area are very different, such as a red rectangle or a yellow circle, the position of the labeled pixel can be obtained, and the data set required by training is finally obtained. And the other part of data sources are obtained through a clustering algorithm after the final result is obtained, and then are fed back to the first-stage target detection model to be used as data for model training.
(3) And inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
Since the key areas obtained in the first stage are not uniformly distributed in the picture and occupy the original picture, the key areas obtained in the first stage need to be recombined and arranged. The logic of the recombination is to arrange the combined key regions in order from left to right and from top to bottom from the top left corner of a completely white background picture, without considering the spatial relationship between the boundaries of the key regions, because the boundaries have no relationship with the defective small objects inside the key regions. We perform an adaptive scaling and compensation algorithm taking into account the excess background and blank margins encountered during the combining process. When the blank area and the key area of the first stage result to be placed are arranged at the leftmost side of the row, the sizes of the blank area and the key area of the first stage result to be placed are calculated, if the size of the key area is too large, the key area is reduced, and if the size of the key area is enough, the key area does not need to be reduced. The same computational comparison is also performed when arranged to the lowermost side of a column. If the residual space is less than 10 pixel points, stopping placing, and if a blank background image is not enough, needing to create a second page background image for continuing placing.
When allocating positions, more positioning area images need to be mixed according to a certain rule. When the positions are redistributed, considering that small objects positioned on the boundary of the positioning area can be cut down or lose the boundary, the edge pixels of each key area are expanded and the detection is carried out again; the purpose is to prevent the condition that when an original image is segmented into key areas, the finally detected small target is cut off, and detection omission is caused.
According to the embodiment, data statistics is carried out on the size of the target object bolt to obtain the length-width mode of the bolt, and the data is used as the basis for expanding the number of pixels, so that the complete target can be effectively ensured to be detected.
In this embodiment, after the identified target defects are re-analyzed in the original image, the NMS non-maximum suppression algorithm is used to suppress the target detection frames with higher coverage at the same position of the original image, and to filter out the frames with lower confidence at the repeated position.
Specifically, after the relative position of the small target detection frame on the canvas in the key area is mapped back to the absolute position on the original image, some frames are overlapped on the original image scale due to expansion, and the frames with low confidence level on the repeated position can be filtered out through the confidence level and the overlapping degree calculated by the neural network, so that the situation of repeated detection cannot occur, and the occurrence of missed detection can be prevented.
In this embodiment, the second-stage target detection model is a convolutional neural network model, and an attention mechanism may also be added to the neural network, and the learning weight is removed through the network according to the loss, so that the effective feature map weight is amplified, and the ineffective or less effective feature map weight is reduced to train the model, so as to achieve better performance.
Fig. 2(a) is a test picture, and fig. 2(b) - (c) respectively perform gradient visualization on the detection result after the attention mechanism is added and the detection result without the attention mechanism added, and the results show that the detection result after the attention mechanism is added is more effective.
In this embodiment, the first-stage target detection model and the second-stage target detection model are respectively designed in a light weight manner through a pruning scheme and a quantization scheme.
The deep learning network model has a large number of redundant parameters from a convolutional layer to a fully-connected layer, the activation value of a large number of neurons approaches to 0, the neurons can show the same model expression capability after being removed, the condition is called over-parameterization, and the corresponding technology is called model pruning. After pruning, the model parameter quantity is reduced by more than 30%, and meanwhile, the precision is not reduced.
In this embodiment, the original neural network target detection framework is pruned, that is, it is not necessary to cover a multi-scale receptive field, and bottleneck layers of three scales can be reduced to bottleneck layers detected by only one scale.
The quantization of the weight is a process of approximating the continuous value of the weight to a finite plurality of discrete values, the binary number of the discrete values determines the quantization precision, and the precision of the NNIE fixed-point operation is 8-bit. The quantization process is to first divide the entire amplitude into a finite set of small amplitudes (quantization steps), classify the samples that fall within a certain step, and assign the same quantized value.
The algorithm of the onelabel series of yolo and the like is model reasoning based on anchors, namely anchor frames, generally comprises three anchors which respectively cover targets with different sizes in the picture. The maximum anchor result is taken as an example, the neural network reasoning will obtain a vector of [1 × 21 × 13 × 13] (type 1 target detection), the second dimension means a data vector with 3 ratios and a length of 4+1+ c, and actually 4 is coordinate information, x and y and relative width and height in grid respectively; 1 is a score as a foreground object, and c is category information. During data extraction, data belonging to scales and grid at the moment are expected to be continuous, so that array access is not required frequently, but the convolution of the neural network cannot meet the requirement, and therefore a reshape layer and a permaute layer (0,3,1 and 2) are added at the end of the neural network, and the data can be processed more efficiently.
(3) Re-analyzing the identified target defects into the original image to obtain final image data; based on the final image data, performing spatial density clustering on the detection targets by adopting a density-based spatial clustering algorithm to obtain a clustering result of the detection targets, wherein the result is the detection targets and background information near the detection targets, and the background information is 'key area' data required by a first-stage detection task, so that the background information can be used as training data of a first-stage target detection model, more data sets are provided for improving the accuracy of the first-stage model, and the defect of the first-stage data is overcome.
Example two
In one or more embodiments, a system for detecting a defect of a small target of a power transmission line is disclosed, which includes:
the data acquisition module is used for acquiring and preprocessing the image data of the power transmission line;
the first-stage image processing module is used for inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
and the second-stage image processing module is used for inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
The specific implementation of the above modules has been described in the first embodiment, and is not described in detail here.
In the embodiment, a low-power-consumption common camera is selected, embedded edge processing equipment is equipped as a core platform for power transmission line image acquisition and video data processing, the defect detection algorithm of the embodiment is deployed together with a storage unit at the front end of the edge side, fine-grained identification and defect real-time detection of a power transmission line body can be realized under the low-power-consumption limit of a solar cell, an abnormal detection result is transmitted to a cloud server/mobile terminal software system through a wireless transmission unit, and a user can further screen images at a mobile terminal.
EXAMPLE III
In one or more embodiments, a terminal device is disclosed, which includes a server, where the server includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the method for detecting a small target defect of a power transmission line in the first embodiment is implemented. For brevity, no further description is provided herein.
It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general purpose processors, digital signal processors DSP, application specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and so on. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may include both read-only memory and random access memory, and may provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.
In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or by instructions in the form of software.
Example four
In one or more embodiments, a computer-readable storage medium is disclosed, in which a plurality of instructions are stored, and the instructions are adapted to be loaded by a processor of a terminal device and execute the method for detecting the defect of the small target of the power transmission line described in the first embodiment.
Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.
Claims (10)
1. A method for detecting defects of small targets of a power transmission line is characterized by comprising the following steps:
acquiring and preprocessing image data of the power transmission line;
inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
and inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
2. The method for detecting the defect of the small target of the power transmission line according to claim 1, wherein the method for obtaining the image data of the power transmission line and preprocessing the image data comprises the following specific processes:
acquiring original data of the power transmission line under the first resolution, and performing spatial preprocessing of image turning and cutting, and pixel preprocessing of blurring and saturation changing to obtain image data under the second resolution.
3. The method for detecting the small target defect of the power transmission line according to claim 1, wherein after the target defect identification result is obtained, the method further comprises the following steps:
re-analyzing the identified target defects into the original image to obtain final image data;
and based on the final image data, performing spatial density clustering on the detection target by adopting a density-based spatial clustering algorithm to obtain a clustering result of the detection target, and taking the clustering result as training data of the first-stage target detection model.
4. The method for detecting the defects of the small targets of the power transmission line according to claim 3, wherein the identified target defects are re-analyzed in the original image, a non-maximum suppression algorithm is adopted to suppress the target detection frames with higher coverage rate at the same position of the original image, and the frames with lower confidence level at the repeated position are filtered.
5. The method for detecting the defect of the small target of the power transmission line according to claim 1, wherein the first-stage target detection model or the first-stage target detection model is a convolutional neural network model, and an attention mechanism is added before a bottleneck layer of the convolutional neural network.
6. The method for detecting the defect of the small target of the power transmission line according to claim 1, wherein the obtained key areas are recombined and arranged according to a set principle, and specifically comprises the following steps:
sequentially arranging and combining the key areas from a set position of a full white background image according to a set arrangement sequence without considering the spatial relationship between the boundaries of the key areas;
when the key areas are arranged at the boundary positions of a row or a column, comparing the sizes of the key areas to be placed with the sizes of the currently remaining blank areas, if the key areas are larger, reducing the key areas, and if the blank areas are larger, directly placing the key areas;
if the residual space of the current background image is smaller than the set threshold value, stopping placement; and if the key area is not placed completely, creating a second completely white background image and continuing to place.
7. The method for detecting the defect of the small target of the power transmission line according to claim 6,
expanding the edge pixels of each key area;
the number of edge pixels to be expanded is determined based on size data of the detection target.
8. A system for detecting defects of small targets of power transmission lines is characterized by comprising:
the data acquisition module is used for acquiring and preprocessing the image data of the power transmission line;
the first-stage image processing module is used for inputting the preprocessed image data into a trained first-stage target detection model to obtain a plurality of key areas containing target objects, and recombining and arranging the obtained key areas according to a set principle to form a new image;
and the second-stage image processing module is used for inputting the new image into the trained second-stage target detection model to obtain a target defect identification result.
9. A terminal device comprising a processor and a memory, the processor being arranged to implement instructions; the memory is used for storing a plurality of instructions, wherein the instructions are suitable for being loaded by the processor and executing the method for detecting the defect of the small target of the power transmission line according to any one of claims 1 to 7.
10. A computer-readable storage medium having stored thereon a plurality of instructions, wherein the instructions are adapted to be loaded by a processor of a terminal device and to perform the method for detecting a small target defect of a power transmission line according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210233188.9A CN114612742A (en) | 2022-03-09 | 2022-03-09 | Method and system for detecting defect of small target of power transmission line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210233188.9A CN114612742A (en) | 2022-03-09 | 2022-03-09 | Method and system for detecting defect of small target of power transmission line |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114612742A true CN114612742A (en) | 2022-06-10 |
Family
ID=81861051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210233188.9A Pending CN114612742A (en) | 2022-03-09 | 2022-03-09 | Method and system for detecting defect of small target of power transmission line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114612742A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846340A (en) * | 2017-02-17 | 2017-06-13 | 大连理工大学 | A kind of striation boundary extraction method based on on-fixed characteristic point |
| CN107358190A (en) * | 2017-07-07 | 2017-11-17 | 广东中星电子有限公司 | A kind of image key area management method and device |
| CN109343918A (en) * | 2018-08-29 | 2019-02-15 | 北京小米移动软件有限公司 | Icon arrangement method, device, terminal and storage medium |
| CN111401424A (en) * | 2020-03-10 | 2020-07-10 | 北京迈格威科技有限公司 | Target detection method, device and electronic system |
| CN113763424A (en) * | 2021-08-13 | 2021-12-07 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Real-time intelligent target detection method and system based on embedded platform |
-
2022
- 2022-03-09 CN CN202210233188.9A patent/CN114612742A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846340A (en) * | 2017-02-17 | 2017-06-13 | 大连理工大学 | A kind of striation boundary extraction method based on on-fixed characteristic point |
| CN107358190A (en) * | 2017-07-07 | 2017-11-17 | 广东中星电子有限公司 | A kind of image key area management method and device |
| CN109343918A (en) * | 2018-08-29 | 2019-02-15 | 北京小米移动软件有限公司 | Icon arrangement method, device, terminal and storage medium |
| CN111401424A (en) * | 2020-03-10 | 2020-07-10 | 北京迈格威科技有限公司 | Target detection method, device and electronic system |
| CN113763424A (en) * | 2021-08-13 | 2021-12-07 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Real-time intelligent target detection method and system based on embedded platform |
Non-Patent Citations (1)
| Title |
|---|
| RUNHAI JIAO ET AL.: "A Deep Learning Model for Small-size Defective Components Detection in Power Transmission Tower", 《IEEE TRANSACTIONS ON POWER DELIVERY》, 14 September 2021 (2021-09-14), pages 2551, XP011915279, DOI: 10.1109/TPWRD.2021.3112285 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059694B (en) | Intelligent identification method for character data in complex scene of power industry | |
| CN111080620B (en) | Road disease detection method based on deep learning | |
| CN111179217A (en) | Attention mechanism-based remote sensing image multi-scale target detection method | |
| CN109886159B (en) | Face detection method under non-limited condition | |
| CN110781882A (en) | License plate positioning and identifying method based on YOLO model | |
| CN110458812B (en) | Quasi-circular fruit defect detection method based on color description and sparse expression | |
| CN111476758B (en) | Defect detection method and device for AMOLED display screen, computer equipment and storage medium | |
| CN111291826A (en) | Multi-source remote sensing image pixel-by-pixel classification method based on correlation fusion network | |
| CN114743119A (en) | High-speed rail contact net dropper nut defect detection method based on unmanned aerial vehicle | |
| CN113486819A (en) | Ship target detection method based on YOLOv4 algorithm | |
| CN113256624A (en) | Continuous casting round billet defect detection method and device, electronic equipment and readable storage medium | |
| CN114548208A (en) | Improved plant seed real-time classification detection method based on YOLOv5 | |
| CN116958827A (en) | Deep learning-based abandoned land area extraction method | |
| CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
| CN117456358A (en) | Method for detecting plant diseases and insect pests based on YOLOv5 neural network | |
| CN115170988A (en) | Power grid line fault identification method and system based on deep learning | |
| CN115861922B (en) | Sparse smoke detection method and device, computer equipment and storage medium | |
| CN114612742A (en) | Method and system for detecting defect of small target of power transmission line | |
| Zacheilas et al. | An FPGA-based system for video processing to detect holes in aquaculture nets | |
| CN110751150A (en) | FPGA-based binary neural network license plate recognition method and system | |
| CN112818987B (en) | Method and system for identifying and correcting display content of electronic bus stop board | |
| CN113505699A (en) | Ship detection method based on RetinaNet algorithm | |
| CN113780189A (en) | Lane line detection method based on U-Net improvement | |
| CN114037840A (en) | Power transmission line visual object extraction method and system based on multi-feature fusion | |
| CN111797782A (en) | Vehicle detection method and system based on image features |
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 |