CN113627299B - Wire floater intelligent recognition method and device based on deep learning - Google Patents
Wire floater intelligent recognition method and device based on deep learning Download PDFInfo
- Publication number
- CN113627299B CN113627299B CN202110879169.9A CN202110879169A CN113627299B CN 113627299 B CN113627299 B CN 113627299B CN 202110879169 A CN202110879169 A CN 202110879169A CN 113627299 B CN113627299 B CN 113627299B
- Authority
- CN
- China
- Prior art keywords
- target
- wire
- data
- floater
- training set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000013135 deep learning Methods 0.000 title claims abstract description 18
- 238000012549 training Methods 0.000 claims abstract description 49
- 239000004020 conductor Substances 0.000 claims abstract description 21
- 238000003708 edge detection Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 17
- 230000006798 recombination Effects 0.000 claims abstract description 7
- 238000005215 recombination Methods 0.000 claims abstract description 7
- 238000001914 filtration Methods 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 7
- 230000008521 reorganization Effects 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 4
- 230000000149 penetrating effect Effects 0.000 claims description 4
- 238000001514 detection method Methods 0.000 abstract description 17
- 208000034699 Vitreous floaters Diseases 0.000 abstract 6
- 230000005540 biological transmission Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 239000000725 suspension Substances 0.000 description 4
- 238000005070 sampling Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—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
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a wire floater intelligent identification method and device based on deep learning, wherein the method comprises the following steps: carrying out data enhancement on the acquired satellite image data by adopting an image recombination mode to acquire a training set; processing the training set by adopting a Canny edge detection operator to obtain a target conductor; fusing deep features and shallow features in a preset SSD model according to a feature pyramid model, acquiring an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and acquiring a target conducting wire floater; and if the target conductor is successfully compared with the target conductor floater, acquiring a final target conductor floater. According to the invention, data enhancement is performed through image recombination, and the target wire floaters obtained through training by combining with the improved SSD model are compared with the wire standard, so that the recognition accuracy and the detection efficiency are improved.
Description
Technical Field
The invention relates to the technical field of target recognition, in particular to a wire floater intelligent recognition method and device based on deep learning.
Background
The modern society relies heavily on electric power service, only reliable electric power service can maintain the electricity demand in normal production, life, in order to provide reliable electric power service for cities and rural areas, the maintenance of important grid components such as electric power lines, transmission towers and related accessories is crucial, however, the facilities are easily damaged when exposed in outdoor environments, such as hanging kites, balloons, plastic bags, advertising cloths and other floating garbage on transmission wires, and huge safety accidents and economic losses can be caused. Therefore, the power line system is required to be monitored in real time so that maintenance departments can clean the power line system in time when dangerous situations such as wire floating object suspension occur, and due to the lack of an effective power transmission line detection method, many power transmission line detection still depends on manual inspection, and the method has high cost and low efficiency.
In recent years, with the development of deep learning and the popularization and application of satellite remote sensing images in civil use, the detection of the state of a transmission line by means of the satellite images is possible, however, due to the fact that the satellite remote sensing images are large in resolution and complex in background, a deep learning network is trained by lacking a sufficient number of sample images with wire hangers, a sufficient positive sample and a sufficient negative sample are keys for guaranteeing the accuracy of a target detection algorithm, the insufficient samples can cause the insufficient performance of the trained network, the relative area of the wire hangers in the satellite remote sensing images is small, the characteristics of small targets are presented, and the detection accuracy of the existing network model on the small targets is low. The existing researches focus on identifying a wire suspension based on an unmanned aerial vehicle aerial image, such as an image segmentation algorithm research based on a transmission line suspension foreign matter aerial image, and a method for segmenting the wire suspension from an image background and a transmission line foreign matter detection method based on YOLOv4 are provided based on the unmanned aerial vehicle aerial image. The target extracted by the method is just similar to the wire floating object, the position of the target is uncertain with the relative position of the wire, and whether the target is the wire floating object cannot be distinguished.
Disclosure of Invention
The invention aims to provide a wire floater intelligent recognition method based on deep learning, which aims to solve the problems that a few sample training set is insufficient and whether a final floater is acquired or not can not be distinguished.
In order to achieve the above purpose, the invention provides a wire floater intelligent identification method based on deep learning, comprising the following steps:
carrying out data enhancement on the acquired satellite image data by adopting an image recombination mode to acquire a training set;
processing the training set by adopting a Canny edge detection operator to obtain a target conductor;
fusing deep features and shallow features in a preset SSD model according to a feature pyramid model, acquiring an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and acquiring a target conducting wire floater;
and if the target conductor is successfully compared with the target conductor floater, acquiring a final target conductor floater.
Preferably, the acquiring the final target conductor floater is specifically:
if the position of the target wire overlaps with the position of the target wire float, determining that the target wire float is the final target wire float if the position of the target wire overlaps with the position of the target wire float, otherwise, determining that the training set is input into the improved SSD model to perform wire float identification inaccurately.
Preferably, the method of image reorganization is used for data enhancement of the acquired satellite image data, specifically:
preprocessing the satellite image data to obtain preprocessed data, dividing target data in the preprocessed data from background data, and fusing the divided target data with preset background data.
Preferably, the preprocessing is performed on the satellite image data, specifically: and carrying out graying, median filtering and Gaussian filtering treatment on the satellite image data.
Preferably, the dividing the target data from the preprocessed data from the background data specifically includes:
traversing the preprocessed data, selecting the target data by adopting a template matching frame, establishing a target frame of the preprocessed data, and dividing the target frame by adopting a Grabcut algorithm.
Preferably, the fusing of the segmented target data and preset background data is specifically:
pixel point P combining the segmented target data t Pixel point P of the preset background data b And the corresponding weight coefficients are fused as follows:
wherein P is new And representing the pixel points of the training set, wherein alpha and beta represent weight coefficients.
Preferably, the training set is processed by adopting a Canny edge detection operator to obtain a target wire, which is specifically:
denoising the training set by adopting Gaussian filtering, calculating a gradient module and a gradient direction, and reserving a maximum gradient value in the gradient direction;
filling missing values of the target edges broken due to noise interference in the training set by adopting a hysteresis threshold to obtain the Canny edge detection operator processing result;
and extracting the lead in the Canny edge detection operator processing result according to the lead characteristics, and acquiring the target lead, wherein the lead characteristics comprise images penetrating through the training set and a plurality of parallel lines.
Preferably, the fusing deep features and shallow features in the preset SSD model according to the feature pyramid model specifically includes:
and acquiring semantic information of the deep features by adopting a bilinear interpolation method, and transmitting the semantic information to the shallow features containing position information for fusion.
The invention also provides a wire floater intelligent identification device based on deep learning, which comprises:
the data processing module is used for carrying out data enhancement on the acquired satellite image data in an image reorganization mode to acquire a training set;
the first extraction module is used for processing the training set by adopting a Canny edge detection operator to obtain a target wire;
the second extraction module is used for fusing deep features and shallow features in a preset SSD model according to the feature pyramid model, obtaining an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and obtaining a target conducting wire floater;
and the comparison module is used for acquiring a final target wire floater if the target wire is successfully compared with the target wire floater.
Preferably, the comparison module is further configured to:
if the position of the target wire overlaps with the position of the target wire float, determining that the target wire float is the final target wire float if the position of the target wire overlaps with the position of the target wire float, otherwise, determining that the training set is input into the improved SSD model to perform wire float identification inaccurately.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, the recombination of the target and the background is realized according to the image recombination mode, so that the problem of insufficient training set in the prior art is solved, meanwhile, a foundation is provided for a better training model, the deep feature map containing rich semantic information is effectively fused with the shallow feature map containing rich position information after being up-sampled, the detection precision of the shallow feature semantic information when detecting small targets is improved, an improved SSD model is obtained, and the extracted lead is compared with the target lead floating object of the improved SSD model, so that the output result of the improved SSD model is the final target lead floating object, and the detection precision and the detection efficiency are improved.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a method for intelligent recognition of a wire floater based on deep learning according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a Canny edge detection operator process according to another embodiment of the present invention;
FIG. 3 is a schematic diagram of an improved SSD model provided by another embodiment of the present invention;
FIG. 4 is a schematic diagram of a bilinear interpolation algorithm according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a wire floater intelligent recognition device based on deep learning according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the step numbers used herein are for convenience of description only and are not limiting as to the order in which the steps are performed.
It is to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
The terms "comprises" and "comprising" indicate the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The term "and/or" refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Referring to fig. 1, an embodiment of the invention provides a method for intelligently identifying a wire floater based on deep learning, which comprises the following steps:
s101: and carrying out data enhancement on the acquired satellite image data by adopting an image recombination mode to acquire a training set.
Specifically, the satellite image data is preprocessed, image segmented and fused, and the method specifically comprises the following steps:
preprocessing the input satellite image data, such as graying, median filtering, advanced filtering and the like, traversing the preprocessed data, selecting target data by adopting a template matching frame, establishing a target frame of the preprocessed data, segmenting the target frame by adopting a Grabcut algorithm, and combining pixel points P of the segmented target data t Pixel point P of preset background data b And the corresponding weight coefficients are fused as follows:
wherein P is new The pixels representing the training set, alpha and beta represent the weighting coefficients. And randomly rotating, reversing and scaling the fused new image sample.
S102: and processing the training set by adopting a Canny edge detection operator to acquire a target conductor.
Referring to fig. 2, based on the above steps, a Canny edge detection operator is used to extract a wire, which is used as a reference for comparing the output results of the following models, and the specific wire extraction process is as follows:
after the training set is denoised by Gaussian filtering, an edge difference operator Sobel is adopted to calculate the difference between the horizontal direction and the vertical direction to obtain a gradient mode and a direction, a non-maximum value is adopted to restrain the maximum gradient value in the reserved gradient direction, other pixels are deleted, a hysteresis threshold is adopted to fill the missing value of the target edge which is broken due to noise interference in the training set, a Canny edge detection operator processing result is obtained, wires in the Canny edge detection operator processing result are extracted according to wire characteristics, and the target wires are obtained, wherein the wire characteristics comprise approximate straight lines, images penetrating through the training set and a plurality of parallel lines.
S103: and fusing deep features and shallow features in a preset SSD model according to the feature pyramid model, acquiring an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and acquiring a target conducting wire floater.
Specifically, when a preset SSD model detects a target, firstly, a priori frame with different scales and aspect ratios is adopted to predict a boundary frame of the target, then VGG is adopted as a feature extraction network, an input image enters the network, 6 feature graphs with different layers are respectively adopted to predict the category and the coordinate of the target, and finally, a non-maximum value suppression (NMS) method is adopted to carry out final detection. The main idea of the preset SSD is to detect large targets based on deep low resolution feature maps and small targets with shallow high resolution feature maps. However, the shallow feature map does not contain abundant semantic information, so that the model has the problem of low detection accuracy when detecting a small target.
The object in the data set cannot cover all scales, so that the image pyramid (downsampling with different resolutions) is used to help CNN learning, but the speed is too slow, so that only a single scale is usually used for prediction, and an intermediate result is also used for prediction, a layer of transposed convolution is added after several layers of residual modules, the resolution is improved, a segmented result is obtained, or a classified result is obtained through 1x1 convolution or GlobalPool, and the architecture is largely used when auxiliary information and auxiliary loss functions exist.
The authors of the feature pyramid model have improved the above-mentioned method with a very ingenious method, except the lateral connection, also add the connection from top to bottom, fuse the result from top to bottom and result obtained laterally together through the way of adding, the focus here is that the low-level characteristic pattern semanteme is not abundant enough, can not be used for classifying directly, and the deep characteristic is more trustworthy, connect laterally and connect from top to bottom and combine, can get the characteristic pattern of different resolutions, and these characteristic patterns all include the semantic information of the original deepest characteristic pattern.
The feature pyramid model is used for solving the problem of different scales of pictures in object detection, is combined with the feature pyramid model and the preset SSD model, an improved SSD model is constructed, semantic information of deep features is obtained by adopting a bilinear interpolation method, the semantic information is transferred to shallow features containing position information for fusion, the deep features are up-sampled, and the deep feature information is transferred to the shallow features, so that the shallow feature map contains more semantic features, and the small objects can be accurately captured.
Referring to fig. 3, block11 corresponding to conv11_2 in the ssd model is fused with conv10_2 to form Block10 by bilinear interpolation up-sampling, and blocks 9, 8, 7, and 4 are sequentially obtained. By means of the method, semantic information of deep features and high resolution of shallow features are fused, and small target detection accuracy can be improved through prediction of the fused features.
Referring to fig. 4, in the up-sampling, known points in the image are predicted to expand the image size, the up-sampling is implemented by a bilinear interpolation method, assuming that the Q point in fig. 4 is a known point and the R point is an x-axis direction insertion point, the calculation is as follows:
wherein P is the unknown point, x 1 ,x 2 ,y,y 1 ,y 2 Each of the horizontal and vertical coordinates represents a pixel value of each point.
S104: and if the target conductor is successfully compared with the target conductor floater, acquiring a final target conductor floater.
Specifically, since the target wire floats solved in the above step are similar targets, whether the obtained target wire floats are targets actually floating on the wire cannot be judged by using the network model, so that the wire is extracted by using the Canny edge detection operator based on the processing method of the satellite image data in the step S101, and is used as a reference for screening the final target in the step, if the position of the target wire overlaps with the position of the target wire float, the target wire float is judged to be the final target wire float, otherwise, the training set is judged to be input into the improved SSD model to identify the wire float inaccurately, and the target float obtained by the model is not the final target, so that irrelevant suspicious targets are removed, and the detection accuracy is improved.
According to the invention, a target image only containing a small amount of background information is obtained by means of template matching and Grabcut algorithm segmentation, then the obtained target image and a new background image are subjected to pixel point fusion to obtain a new image so as to expand a training set, the problems of few samples and insufficient training sets of satellite remote sensing images are solved, a feature pyramid model is used for reference, deep feature images are up-sampled and then fused with shallow features, semantic information of the shallow features is increased so as to improve the detection precision of the model on small targets such as wire floats in the satellite remote sensing images, and the wire positions and suspicious target positions are compared to serve as conditions for screening final wire floats, so that irrelevant targets extracted by deep learning can be removed, and the detection efficiency is improved.
Referring to fig. 5, another embodiment of the present invention provides an intelligent recognition device for a wire floater based on deep learning, including:
the data processing module 11 is configured to perform data enhancement on the acquired satellite image data by using an image reorganization manner, so as to acquire a training set.
A first extraction module 12 is configured to process the training set using a Canny edge detection operator to obtain a target conductor.
The second extraction module 13 is configured to fuse deep features and shallow features in a preset SSD model according to a feature pyramid model, obtain an improved SSD model, input the training set into the improved SSD model to identify a wire floater, and obtain a target wire floater.
And the comparison module 14 is used for acquiring a final target wire floater if the target wire is successfully compared with the target wire floater.
Specific limitations regarding the deep learning-based wire float intelligent recognition device may be found in the above-mentioned limitations, and will not be described herein. The above-mentioned intelligent recognition device for wire floats based on deep learning can be implemented by all or part of software, hardware and their combination. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.
Claims (3)
1. The intelligent recognition method for the wire floats based on deep learning is characterized by comprising the following steps of:
carrying out data enhancement on the acquired satellite image data by adopting an image recombination mode to acquire a training set; the specific process of carrying out data enhancement on the acquired satellite image data by adopting an image reorganization mode comprises the following steps: preprocessing the satellite image data to obtain preprocessed data, dividing target data in the preprocessed data from background data, and fusing the divided target data with preset background data; the preprocessing of the satellite image data comprises: carrying out graying, median filtering and Gaussian filtering treatment on the satellite image data; the dividing the target data and the background data in the preprocessed data specifically comprises the following steps: traversing the preprocessed data, selecting the target data by adopting a template matching frame, establishing a target frame of the preprocessed data, and dividing the target frame by adopting a Grabcut algorithm; by a means ofThe segmented target data are fused with preset background data, and specifically: pixel point P combining the segmented target data t Pixel point P of the preset background data b And the corresponding weight coefficients are fused, and the specific formula is as follows:wherein P is new Representing the pixel points of the training set, wherein alpha and beta represent weight coefficients;
processing the training set by adopting a Canny edge detection operator to obtain a target conductor; the specific process of acquiring the target wire is as follows: denoising the training set by adopting Gaussian filtering, calculating a gradient module and a gradient direction, and reserving a maximum gradient value in the gradient direction; filling missing values of the target edges broken due to noise interference in the training set by adopting a hysteresis threshold to obtain the Canny edge detection operator processing result; extracting a wire in the Canny edge detection operator processing result according to wire characteristics, and acquiring the target wire, wherein the wire characteristics comprise images penetrating through the training set and a plurality of parallel lines;
fusing deep features and shallow features in a preset SSD model according to a feature pyramid model, acquiring an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and acquiring a target conducting wire floater;
if the target conductor is successfully compared with the target conductor floater, acquiring a final target conductor floater; and if the position of the target wire is overlapped with the position of the target wire floater, determining that the target wire floater is the final target wire floater, otherwise, determining that the training set is input into the improved SSD model to identify the wire floater inaccurately.
2. The intelligent recognition method of the wire floater based on the deep learning according to claim 1, wherein the fusing of deep features and shallow features in a preset SSD model according to a feature pyramid model is specifically as follows:
and acquiring semantic information of the deep features by adopting a bilinear interpolation method, and transmitting the semantic information to the shallow features containing position information for fusion.
3. Wire floater intelligent recognition device based on degree of depth study, its characterized in that includes:
the data processing module is used for carrying out data enhancement on the acquired satellite image data in an image reorganization mode to acquire a training set; the specific process of carrying out data enhancement on the acquired satellite image data by adopting an image reorganization mode comprises the following steps: preprocessing the satellite image data to obtain preprocessed data, dividing target data in the preprocessed data from background data, and fusing the divided target data with preset background data; the preprocessing of the satellite image data comprises: carrying out graying, median filtering and Gaussian filtering treatment on the satellite image data; the dividing the target data and the background data in the preprocessed data specifically comprises the following steps: traversing the preprocessed data, selecting the target data by adopting a template matching frame, establishing a target frame of the preprocessed data, and dividing the target frame by adopting a Grabcut algorithm; the segmented target data are fused with preset background data, and specifically: pixel point P combining the segmented target data t Pixel point P of the preset background data b And the corresponding weight coefficients are fused, and the specific formula is as follows:wherein P is new Representing the pixel points of the training set, wherein alpha and beta represent weight coefficients;
the first extraction module is used for processing the training set by adopting a Canny edge detection operator to obtain a target wire; the specific process of acquiring the target wire is as follows: denoising the training set by adopting Gaussian filtering, calculating a gradient module and a gradient direction, and reserving a maximum gradient value in the gradient direction; filling missing values of the target edges broken due to noise interference in the training set by adopting a hysteresis threshold to obtain the Canny edge detection operator processing result; extracting a wire in the Canny edge detection operator processing result according to wire characteristics, and acquiring the target wire, wherein the wire characteristics comprise images penetrating through the training set and a plurality of parallel lines;
the second extraction module is used for fusing deep features and shallow features in a preset SSD model according to the feature pyramid model, obtaining an improved SSD model, inputting the training set into the improved SSD model for conducting wire floater identification, and obtaining a target conducting wire floater;
the comparison module is used for acquiring a final target conductor floater if the target conductor is successfully compared with the target conductor floater; and if the position of the target wire is overlapped with the position of the target wire floater, determining that the target wire floater is the final target wire floater, otherwise, determining that the training set is input into the improved SSD model to identify the wire floater inaccurately.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879169.9A CN113627299B (en) | 2021-07-30 | 2021-07-30 | Wire floater intelligent recognition method and device based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110879169.9A CN113627299B (en) | 2021-07-30 | 2021-07-30 | Wire floater intelligent recognition method and device based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113627299A CN113627299A (en) | 2021-11-09 |
| CN113627299B true CN113627299B (en) | 2024-04-09 |
Family
ID=78382122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110879169.9A Active CN113627299B (en) | 2021-07-30 | 2021-07-30 | Wire floater intelligent recognition method and device based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113627299B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114842006B (en) * | 2022-07-04 | 2022-10-14 | 南方电网科学研究院有限责任公司 | Digital twin station power line detection method and related device thereof |
| CN115432331A (en) * | 2022-10-10 | 2022-12-06 | 浙江绿达智能科技有限公司 | Intelligent classification dustbin |
| CN115861359B (en) * | 2022-12-16 | 2023-07-21 | 兰州交通大学 | Self-adaptive segmentation and extraction method for water surface floating garbage image |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447530A (en) * | 2016-01-05 | 2016-03-30 | 国网四川省电力公司检修公司 | Power transmission line hidden risk and fault detection method based on image identification technology |
| CN108985170A (en) * | 2018-06-15 | 2018-12-11 | 山东信通电子股份有限公司 | Transmission line of electricity hanger recognition methods based on Three image difference and deep learning |
| CN111881937A (en) * | 2020-06-22 | 2020-11-03 | 深圳金三立视频科技股份有限公司 | Transmission line hardware target detection and defect identification method and terminal |
| CN111967335A (en) * | 2020-07-21 | 2020-11-20 | 广东工业大学 | Method for identifying foreign matters on power transmission line based on image processing |
| CN111985499A (en) * | 2020-07-23 | 2020-11-24 | 东南大学 | High-precision bridge apparent disease identification method based on computer vision |
| CN112613343A (en) * | 2020-12-01 | 2021-04-06 | 浙江大学 | Improved YOLOv 4-based river waste monitoring method |
| CN112800952A (en) * | 2021-01-27 | 2021-05-14 | 山东大学 | Marine organism identification method and system based on improved SSD algorithm |
-
2021
- 2021-07-30 CN CN202110879169.9A patent/CN113627299B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105447530A (en) * | 2016-01-05 | 2016-03-30 | 国网四川省电力公司检修公司 | Power transmission line hidden risk and fault detection method based on image identification technology |
| CN108985170A (en) * | 2018-06-15 | 2018-12-11 | 山东信通电子股份有限公司 | Transmission line of electricity hanger recognition methods based on Three image difference and deep learning |
| CN111881937A (en) * | 2020-06-22 | 2020-11-03 | 深圳金三立视频科技股份有限公司 | Transmission line hardware target detection and defect identification method and terminal |
| CN111967335A (en) * | 2020-07-21 | 2020-11-20 | 广东工业大学 | Method for identifying foreign matters on power transmission line based on image processing |
| CN111985499A (en) * | 2020-07-23 | 2020-11-24 | 东南大学 | High-precision bridge apparent disease identification method based on computer vision |
| CN112613343A (en) * | 2020-12-01 | 2021-04-06 | 浙江大学 | Improved YOLOv 4-based river waste monitoring method |
| CN112800952A (en) * | 2021-01-27 | 2021-05-14 | 山东大学 | Marine organism identification method and system based on improved SSD algorithm |
Non-Patent Citations (1)
| Title |
|---|
| 基于航拍图像的输电线路异物识别;金立军 等;同济大学学报(自然科学版);第41卷(第2期);第277-281页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113627299A (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113627299B (en) | Wire floater intelligent recognition method and device based on deep learning | |
| CN111784685B (en) | Power transmission line defect image identification method based on cloud edge cooperative detection | |
| CN109583369B (en) | Target identification method and device based on target area segmentation network | |
| CN109029381B (en) | Tunnel crack detection method and system and terminal equipment | |
| CN110310264B (en) | DCNN-based large-scale target detection method and device | |
| CN112199993B (en) | Method for identifying transformer substation insulator infrared image detection model in any direction based on artificial intelligence | |
| CN109711295B (en) | Optical remote sensing image offshore ship detection method | |
| CN112232349A (en) | Model training method, image segmentation method and device | |
| CN111753677B (en) | Multi-angle remote sensing ship image target detection method based on characteristic pyramid structure | |
| CN106651836B (en) | A kind of ground level detection method based on binocular vision | |
| CN112666553B (en) | Road ponding identification method and equipment based on millimeter wave radar | |
| CN114049356B (en) | Method, device and system for detecting structure apparent crack | |
| CN111639530B (en) | Method and system for detecting and identifying power transmission tower and insulator of power transmission line | |
| CN111160293A (en) | Small target ship detection method and system based on characteristic pyramid network | |
| CN110807758B (en) | Method, device, equipment and storage medium for detecting uncovered area of heat preservation quilt | |
| CN116279592A (en) | Method for dividing travelable area of unmanned logistics vehicle | |
| CN111814773A (en) | Lineation parking space identification method and system | |
| CN113628180A (en) | Semantic segmentation network-based remote sensing building detection method and system | |
| CN113311507A (en) | Typhoon path identification method and device | |
| CN105095846B (en) | Region growing seed point extracting method and system towards the segmentation of remote sensing images sea land | |
| CN112116561B (en) | Power grid transmission line detection method and device based on image processing fusion network weight | |
| CN110472472B (en) | Airport detection method and device based on SAR remote sensing image | |
| CN113095224A (en) | Method and device for identifying floaters in unmanned aerial vehicle patrolling river channel | |
| CN112633158A (en) | Power transmission line corridor vehicle identification method, device, equipment and storage medium | |
| CN111080562A (en) | Substation suspender identification method based on enhanced image contrast |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |