CN116385952B - Distribution network line small target defect detection method, device, equipment and storage medium - Google Patents
Distribution network line small target defect detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN116385952B CN116385952B CN202310636354.4A CN202310636354A CN116385952B CN 116385952 B CN116385952 B CN 116385952B CN 202310636354 A CN202310636354 A CN 202310636354A CN 116385952 B CN116385952 B CN 116385952B
- Authority
- CN
- China
- Prior art keywords
- region
- small target
- image
- defect detection
- interest
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 228
- 230000007547 defect Effects 0.000 title claims abstract description 209
- 238000000034 method Methods 0.000 claims abstract description 51
- 238000012545 processing Methods 0.000 claims abstract description 41
- 230000004927 fusion Effects 0.000 claims description 47
- 230000003287 optical effect Effects 0.000 claims description 34
- 238000005259 measurement Methods 0.000 claims description 12
- 230000001629 suppression Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 4
- 238000007689 inspection Methods 0.000 abstract description 20
- 230000008901 benefit Effects 0.000 abstract description 4
- 238000013135 deep learning Methods 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 20
- 230000008569 process Effects 0.000 description 17
- 238000010586 diagram Methods 0.000 description 15
- 239000012212 insulator Substances 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 239000004973 liquid crystal related substance Substances 0.000 description 4
- 230000002093 peripheral effect Effects 0.000 description 4
- 101100012526 Arabidopsis thaliana FAF3 gene Proteins 0.000 description 3
- 101100012527 Arabidopsis thaliana FAF4 gene Proteins 0.000 description 3
- 230000009471 action Effects 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000005764 inhibitory process Effects 0.000 description 3
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 3
- 102100022354 FAS-associated factor 2 Human genes 0.000 description 2
- 101000824586 Homo sapiens FAS-associated factor 2 Proteins 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/766—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using regression, e.g. by projecting features on hyperplanes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/70—Labelling scene content, e.g. deriving syntactic or semantic representations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
-
- 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
Abstract
The application provides a method, a device, equipment and a storage medium for detecting small target defects of a distribution network line, which relate to the technical field of deep learning and comprise the following steps: detecting a distribution network line image to obtain an interested region frame image; processing the block image of the region of interest to obtain a small target defect detection region image; and performing defect detection on the small target defect detection area image to obtain a small target defect detection result. The method has the advantages that the region of interest detection network and the improved defect detection network are used for detecting the distribution network line image, the region of interest frame image is obtained, the small target defect detection region image is further obtained, finally, the detection of the small target defect is realized, the working time of electric power inspection personnel is greatly shortened, false detection and missing detection phenomena are effectively prevented, the discovery rate is improved, and the false alarm rate is reduced.
Description
Technical Field
The application relates to the technical field of deep learning, in particular to a method, a device, equipment and a storage medium for detecting small target defects of a distribution network line.
Background
With the extremely rapid increase of the power consumption of the power grid, the defect overhaul difficulty of the distribution network line is also increased. The safety pin in the distribution network line falls off, the spring pin falls off, the connecting fitting ball head is corroded, the wire clamp insulating cover falls off and the like, and the safety pin is used as an important small target defect type of the distribution network, and if the small target defect type cannot be found in time and the defect is overhauled in time, faults such as line leakage and power failure are easily generated, so that the safe operation of the distribution network line is seriously threatened. In order to quickly find small target defect types in distribution network lines and timely eliminate hidden dangers of the distribution network defects, the power grid departments develop regular or irregular distribution network line inspection work.
At present, a one-stage detection algorithm or a multi-stage detection algorithm is often adopted for detecting the defects of the distribution network, but the small targets of the distribution network have the characteristics of small duty ratio, large target quantity and the like, so that the situation that the discovery rate is low and the false alarm rate is high often occurs when the one-stage detection algorithm or the multi-stage detection algorithm is directly adopted for detecting the defects of the distribution network.
Disclosure of Invention
Accordingly, an object of the embodiments of the present application is to provide a method, an apparatus, a device, and a storage medium for detecting a small target defect of a distribution network line, which detect a distribution network line image by using a PP-YOLOE detection network and an improved cascades-RCNN detection network to obtain a region frame image of interest, further obtain a small target defect detection region image, and finally realize detection of a small target defect, so that the working time of an electric power inspection personnel is greatly reduced, and false inspection and omission inspection phenomena are effectively prevented, thereby solving the technical problems described above.
In a first aspect, an embodiment of the present application provides a method for detecting a small target defect of a distribution network line, where the method includes: detecting a distribution network line image to obtain an interested region frame image; processing the block image of the region of interest to obtain a small target defect detection region image; and performing defect detection on the small target defect detection area image to obtain a small target defect detection result.
In the implementation process, the region of interest detection network and the improved defect detection network are used for detecting the distribution network line image to obtain the region of interest frame image, the small target defect detection region image is further obtained, and finally the detection of the small target defect is realized, so that the working time of electric power inspection personnel is greatly reduced, the false detection and missing detection phenomena are effectively prevented, the discovery rate is improved, and the false alarm rate is reduced.
Optionally, the detecting the distribution network line image to obtain a region of interest frame image includes: detecting the distribution network line image by adopting a PP-YOLOE detection network to obtain an overlapping frame image of the region of interest; and performing non-maximum suppression processing on the region of interest overlapped frame image to obtain the region of interest frame image.
In the implementation process, the PP-YOLOE detection network algorithm is used for detecting the distribution network line image, so that the region of interest where the small target is located can be accurately detected, the region-of-interest frame image is obtained, subsequent detection of the small target defect is facilitated, and the defect detection accuracy is improved.
Optionally, the processing the block image of the region of interest to obtain a small target defect detection region image includes: merging and sliding window processing are carried out on the region of interest frames in the region of interest frame image, so that a region of interest sliding window frame is obtained; IOU calculation is carried out on the sliding window frame of the region of interest and the region of interest frame, and a measurement standard value is obtained; and if the measurement standard value corresponding to the region of interest sliding window frame exceeds a preset threshold value, determining the image of the region where the region of interest sliding window frame is located as a small target defect detection region image.
In the implementation process, the small target defect detection area image is obtained by combining the block images of the region of interest, sliding the window and performing equal correlation processing on the threshold ratio, so that the small target defect detection area image is convenient to detect the small target defect later, the resolution ratio of an input network is unified, the discovery rate of the small target defect is improved, the false detection ratio of the small target defect is reduced, and the speed of detecting the small target defect is improved.
Optionally, the processing the block image of the region of interest to obtain a small target defect detection region image further includes: and if the region-of-interest frame image is not detected, determining an image in which the central region of the distribution network line image is positioned as a small target defect detection region image.
In the implementation process, when the region-of-interest frame image is not detected, the central region of the distribution network line image is cut to serve as a small target defect detection region image, so that the discovery rate lost when the small target region-of-interest is not detected is compensated, and the discovery rate of small target defect detection is improved.
Optionally, the performing defect detection on the small target defect detection area image to obtain a small target defect detection result includes: inputting the small target defect detection area image into a ResNet50 backbone network convolution layer to perform feature extraction, and obtaining a feature map; inputting the feature map into an improved pyramid network for fusion to obtain a cross-level fusion feature map; and inputting the cross-level fusion feature map into a cascade detection network for classification and regression to obtain a small target defect detection result.
In the implementation process, the small target defect detection result is obtained by carrying out feature extraction, feature fusion, classification and regression on the small target defect detection area image, so that the small target defect of the distribution network can be effectively detected, and the speed and efficiency of small target defect detection are improved.
Optionally, inputting the feature map to an improved pyramid network for fusion to obtain a cross-level fusion feature map, including: convolving and upsampling the high-layer low-resolution feature map in the feature map to obtain a high-layer same-resolution feature map; convolving a low-layer high-resolution feature map in the feature map to obtain a low-layer same-resolution feature map; performing channel splicing and convolution processing on the high-layer same-resolution feature map and the low-layer same-resolution feature map to obtain an optical flow field feature map; and performing distortion alignment processing on the optical flow field feature images, and fusing the optical flow field feature images with the low-layer high-resolution feature images to obtain cross-layer fusion feature images.
In the implementation process, the cross-level fusion feature map is obtained by carrying out optical flow alignment and feature fusion on the feature map of the small target defect detection area, so that alignment learning of the high-level and low-level feature map is realized, semantic information of the high-level feature map is obtained, detail information of the low-level feature map is fused, cross-level high-level semantic feature fusion is realized, and further the accuracy of detecting the small target defects of power distribution is effectively improved.
Optionally, inputting the cross-level fusion feature map to a cascade detection network for classification and regression to obtain a small target defect detection result, including: performing three-level classification and regression on the cross-level fusion feature map to obtain a high-quality detection frame image; performing non-maximum suppression processing on the high-quality detection block image to obtain a small target defect detection result; wherein the small target defect detection result includes: the safety pin falls off, the spring pin falls off, the connecting fitting ball head is corroded, and the wire clamp insulating cover falls off.
In the implementation process, the small target defect detection is completed by classifying and regressing the fusion feature map of the small target defect detection area and performing non-maximum value inhibition processing, so that the accuracy of the small target defect detection of the power distribution is effectively improved.
In a second aspect, an embodiment of the present application provides a device for detecting a small target defect of a distribution network line, where the device includes: the area detection module is used for detecting the distribution network line image to obtain an interested area frame image; the small target area determining module is used for processing the block images of the region of interest to obtain small target defect detection area images; and the small target defect detection module is used for carrying out defect detection on the small target defect detection area image to obtain a small target defect detection result.
In a third aspect, an embodiment of the present application further provides an electronic device, including: a processor, a memory storing machine-readable instructions executable by the processor, which when executed by the processor perform the steps of the method described above when the electronic device is run.
In a fourth aspect, embodiments of the present application provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method described above.
The application has the advantages that: the PP-YOLOE detection network and the improved Cascade-RCNN detection network are used for detecting the distribution network line image to obtain the region-of-interest frame image, further obtain the small target defect detection region image, finally realize the detection of the small target defect, greatly reduce the working time of electric power inspection personnel and effectively prevent false inspection and omission inspection. In particular, by adopting the FAMFPN improved pyramid network to perform optical flow alignment and feature fusion on the small target defect detection area feature images, cross-level fusion feature images are obtained, alignment learning of high-level and low-level feature images is realized, semantic information of the high-level feature images is obtained, detail information of the low-level feature images is fused, cross-level high-level semantic feature fusion is realized, and further accuracy of power distribution small target defect detection is effectively improved.
In order to make the above objects, features and advantages of the present application more comprehensible, embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and should not be considered as limiting the scope, and other related drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for detecting a small target defect of a distribution network line according to an embodiment of the present application;
fig. 2 is an exemplary diagram of a method for detecting a small target defect of a distribution network line according to an embodiment of the present application;
FIG. 3 is a network structure diagram of an improved Cascade-RCNN algorithm provided by an embodiment of the application;
fig. 4 is a network structure diagram of a FAM module provided in an embodiment of the present application;
fig. 5 is a schematic functional block diagram of a small target defect detection device for a distribution network line according to an embodiment of the present application;
fig. 6 is a schematic block diagram of an electronic device of a small target defect detection device for a distribution network according to an embodiment of the present application.
Icon: 210-a region detection module; 220-a small target area determination module; 230-a small target defect detection module; 300-an electronic device; 311-memory; 312-a storage controller; 313-processor; 314-peripheral interface; 315-an input-output unit; 316-display unit.
Detailed Description
The following description of the embodiments of the present application 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 application, but not all embodiments. The components of the embodiments of the present application generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the application, as presented in the figures, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present application.
It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. 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 one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. The terms "first," "second," and the like are used merely to distinguish between descriptions and are not to be construed as indicating or implying relative importance.
Before describing the embodiments of the present application, a brief description will be first made of the technical concept related to the present application.
Intersection Over Union (IOU): is a standard for measuring the accuracy of detecting a corresponding object in a specific dataset, IOU is a simple measurement standard, and IOU can be used for measurement as long as a task of deriving a prediction horizon (prediction boxes) in the output.
The inventors noted that there are two defect detection modes in the prior art: (1) The convolutional neural network is adopted to carry out model training on the defect samples of the power distribution equipment and the defect simulation samples generated by the generation countermeasure network, the trained defect recognition model is deployed on the mobile terminal, and automatic defect recognition of the inspection image of the power distribution unmanned aerial vehicle is completed, so that inspection workers can effectively carry out manual verification, and defect parts can be rapidly developed and overhauled, and the intelligent inspection efficiency of the power grid is greatly improved. However, the method only adopts a convolutional neural network to identify the distribution network defects, and the method directly identifies the small target defects (such as falling-off of a safety pin, falling-off of a spring pin, rust of a connecting fitting ball, falling-off of a wire clamp insulating cover and the like) of the distribution network, so that the discovery rate is very low, and the requirement of automatic detection of the small target defects cannot be met. (2) And detecting the insulator region by using a yolov5 algorithm, and detecting the insulator defect of the cut insulator region by using the yolov5 algorithm again, so that the defect detection precision and the detection efficiency of the insulator are improved. However, the mode only adopts the yolov5 algorithm to identify the defects of the distribution network insulators, and identifies the defects of the small targets of the distribution network, and the discovery rate is very low because the small target defects are not covered in the first-stage interested region, so that the requirement of automatic defect detection of the small targets cannot be met. In view of the above, the embodiment of the application provides a distribution network line small target defect detection method, which can greatly reduce the working time of power inspection personnel and effectively prevent false inspection and missing inspection by using a deep learning algorithm to automatically detect the distribution network unmanned aerial vehicle image small target defect.
Referring to fig. 1, fig. 1 is a flowchart of a method for detecting a small target defect of a distribution network line according to an embodiment of the present application, and the method is explained in detail below, and includes: step 100, step 120 and step 140.
Step 100: detecting a distribution network line image to obtain an interested region frame image;
step 120: processing the block image of the region of interest to obtain a small target defect detection region image;
step 140: and performing defect detection on the small target defect detection area image to obtain a small target defect detection result.
Illustratively, the distribution network line image may be: the unmanned aerial vehicle shooting equipment is utilized to carry out the unmanned aerial vehicle inspection image of the distribution network or the distribution network line inspection image shot by other shooting equipment on the distribution network, and the image can generally comprise various small target defects of the distribution network, such as falling off of a safety pin, falling off of a spring pin, rust of a connecting fitting ball head, falling off of a wire clamp insulating cover and the like. The region of interest frame image may be: and utilizing the image of the region of interest detected by the target detection network algorithm.
Alternatively, as shown in fig. 2, the process of performing small target defect detection on the distribution network line image may be generally divided into three parts: region of interest detection, small target defect detection region image generation, small target defect detection. (1) region of interest detection. Detecting a small target region of interest of the inspection image of the unmanned aerial vehicle of the distribution network through a target detection network algorithm, and if the small target region of interest is not detected, the small target region of interest list of the image is empty and is stored; if the small target region of interest is detected, the small target region of interest type and the position information are recorded into a small target region of interest list of the image and stored. (2) small target defect detection area image generation. If the small target region of interest is empty, in order to improve the small target defect discovery rate, a certain-scale image center region can be selected as a small target detection region, and a small target center region image can be generated by cutting and is used as a small target detection region image. If the small target region of interest list is not empty, all the regions of interest are processed and then determined to be small target detection region images. (3) small target defect detection. Detecting a small target defect detection area image through an improved detection network algorithm, and if a small target defect result is not detected, recording that a small target defect list of the area image is empty; if the small target defect is detected, the small target defect type and the position information are recorded in the small target defect list of the area image, and when all the small target area images are detected, the small target defect lists of all the area images are combined to obtain a final small target defect detection result of the distribution network, the final small target defect detection result is stored, and the defect detection flow of the images is completed.
The method has the advantages that the region of interest detection network and the improved defect detection network algorithm are used for detecting the distribution network line image, the region of interest frame image is obtained, the small target defect detection region image is further obtained, finally, the detection of the small target defect is realized, the working time of electric power inspection personnel is greatly reduced, false detection and missing detection phenomena are effectively prevented, the discovery rate is improved, and the false alarm rate is reduced.
In one embodiment, step 100 may include: step 101 and step 102.
Step 101: detecting the distribution network line image by adopting a PP-YOLOE detection network to obtain an overlapping frame image of the region of interest;
step 102: and performing non-maximum suppression processing on the region of interest overlapped frame image to obtain the region of interest frame image.
For example, as the resolution of the inspection image of the distribution network unmanned aerial vehicle is high, the scene is complex and changeable, the structure of the small target region of interest is relatively simple, the scale change is larger, the number of the regions of interest is larger, the dense overlapping rate is high, and in order to accurately detect the small target region of interest so as to detect the small target defect subsequently, the PP-YOLOE detection network algorithm can be adopted to detect the distribution network line image.
Optionally, the PP-YOLOE detection network adopts an Anchor-free Anchor box-free mechanism, and the extensible backbones and Neck are designed, so that the configuration of multiple sizes is more convenient and flexible. Meanwhile, a dynamic matching strategy of a more efficient label distribution strategy TAL (Task Alignment Learning) is introduced, so that the common problem of unbalanced classification regression in a target detection task is solved, and the detection precision is improved. On the basis, the ET-Head (effective Task-aligned Head) is designed more simply, and the precision is improved at the cost of a small speed loss. The distribution network line image is detected through a PP-YOLOE detection network algorithm, a large number of accurate small target region of interest overlapping frames can be obtained, and then an ideal region of interest frame can be obtained through non-maximum suppression (NMS) processing, so that a region of interest frame image is obtained.
By detecting the distribution network line image by using the PP-YOLOE detection network algorithm, the region of interest where the small target is located can be accurately detected, the region-of-interest frame image is obtained, the subsequent detection of the small target defect is facilitated, and the defect detection accuracy is improved.
In one embodiment, step 120 may include: step 121, step 122 and step 123.
Step 121: merging and sliding window processing are carried out on the region of interest frames in the region of interest frame image, so that a region of interest sliding window frame is obtained;
step 122: IOU calculation is carried out on the sliding window frame of the region of interest and the region of interest frame, and a measurement standard value is obtained;
step 123: if the measurement standard value corresponding to the region of interest sliding window frame exceeds a preset threshold value, determining an image of the region where the region of interest sliding window frame is located as a small target defect detection region image.
For example, because the resolutions of the small target interested areas are different and the number of the small target interested areas is dense, the small target interested areas are directly cut for detection, so that the small target detection is easy to be caused, the small target detection is easy to be caused in the dense areas, and the quality and the efficiency of the small target defect detection are reduced. Therefore, in order to unify the resolution of the input network, the discovery rate of the small target defects is improved, the false detection ratio of the small target defects is reduced, the detection speed of the small target defects is accelerated, and the region image generation algorithm can be used for processing the region block image of interest, so that the small target defect detection region image is generated.
Optionally, firstly, merging all the region frames of interest, counting the size range of the small defect target, the size range of the region frames of interest and the size range of the merged frame, comprehensively selecting a certain scale window, and carrying out sliding window processing on the merged frame, wherein the size of the scale window is 1200 x 1200; then, calculating IOU of the sliding window frame and the interested region frame, if the IOU is lower than a certain threshold value, for example, the threshold value can be set to 0.1, considering that a small power distribution target is hardly existed in the sliding window frame region, discarding the sliding window frame, and not carrying out subsequent defect detection; if the IOU is not lower than the threshold value of 0.1, the sliding window frame area possibly contains a small power distribution target, the sliding window frame is an effective sliding window frame of a small target region of interest, and an effective sliding window image of the small target region of interest is generated through clipping and is used as a small target defect detection area image.
Particularly, a sliding window detection strategy is adopted, the small target interested areas are combined firstly, then the combining frame is subjected to sliding window, and then the sliding window frame is filtered, so that the resolution of the small target defect detection network input is unified, the detection quantity of the small target defect detection network input is effectively controlled, and a foundation is laid for the accuracy and the speed of small target defect detection.
The small target defect detection area image is obtained by combining the block images of the interested area, sliding the window and comparing the threshold value with the equivalent correlation processing, so that the small target defect can be detected conveniently and subsequently, the resolution of an input network is unified, the discovery rate of the small target defect is improved, the false detection ratio of the small target defect is reduced, and the detection speed of the small target defect is improved.
In one embodiment, step 120 may further comprise: step 121.
Step 121: if the region-of-interest frame image is not detected, determining an image in which the central region of the distribution network line image is located as a small target defect detection region image.
For example, when the distribution network line image belongs to a non-matched failure image, or the real and effective distribution network line image cannot detect the region-of-interest frame image due to various reasons such as shielding, blurring, and clipping, that is, when the small target region-of-interest is empty, in order to improve the small target defect discovery rate and compensate for the loss discovery rate caused by the fact that the small target region-of-interest is not detected, a certain-scale image center region may be selected as the small target detection region, for example, the center region has the size of: 1200 by 1200, and cropping to generate a small target center area image as a small target detection area image.
When the region-of-interest frame image is not detected, the central region of the distribution network line image is cut to serve as a small target defect detection region image, so that the discovery rate lost due to the fact that the small target region-of-interest is not detected is made up, and the discovery rate of small target defect detection is improved.
In one embodiment, step 140 may include: step 141, step 142 and step 143.
Step 141: inputting the small target defect detection area image into a ResNet50 backbone network convolution layer to perform feature extraction, and obtaining a feature map;
step 142: inputting the feature map into an improved pyramid network for fusion to obtain a cross-level fusion feature map;
step 143: and inputting the cross-level fusion feature map into a cascade detection network for classification and regression to obtain a small target defect detection result.
Illustratively, because the distribution network line image resolution is high, for example: 5000 x 3000, small power distribution target defects occupy smaller space in an image, and in addition, the scene where the small power distribution target defects are located is complex and changeable, the effect of directly detecting the small power distribution target is not good, a dense region of interest is cut for detection, and the small target defects are not found high due to different resolutions of an input network and more dense regions, so that the detection speed is slow. Therefore, the embodiment of the application firstly combines the target areas, obtains the network input images with consistent resolution by utilizing the sliding window idea, and then detects by utilizing the improved Cascade-RCNN algorithm, thereby effectively detecting the small target defects of the distribution network.
Alternatively, as shown in fig. 3, defect detection is performed using a detection algorithm combined with a res net50 backbone network, a FAMFPN improvement pyramid, RPN network detection, cascades detection, NMS processing, and the like. Firstly, a small target defect detection area image is subjected to ResNet50 backbone network convolution layers (Conv 1-Conv 5) to obtain 4-layer feature graphs (F2-F5), and the 4-layer feature graphs are input into an improved feature pyramid network FAMFPN. Then, the feature map F5 passes through a 1x1 convolution layer to obtain a fusion feature map P5; the feature images F4 and P5 generate an optical flow alignment feature image FAF4 through an FAM (optical flow alignment module), and the FAF4 and F4 are fused to generate a fusion feature image P4; the feature map F3 and the feature map FAF4 generate an optical flow alignment feature map FAF3 through an FAM (optical flow alignment module), and the FAF3 and the F3 are fused to generate a fusion feature map P3; the feature map F2 and the feature map FAF3 generate an optical flow alignment feature map FAF2 through an FAM (optical flow alignment module), the FAF2 and the F2 are fused, and then a fusion feature map P2 is generated, so that cross-layer fusion feature maps (P2-P5) with stronger expression capability are obtained. Then, the feature maps P2-P5 can obtain preliminary candidate frames by utilizing an RPN detection network, and the preliminary candidate frames are sent into a Cascade detection network Cascade for classification detection, so that a small target defect detection result is obtained.
The small target defect detection result is obtained by carrying out feature extraction, feature fusion, classification and regression on the small target defect detection area image, so that the small target defect of the distribution network can be effectively detected, and the speed and efficiency of small target defect detection are improved.
In one embodiment, step 142 may include: step 1421, step 1422, step 1423, and step 1424.
Step 1421: convolving and upsampling the high-layer low-resolution feature map in the feature map to obtain a high-layer same-resolution feature map;
step 1422: convolving the low-layer high-resolution feature map in the feature map to obtain a low-layer same-resolution feature map;
step 1423: performing channel splicing and convolution processing on the high-layer same-resolution feature map and the low-layer same-resolution feature map to obtain an optical flow field feature map;
step 1424: and performing distortion alignment treatment on the optical flow field characteristic diagram, and fusing the optical flow field characteristic diagram with the low-layer high-resolution characteristic diagram to obtain a cross-layer fusion characteristic diagram.
Illustratively, as shown in fig. 4, which shows a network structure diagram of a FAM module, the process of inputting a feature map into an improved pyramid network for fusion may be: firstly, a high-layer low-resolution characteristic diagram is convolved and up-sampled through 1x1 to generate a characteristic diagram with the same resolution as that of the low-layer high-resolution characteristic diagram; then, carrying out channel splicing on the characteristic images subjected to 1x1 convolution with the low-layer high-resolution characteristic images, and then carrying out 3x3 convolution to form an optical flow field with the same resolution as the low-layer high-resolution characteristic images, wherein the characteristic value generated by the optical flow field is the offset from the low-resolution characteristic images to the high-resolution characteristic images; finally, the optical flow field and the high-layer low-resolution feature map are subjected to distortion alignment treatment to obtain an optical flow alignment feature map (for example, FAF 2-FAF 5 in step 142), and then are fused with the low-layer high-resolution feature map (for example, F2-F5 in step 142) to obtain a cross-layer fusion feature map (for example, P2-P5 in step 142); wherein the resolution of the optical flow alignment feature map is consistent with the resolution of the low-level high-resolution feature map.
The process of the warp alignment process may be: first, the optical flow field coordinates and the optical flow field coordinate offset generated by the optical flow field are mapped into the low resolution feature map to generate corresponding low resolution feature map coordinates. Then, the low-resolution feature map feature values are calculated through bilinear interpolation, and the optical flow field feature value offset is added, so that a high-resolution feature map distortion alignment value is obtained, and the specific calculation formula is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,，/>warp alignment values for high resolution feature map, +.>For the low-resolution feature map feature values,for the optical flow field characteristic value offset, +.>For high resolution feature map coordinates, +.>For optical flow field coordinate offset, +.>For low resolution feature map coordinates, +.>Bilinear interpolation neighborhood coordinate set for low resolution feature map,/for low resolution feature map>Neighborhood coordinate weights are bilinear interpolated for low resolution feature maps.
By carrying out optical flow alignment and feature fusion on the feature map of the small target defect detection area, a cross-level fusion feature map is obtained, alignment learning of high-low-level feature maps is carried out by using a FAMFPN network, so that semantic information of the high-level feature map is obtained, detail information of the low-level feature map is fused, cross-level high-level semantic feature fusion is realized, and further accuracy of power distribution small target defect detection is effectively improved.
In one embodiment, step 143 may include: step 1431 and step 1432.
Step 1431: three-level classification and regression are carried out on the cross-level fusion feature images, and a high-quality detection frame image is obtained;
step 1432: performing non-maximum suppression processing on the high-quality detection block image to obtain a small target defect detection result; the small target defect detection result comprises: the safety pin falls off, the spring pin falls off, the connecting fitting ball head is corroded, and the wire clamp insulating cover falls off.
After optical flow alignment and feature fusion are carried out on the small target defect detection region feature map, the ROI alignment is utilized to unify feature scale, three-level classification and regression are continuously carried out, and the obtained detection frame with higher IOU quality is subjected to non-maximum value inhibition processing, so that accurate detection of the small target defects of the distribution network is completed. If the detection result has no small target defect, recording that a small target defect list of the image is empty, and storing, wherein the image detection flow is finished; if any small target defects such as the falling of the safety pin, the falling of the spring pin, the rust of the bulb of the connecting fitting, the falling of the insulating cover of the wire clamp and the like exist in the detection result, the type and the position information of the small target defects are recorded into a small target defect list of the image and stored, and the image detection flow is finished.
And the small target defect detection is completed by classifying and regressing the fusion feature map of the small target defect detection area and performing non-maximum value inhibition treatment, so that the accuracy of the small target defect detection of the power distribution is effectively improved.
Referring to fig. 5, fig. 5 is a schematic functional block diagram of a small target defect detection device for a distribution network according to an embodiment of the present application. The device comprises: the region detection module 210, the small target region determination module 220, and the small target defect detection module 230.
The area detection module 210 is configured to detect a distribution network line image to obtain an area frame image of interest;
the small target area determining module 220 is configured to process the block image of the region of interest to obtain a small target defect detection area image;
and the small target defect detection module 230 is configured to detect the small target defect detection area image to obtain a small target defect detection result.
Alternatively, the region detection module 210 may be configured to:
detecting the distribution network line image by adopting a PP-YOLOE detection network to obtain an overlapping frame image of the region of interest;
and performing non-maximum suppression processing on the region of interest overlapped frame image to obtain the region of interest frame image.
Alternatively, the small target area determination module 220 may be configured to:
merging and sliding window processing are carried out on the region of interest frames in the region of interest frame image, so that a region of interest sliding window frame is obtained;
IOU calculation is carried out on the sliding window frame of the region of interest and the region of interest frame, and a measurement standard value is obtained;
and if the measurement standard value corresponding to the region of interest sliding window frame exceeds a preset threshold value, determining the image of the region where the region of interest sliding window frame is located as a small target defect detection region image.
Alternatively, the small target area determination module 220 may be configured to:
and if the region-of-interest frame image is not detected, determining an image in which the central region of the distribution network line image is positioned as a small target defect detection region image.
Alternatively, the small target defect detection module 230 may be configured to:
inputting the small target defect detection area image into a ResNet50 backbone network convolution layer to perform feature extraction, and obtaining a feature map;
inputting the feature map into an improved pyramid network for fusion to obtain a cross-level fusion feature map;
and inputting the cross-level fusion feature map into a cascade detection network for classification and regression to obtain a small target defect detection result.
Alternatively, the small target defect detection module 230 may be configured to:
convolving and upsampling the high-layer low-resolution feature map in the feature map to obtain a high-layer same-resolution feature map;
convolving a low-layer high-resolution feature map in the feature map to obtain a low-layer same-resolution feature map;
performing channel splicing and convolution processing on the high-layer same-resolution feature map and the low-layer same-resolution feature map to obtain an optical flow field feature map;
and performing distortion alignment processing on the optical flow field feature images, and fusing the optical flow field feature images with the low-layer high-resolution feature images to obtain cross-layer fusion feature images.
Alternatively, the small target defect detection module 230 may be configured to:
performing three-level classification and regression on the cross-level fusion feature map to obtain a high-quality detection frame image;
performing non-maximum suppression processing on the high-quality detection block image to obtain a small target defect detection result; wherein the small target defect detection result includes: the safety pin falls off, the spring pin falls off, the connecting fitting ball head is corroded, and the wire clamp insulating cover falls off.
Referring to fig. 6, fig. 6 is a block schematic diagram of an electronic device. The electronic device 300 may include a memory 311, a memory controller 312, a processor 313, a peripheral interface 314, an input output unit 315, a display unit 316. It will be appreciated by those of ordinary skill in the art that the configuration shown in fig. 6 is merely illustrative and is not limiting of the configuration of the electronic device 300. For example, electronic device 300 may also include more or fewer components than shown in FIG. 6, or have a different configuration than shown in FIG. 6.
The above-mentioned memory 311, memory controller 312, processor 313, peripheral interface 314, input/output unit 315, and display unit 316 are electrically connected directly or indirectly to each other to realize data transmission or interaction. For example, the components may be electrically connected to each other via one or more communication buses or signal lines. The processor 313 is used to execute executable modules stored in the memory.
The Memory 311 may be, but is not limited to, a random access Memory (Random Access Memory, RAM), a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc. The memory 311 is configured to store a program, and the processor 313 executes the program after receiving an execution instruction, and a method executed by the electronic device 300 defined by the process disclosed in any embodiment of the present application may be applied to the processor 313 or implemented by the processor 313.
The processor 313 may be an integrated circuit chip having signal processing capabilities. The processor 313 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but also digital signal processors (digital signal processor, DSP for short), application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), field Programmable Gate Arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The peripheral interface 314 couples various input/output devices to the processor 313 and the memory 311. In some embodiments, the peripheral interface 314, the processor 313, and the memory controller 312 may be implemented in a single chip. In other examples, they may be implemented by separate chips.
The input/output unit 315 is used for providing input data to a user. The input/output unit 315 may be, but is not limited to, a mouse, a keyboard, and the like.
The display unit 316 provides an interactive interface (e.g., a user interface) between the electronic device 300 and a user for reference. In this embodiment, the display unit 316 may be a liquid crystal display or a touch display. The liquid crystal display or the touch display may display a process of executing the program by the processor.
The electronic device 300 in this embodiment may be used to perform each step in each method provided in the embodiment of the present application.
Furthermore, the embodiment of the present application also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, performs the steps in the above-mentioned method embodiments.
The computer program product of the above method according to the embodiments of the present application includes a computer readable storage medium storing a program code, where instructions included in the program code may be used to execute steps in the above method embodiment, and specifically, reference may be made to the above method embodiment, which is not described herein.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, and the division of the modules is merely a logical function division, and there may be additional divisions when actually implemented, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form. The functional modules in the embodiment of the application can be integrated together to form a single part, or each module can exist alone, or two or more modules can be integrated to form a single part.
It should be noted that the functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM) random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
In this document, 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.
The above description is only an example of the present application and is not intended to limit the scope of the present application, and various modifications and variations will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (8)
1. The method for detecting the small target defects of the distribution network line is characterized by comprising the following steps:
detecting a distribution network line image to obtain an interested region frame image;
processing the block image of the region of interest to obtain a small target defect detection region image;
performing defect detection on the small target defect detection area image to obtain a small target defect detection result;
the processing the block image of the region of interest to obtain a small target defect detection region image includes: merging and sliding window processing are carried out on the region of interest frames in the region of interest frame image, so that a region of interest sliding window frame is obtained; IOU calculation is carried out on the sliding window frame of the region of interest and the region of interest frame, and a measurement standard value is obtained; if the measurement standard value corresponding to the region of interest sliding window frame exceeds a preset threshold value, determining an image of the region where the region of interest sliding window frame is located as a small target defect detection region image; the processing the block image of the region of interest to obtain a small target defect detection region image, further includes: and if the region-of-interest frame image is not detected, determining an image in which the central region of the distribution network line image is positioned as a small target defect detection region image.
2. The method of claim 1, wherein detecting the distribution network line image to obtain the region of interest frame image comprises:
detecting the distribution network line image by adopting a PP-YOLOE detection network to obtain an overlapping frame image of the region of interest;
and performing non-maximum suppression processing on the region of interest overlapped frame image to obtain the region of interest frame image.
3. The method according to claim 1, wherein performing defect detection on the small target defect detection area image to obtain a small target defect detection result comprises:
inputting the small target defect detection area image into a ResNet50 backbone network convolution layer to perform feature extraction, and obtaining a feature map;
inputting the feature map into an improved pyramid network for fusion to obtain a cross-level fusion feature map;
and inputting the cross-level fusion feature map into a cascade detection network for classification and regression to obtain a small target defect detection result.
4. A method according to claim 3, wherein said inputting the feature map into an improved pyramid network for fusion to obtain a cross-level fusion feature map comprises:
Convolving and upsampling the high-layer low-resolution feature map in the feature map to obtain a high-layer same-resolution feature map;
convolving a low-layer high-resolution feature map in the feature map to obtain a low-layer same-resolution feature map;
performing channel splicing and convolution processing on the high-layer same-resolution feature map and the low-layer same-resolution feature map to obtain an optical flow field feature map;
and performing distortion alignment processing on the optical flow field feature images, and fusing the optical flow field feature images with the low-layer high-resolution feature images to obtain cross-layer fusion feature images.
5. The method of claim 3, wherein inputting the cross-level fusion feature map to a cascade detection network for classification and regression to obtain a small target defect detection result comprises:
performing three-level classification and regression on the cross-level fusion feature map to obtain a high-quality detection frame image;
performing non-maximum suppression processing on the high-quality detection block image to obtain a small target defect detection result; wherein the small target defect detection result includes: the safety pin falls off, the spring pin falls off, the connecting fitting ball head is corroded, and the wire clamp insulating cover falls off.
6. A distribution network line small target defect detection device, the device comprising:
the area detection module is used for detecting the distribution network line image to obtain an interested area frame image;
the small target area determining module is used for processing the block images of the region of interest to obtain small target defect detection area images; the small target area determining module is used for: merging and sliding window processing are carried out on the region of interest frames in the region of interest frame image, so that a region of interest sliding window frame is obtained; IOU calculation is carried out on the sliding window frame of the region of interest and the region of interest frame, and a measurement standard value is obtained; if the measurement standard value corresponding to the region of interest sliding window frame exceeds a preset threshold value, determining an image of the region where the region of interest sliding window frame is located as a small target defect detection region image; the small target area determination module is further configured to: if the region-of-interest frame image is not detected, determining an image in which a central region of the distribution network line image is located as a small target defect detection region image;
and the small target defect detection module is used for carrying out defect detection on the small target defect detection area image to obtain a small target defect detection result.
7. An electronic device, comprising: a processor, a memory storing machine-readable instructions executable by the processor, which when executed by the processor perform the steps of the method of any of claims 1 to 5 when the electronic device is run.
8. A computer-readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, performs the steps of the method according to any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310636354.4A CN116385952B (en) | 2023-06-01 | 2023-06-01 | Distribution network line small target defect detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310636354.4A CN116385952B (en) | 2023-06-01 | 2023-06-01 | Distribution network line small target defect detection method, device, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116385952A CN116385952A (en) | 2023-07-04 |
| CN116385952B true CN116385952B (en) | 2023-09-01 |
Family
ID=86971379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310636354.4A Active CN116385952B (en) | 2023-06-01 | 2023-06-01 | Distribution network line small target defect detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116385952B (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107493432A (en) * | 2017-08-31 | 2017-12-19 | 广东欧珀移动通信有限公司 | Image processing method, device, mobile terminal and computer-readable recording medium |
| CN109165644A (en) * | 2018-07-13 | 2019-01-08 | 北京市商汤科技开发有限公司 | Object detection method and device, electronic equipment, storage medium, program product |
| CN110210474A (en) * | 2019-04-30 | 2019-09-06 | 北京市商汤科技开发有限公司 | Object detection method and device, equipment and storage medium |
| CN111339891A (en) * | 2020-02-20 | 2020-06-26 | 苏州浪潮智能科技有限公司 | Target detection method of image data and related device |
| 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 |
| CN112801230A (en) * | 2021-04-07 | 2021-05-14 | 国网江西省电力有限公司电力科学研究院 | Intelligent acceptance method for unmanned aerial vehicle of power distribution line |
| CN113536981A (en) * | 2021-06-28 | 2021-10-22 | 华雁智能科技(集团)股份有限公司 | Transformer substation foreign matter identification detection method, device and system |
| CN113642486A (en) * | 2021-08-18 | 2021-11-12 | 国网江苏省电力有限公司泰州供电分公司 | Unmanned aerial vehicle distribution network inspection method with airborne front-end identification model |
| WO2022036953A1 (en) * | 2020-08-19 | 2022-02-24 | 上海商汤智能科技有限公司 | Defect detection method and related apparatus, device, storage medium, and computer program product |
| CN114170144A (en) * | 2021-11-11 | 2022-03-11 | 国网福建省电力有限公司漳州供电公司 | Power transmission line pin defect detection method, equipment and medium |
| WO2022082856A1 (en) * | 2020-10-19 | 2022-04-28 | 广东科凯达智能机器人有限公司 | Method and system for automatically identifying and tracking inspection target, and robot |
| CN115187940A (en) * | 2022-06-09 | 2022-10-14 | 中汽创智科技有限公司 | Image detection method and device for vehicle |
| CN115239642A (en) * | 2022-07-01 | 2022-10-25 | 华雁智能科技(集团)股份有限公司 | Detection method, detection device and equipment for hardware defects in power transmission line |
| CN115775231A (en) * | 2022-11-22 | 2023-03-10 | 国网山西省电力公司大同供电公司 | Cascade R-CNN-based hardware defect detection method and system |
| CN115908999A (en) * | 2022-11-25 | 2023-04-04 | 合肥中科类脑智能技术有限公司 | Method for detecting corrosion of top hardware fitting of power distribution tower, medium and edge terminal equipment |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20210073692A1 (en) * | 2016-06-12 | 2021-03-11 | Green Grid Inc. | Method and system for utility infrastructure condition monitoring, detection and response |
| US11615523B2 (en) * | 2021-08-18 | 2023-03-28 | Zhejiang Gongshang University | Methods for recognizing small targets based on deep learning networks |
-
2023
- 2023-06-01 CN CN202310636354.4A patent/CN116385952B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107493432A (en) * | 2017-08-31 | 2017-12-19 | 广东欧珀移动通信有限公司 | Image processing method, device, mobile terminal and computer-readable recording medium |
| CN109165644A (en) * | 2018-07-13 | 2019-01-08 | 北京市商汤科技开发有限公司 | Object detection method and device, electronic equipment, storage medium, program product |
| CN110210474A (en) * | 2019-04-30 | 2019-09-06 | 北京市商汤科技开发有限公司 | Object detection method and device, equipment and storage medium |
| CN111339891A (en) * | 2020-02-20 | 2020-06-26 | 苏州浪潮智能科技有限公司 | Target detection method of image data and related device |
| 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 |
| WO2022036953A1 (en) * | 2020-08-19 | 2022-02-24 | 上海商汤智能科技有限公司 | Defect detection method and related apparatus, device, storage medium, and computer program product |
| WO2022082856A1 (en) * | 2020-10-19 | 2022-04-28 | 广东科凯达智能机器人有限公司 | Method and system for automatically identifying and tracking inspection target, and robot |
| CN112801230A (en) * | 2021-04-07 | 2021-05-14 | 国网江西省电力有限公司电力科学研究院 | Intelligent acceptance method for unmanned aerial vehicle of power distribution line |
| CN113536981A (en) * | 2021-06-28 | 2021-10-22 | 华雁智能科技(集团)股份有限公司 | Transformer substation foreign matter identification detection method, device and system |
| CN113642486A (en) * | 2021-08-18 | 2021-11-12 | 国网江苏省电力有限公司泰州供电分公司 | Unmanned aerial vehicle distribution network inspection method with airborne front-end identification model |
| CN114170144A (en) * | 2021-11-11 | 2022-03-11 | 国网福建省电力有限公司漳州供电公司 | Power transmission line pin defect detection method, equipment and medium |
| CN115187940A (en) * | 2022-06-09 | 2022-10-14 | 中汽创智科技有限公司 | Image detection method and device for vehicle |
| CN115239642A (en) * | 2022-07-01 | 2022-10-25 | 华雁智能科技(集团)股份有限公司 | Detection method, detection device and equipment for hardware defects in power transmission line |
| CN115775231A (en) * | 2022-11-22 | 2023-03-10 | 国网山西省电力公司大同供电公司 | Cascade R-CNN-based hardware defect detection method and system |
| CN115908999A (en) * | 2022-11-25 | 2023-04-04 | 合肥中科类脑智能技术有限公司 | Method for detecting corrosion of top hardware fitting of power distribution tower, medium and edge terminal equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于CornerNet-Lite的输电塔与绝缘子目标识别与检测;张庆庆;朱仲杰;高明;葛志峰;白永强;屠仁伟;;《浙江万里学院学报》(第03期);91-96 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116385952A (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108960211B (en) | Multi-target human body posture detection method and system | |
| CN109035304B (en) | Target tracking method, medium, computing device and apparatus | |
| CN113902897B (en) | Training of target detection model, target detection method, device, equipment and medium | |
| CN113139543B (en) | Training method of target object detection model, target object detection method and equipment | |
| CN112364843A (en) | Plug-in aerial image target positioning detection method, system and equipment | |
| CN113299073A (en) | Method, device, equipment and storage medium for identifying illegal parking of vehicle | |
| CN116559177A (en) | Defect detection method, device, equipment and storage medium | |
| CN115471476A (en) | Method, device, equipment and medium for detecting component defects | |
| CN114049488A (en) | Multi-dimensional information fusion remote weak and small target detection method and terminal | |
| CN114445825A (en) | Character detection method and device, electronic equipment and storage medium | |
| CN113643260A (en) | Method, apparatus, device, medium and product for detecting image quality | |
| CN113255555A (en) | Method, system, processing equipment and storage medium for identifying Chinese traffic sign board | |
| CN113505643A (en) | Violation target detection method and related device | |
| CN106778822B (en) | Image straight line detection method based on funnel transformation | |
| CN116385952B (en) | Distribution network line small target defect detection method, device, equipment and storage medium | |
| CN114429631B (en) | Three-dimensional object detection method, device, equipment and storage medium | |
| CN110689556A (en) | Tracking method and device and intelligent equipment | |
| CN116188587A (en) | Positioning method and device and vehicle | |
| CN115035129A (en) | Goods identification method and device, electronic equipment and storage medium | |
| CN114581890B (en) | Method and device for determining lane line, electronic equipment and storage medium | |
| CN114565889B (en) | Method and device for determining vehicle line pressing state, electronic equipment and medium | |
| CN113111852B (en) | Target detection method, training method, electronic equipment and gun and ball linkage system | |
| CN113570607B (en) | Target segmentation method and device and electronic equipment | |
| CN114648736B (en) | Robust engineering vehicle identification method and system based on target detection | |
| CN117392631B (en) | Road boundary extraction method and device, electronic equipment and storage medium |
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 |