CN115908999B - Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment - Google Patents
Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment Download PDFInfo
- Publication number
- CN115908999B CN115908999B CN202211491468.6A CN202211491468A CN115908999B CN 115908999 B CN115908999 B CN 115908999B CN 202211491468 A CN202211491468 A CN 202211491468A CN 115908999 B CN115908999 B CN 115908999B
- Authority
- CN
- China
- Prior art keywords
- convolution kernel
- convolution
- branch
- characteristic data
- fusion
- 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
Classifications
-
- 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
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting rust of a power distribution pole tower top hardware fitting, a medium and edge terminal equipment, and relates to the technical field of image processing. The method comprises the following steps: extracting a hardware fitting area image from the top image of the distribution rod tower; extracting features of the hardware region image through the hardware rust detection model to obtain feature data to be processed of the hardware region image; the hardware rust detection model comprises a neck network structure, a first standard convolution kernel and a first depth separable convolution kernel, wherein the neck network structure of the hardware rust detection model is built by adopting a first general convolution kernel and a first fusion convolution kernel; and fusing the characteristic data through the first standard convolution kernel and the first depth separable convolution kernel to obtain a fusion result, and generating a hardware rust detection result. According to the method, the robustness and the precision of the model are further improved by improving the convolution block of the neck structure in the traditional network model, so that the detection speed of the rust of the hardware fitting at the top of the power distribution rod tower is greatly improved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method for detecting rust of a power distribution pole tower top fitting, a medium and edge terminal equipment.
Background
The distribution line is an important component of the national power grid, and metal accessories (made of iron, aluminum or aluminum alloy) widely used in the line are generally called as electric power fittings, wherein equipment in the distribution device is connected with conductors, conductors and wires, the wires of the transmission line and insulators in series, and the metal accessories (made of iron, aluminum or aluminum alloy) used for protecting the wires and insulators are commonly called as electric power fittings. The power fittings used for wire-to-wire connection, insulator-to-insulator connection, insulator-to-tower connection, and insulator-to-wire connection at the top of the distribution line are called line fittings. Because distribution lines in China are widely distributed, the geographical environment is quite complex, distribution cables and towers are exposed in the field for a long time, and the conditions of abundant rainwater and moist weather exist in many areas, oxygen can gradually oxidize metal parts under the condition of water, so that various hardware fittings are rusted and corroded to different degrees. The rust can not only shorten the life of gold utensil, more probably arouse gold utensil not hard up droing, if can't in time investigation and change, probably arouse the short circuit even, destroy the power supply line, and then cause serious accident such as forest fire. In the related art, whether the distribution line works normally or not is judged by manually and regularly checking the distribution line, the method is low in efficiency and has a certain danger, and hidden dangers cannot be checked in time when the hardware fittings at the top of the distribution rod are corroded.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems in the related art to some extent.
Therefore, a first object of the present invention is to provide a method for detecting rust of top hardware of a power distribution pole tower, which further improves the robustness and precision of the model by improving the convolution block of the neck structure in the traditional network model, so that the detection speed of rust of the top hardware of the power distribution pole tower is greatly improved.
A second object of the present invention is to propose a computer readable storage medium.
A third object of the present invention is to propose an edge termination device.
To achieve the above objective, an embodiment of a first aspect of the present invention provides a method for detecting rust on a top hardware of a power distribution pole tower, the method comprising: extracting a hardware fitting area image from the top image of the distribution rod tower; extracting features of the hardware fitting area image through a hardware fitting corrosion detection model to obtain feature data to be processed of the hardware fitting area image; the hardware rust detection model comprises a hardware rust detection model, a hardware rust detection model and a hardware rust detection model, wherein the neck network structure of the hardware rust detection model is built by adopting a first general convolution kernel and a first fusion convolution kernel, and the first fusion convolution kernel is built by adopting a first standard convolution kernel and a first depth separable convolution kernel; the feature data to be processed is subjected to standard convolution processing through the first standard convolution check to obtain first feature data; performing feature fusion processing on the first feature data through the first depth separable convolution check to obtain second feature data; and generating a hardware rust detection result based on the fusion result of the first characteristic data and the second characteristic data.
According to one embodiment of the invention, the extracting of the hardware area image from the power distribution pole tower top image comprises inputting the power distribution pole tower top image into a tower top detection model; the neck network structure of the tower top detection model comprises a second general convolution kernel and a second fusion convolution kernel, wherein the second fusion convolution kernel comprises a second standard convolution kernel and a second depth separable convolution kernel; performing standard convolution processing on the top image of the distribution pole tower through the second standard convolution check to obtain third characteristic data; performing feature fusion processing on the third feature data through the second depth separable convolution check to obtain fourth feature data; and outputting the hardware fitting area image by the tower top detection model based on the third characteristic data and the fourth characteristic data.
According to one embodiment of the invention, the hardware rust detection model further comprises a backbone network structure, wherein the neck network structure is connected to the backbone network structure; the neck network structure includes a plurality of network branches; the network branch is constructed by adopting the first general convolution kernel and the first fusion convolution kernel; the feature extraction is performed on the hardware fitting area image through the hardware fitting corrosion detection model to obtain feature data to be processed of the hardware fitting area image, and the feature data to be processed comprises the following steps: extracting features of the hardware fitting area image through the backbone network structure to obtain backbone feature data; inputting the trunk feature data into the neck network structure, and carrying out standard convolution processing on the trunk feature data through the first general convolution check of the network branches to obtain standard convolution feature data; and the standard convolution characteristic data is used as the characteristic data to be processed.
According to one embodiment of the invention, the network branch comprises a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence; the first branch convolution kernel and the second branch convolution kernel adopt a first fusion convolution kernel; correspondingly, after the first general convolution check passing through the network branches performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method further comprises: sequentially processing the feature data to be processed through a first standard convolution kernel and a first depth separable convolution kernel in the first branch convolution kernel to obtain second feature data; the second characteristic data obtained based on the first branch convolution kernel corresponds to first intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the first branch convolution kernel; and sequentially processing the first intermediate fusion characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in a second branch convolution kernel of the network branch to obtain target fusion convolution characteristic data of the network branch.
According to one embodiment of the invention, the plurality of network branches includes a first network branch; the first network branch comprises a third branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, and the third branch convolution kernel adopts a first fusion convolution kernel; correspondingly, before the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method further comprises: sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the third branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the third branch convolution kernel corresponds to second intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the third branch convolution kernel; correspondingly, the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, including: and carrying out standard convolution processing on the second intermediate fusion characteristic data through a first general convolution check of the first network branch to obtain the standard convolution characteristic data.
According to one embodiment of the invention, the plurality of network branches includes a second network branch; the second network branch comprises a first general convolution kernel, a first branch convolution kernel, a second branch convolution kernel, a fourth branch convolution kernel and a fifth branch convolution kernel which are sequentially connected, wherein the fourth branch convolution kernel adopts the first general convolution kernel, and the fifth branch convolution kernel adopts the first fusion convolution kernel; correspondingly, the main characteristic data is subjected to standard convolution processing in the first general convolution check passing through the network branch to obtain standard convolution characteristic data, and the method further comprises the following steps: the main characteristic data is checked through a first general convolution of the second network branch to carry out standard convolution processing, and the standard convolution characteristic data is obtained; correspondingly, after the first general convolution check passing through the network branches performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method comprises the following steps: sequentially processing the standard convolution characteristic data through a first branch convolution kernel and a second branch convolution kernel in the second network branch to obtain target fusion convolution characteristic data of the second network branch; performing standard convolution processing on target fusion convolution characteristic data of the second network branch through the fourth branch convolution check to obtain first intermediate standard convolution characteristic data; sequentially processing the first intermediate standard convolution characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the fifth branch convolution kernel to obtain second characteristic data; and the second characteristic data obtained based on the fifth branch convolution kernel corresponds to third intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the fifth branch convolution kernel.
According to one embodiment of the invention, the plurality of network branches includes a third network branch; the third network branch comprises a sixth branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, and the sixth branch convolution kernel adopts a first fusion convolution kernel; correspondingly, before the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method further includes: sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the sixth branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the sixth branch convolution kernel corresponds to fourth intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the sixth branch convolution kernel; correspondingly, the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, including: and carrying out standard convolution processing on the fourth intermediate fusion characteristic data through a first general convolution check of the third network branch to obtain the standard convolution characteristic data.
According to one embodiment of the present invention, the plurality of network branches includes a first intermediate convolution kernel disposed between the second network branch and the third network branch, and a second intermediate convolution kernel disposed between the first network branch and the second network branch, where the first intermediate convolution kernel and the second intermediate convolution kernel respectively use a first fusion convolution kernel; the second intermediate convolution kernel is used for processing the output of the first network branch to obtain fifth intermediate fusion characteristic data, and the fifth intermediate fusion characteristic data and the output of the second network branch are subjected to fusion processing to obtain sixth intermediate fusion characteristic data input to the first intermediate convolution kernel; correspondingly, before the first general convolution check passing through the third network branch performs standard convolution processing on the fourth intermediate fusion feature data to obtain the standard convolution feature data, the method further includes: sequentially processing the sixth intermediate fusion characteristic data through a first general convolution kernel and a first depth separable convolution kernel in the first intermediate convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the first intermediate convolution kernel corresponds to seventh intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the first intermediate convolution kernel; performing fusion processing on the seventh intermediate fusion characteristic data and the fourth intermediate fusion characteristic data to obtain eighth intermediate fusion characteristic data; the standard convolution processing is performed on the fourth intermediate fusion characteristic data and the seventh intermediate fusion characteristic data through the first general convolution check of the third network branch, so as to obtain the standard convolution characteristic data, which comprises the following steps: and carrying out standard convolution processing on the eighth intermediate fusion characteristic data through a first general convolution check of the third network branch to obtain the standard convolution characteristic data.
According to an embodiment of the present invention, based on a fusion result of the first feature data and the second feature data, the method includes: multiplying the first characteristic data by the second characteristic data, and performing information fusion on the multiplied result of the first characteristic data and the second characteristic data through a shuffle function to obtain the fusion result.
According to one embodiment of the invention, the tower top detection model and the hardware rust detection model adopt EIoU Loss as candidate frame regression Loss functions.
According to one embodiment of the invention, the tower top detection model and the hardware rust detection model adopt an exponential sliding average method for parameter weight correction.
To achieve the above embodiments, a second aspect of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a method for detecting rust in a power distribution pole tower top fitting according to the above embodiments of the present invention.
In order to achieve the above-mentioned embodiments, a third aspect of the present invention provides an edge terminal device, which is characterized by comprising a memory, a processor, and a power distribution tower top hardware rust detection program stored in the memory and capable of running on the processor, wherein the processor implements the power distribution tower top hardware rust detection method according to the above-mentioned embodiments of the present invention when executing the power distribution tower top hardware rust detection program.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
Fig. 1 is a schematic flow chart of a method for detecting rust on top of a power distribution pole tower according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the structure of a first fused convolution kernel of one embodiment of the present disclosure;
fig. 3 is a schematic flow chart of a method for detecting rust in a top hardware of a power distribution tower according to a first embodiment of the present invention;
fig. 4 is a schematic flow chart of a method for detecting rust on top of a power distribution tower according to a second embodiment of the present invention;
FIG. 5 is a schematic diagram of the architecture of a network branch according to one embodiment of the invention;
FIG. 6 is a schematic diagram of the structure of a first network branch according to one embodiment of the invention;
FIG. 7 is a schematic diagram of the architecture of a second network branch according to one embodiment of the invention;
FIG. 8 is a schematic diagram of the architecture of a second network branch according to one embodiment of the invention;
fig. 9 is a schematic structural diagram of a neck network of a hardware rust detection model according to an embodiment of the present invention;
fig. 10 is a schematic structural view of an edge terminal device according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative and intended to explain the present invention and should not be construed as limiting the invention.
The method, medium and edge terminal equipment for detecting rust in the top hardware of a power distribution pole tower according to the embodiments of the present invention are described below with reference to fig. 1 to 10.
Fig. 1 is a schematic flow chart of a method for detecting rust on top of a power distribution pole tower according to an embodiment of the present invention.
As shown in fig. 1, the method for detecting rust on the top hardware of the power distribution pole tower may include the following steps:
s110, extracting a hardware fitting area image from the top image of the power distribution pole tower.
It can be understood that the distribution line is an important component of the national power grid, the top of the distribution line is used for connecting wires, connecting insulators with towers and connecting insulators with wires, and connecting equipment in the distribution device with conductors, conductors with wires, wires of the transmission line and insulators to form a string, and when the wires, insulators and the like are in corrosion, the service life of the electric power fitting can be shortened, and hidden trouble investigation is needed in time. In this embodiment, the position of the hardware area is extracted from the overhead of the distribution pole by acquiring an image of the top of the distribution pole.
As an example, after the power distribution pole tower top image is acquired, the part of the hardware fitting area in the power distribution pole tower top image can be manually cut out, the part of the hardware fitting area in the power distribution pole tower top image can be detected through a network detection model, and the detected part of the hardware fitting area image is extracted from the power distribution pole tower top image.
By way of example, the top image of the distribution pole can be obtained by shooting the top of the distribution pole through an unmanned aerial vehicle, and the top image of the distribution pole can also be obtained through other high-altitude shooting equipment, so that the invention is not particularly limited.
S120, extracting features of the hardware region image through a hardware rust detection model to obtain feature data to be processed of the hardware region image; the hardware rust detection model comprises a hardware rust detection model, a neck network structure, a first general convolution kernel and a first fusion convolution kernel, wherein the neck network structure of the hardware rust detection model is built by the first general convolution kernel and the first fusion convolution kernel, and the first fusion convolution kernel is built by the first standard convolution kernel and the first depth separable convolution kernel.
Specifically, an image of a hardware fitting area is input into a hardware fitting corrosion detection model to be detected, to obtain feature data to be processed of the hardware fitting area, after the feature data to be processed are obtained, the feature data to be processed are processed through a neck network of the hardware fitting corrosion detection model, wherein a neck network structure of the hardware fitting corrosion detection model is built by adopting a first general convolution kernel and a first fusion convolution kernel, namely the convolution kernel of the neck network comprises a first general convolution kernel and a first fusion convolution kernel, and the first fusion convolution kernel is built by adopting a first standard convolution kernel and a first depth separable convolution kernel.
And S130, carrying out standard convolution processing on the feature data to be processed through a first standard convolution kernel to obtain first feature data.
And S140, performing feature fusion processing on the first feature data through the first depth separable convolution check to obtain second feature data.
It can be understood that the first standard convolution kernel is used for carrying out standard convolution on the feature data, the first fusion convolution kernel is used for carrying out fusion processing on the feature data, the first fusion convolution kernel firstly carries out processing on the feature data through the first standard convolution kernel therein to obtain first feature data, and further carries out processing on the first feature data through the first depth separable convolution kernel to obtain second feature data.
S150, generating a hardware rust detection result based on the fusion result of the first characteristic data and the second characteristic data.
Further, the first characteristic data and the second characteristic data are fused to obtain a fusion result, and the fusion result is used as characteristic data for generating a hardware rust detection model in the hardware rust detection model.
As a possible implementation manner, a structural schematic diagram of the first fusion convolution kernel is shown in fig. 2, the first fusion convolution kernel GSConv is constructed by adopting a first standard convolution kernel Conv and a first depth separable convolution kernel DSC, standard convolution processing is performed on feature data through the first standard convolution kernel Conv, channel-by-channel convolution processing is performed on the feature image through the first depth separable convolution kernel DSC, namely, first feature data is generated through the first standard convolution kernel Conv, second feature data is generated through the first depth separable convolution kernel DSC, and the first feature data and the second feature data are multiplied to obtain a fusion result finally.
As an example, before a hardware rust detection model is obtained, a data set is manufactured for hardware rust target detection at the top of a power distribution pole, a Labelme tool is used for marking an image, wherein the marked area in the image is the position area of hardware rust in the pole top, a label is generated after the image is normalized, and the format of the label is the center point and the width and the height of a category and a rectangle. The data normalization mode adopts a mode of dividing the coordinates of the marking frame by the image width and height values, and the data enhancement mode can adopt modes of horizontal overturn, left-right overturn, tone change, rotation enhancement and the like. The Yolov5 model with the improved neck convolution structure is trained by adopting the distribution rod top hardware rust target detection data set, and a hardware rust detection model is obtained.
Alternatively, the distribution pole top object detection dataset may be made by an unmanned aerial vehicle capturing a preset number of distribution pole top hardware rust images.
Fig. 3 is a flow chart of a method for detecting rust in a top hardware of a power distribution tower according to a first embodiment of the present invention.
As shown in fig. 3, the extracting the image of the hardware area from the image of the top of the power distribution pole tower may include the following steps:
s310, inputting the power distribution pole tower top image into a tower pole top detection model; the neck network structure of the tower top detection model comprises a second general convolution kernel and a second fusion convolution kernel, wherein the second fusion convolution kernel comprises a second standard convolution kernel and a second depth separable convolution kernel.
Specifically, after a distribution pole overhead image is acquired, firstly inputting the distribution pole overhead image into a pole overhead detection model, wherein the model is used for detecting the position of a hardware fitting in the distribution pole overhead image, so as to acquire an area image in which the hardware fitting in the distribution pole overhead image is positioned; and inputting the hardware rust detection model into the hardware rust detection model, wherein the hardware rust detection model is used for detecting the position of hardware rust in the hardware region image.
Illustratively, the tower top detection model and the hardware rust detection model have the same structure, i.e., the second general convolution kernel has the same structure as the first general convolution kernel, and the second fusion convolution kernel has the same structure as the first fusion convolution kernel.
As an example, before obtaining a tower top detection model, firstly, making a data set for power distribution pole top target detection, and marking an image by using a Labelme tool, wherein the marked area in the image is a position area of the power distribution pole top, and generating a label after normalizing the image, wherein the label format is a category, a rectangular center point and a width and height. The data normalization mode adopts a mode of dividing the coordinates of the marking frame by the image width and height values, and the data enhancement mode can adopt modes of horizontal overturn, left-right overturn, tone change, rotation enhancement and the like. The improved model of the neck convolution structure is trained by adopting the distribution rod top target detection data set, and a tower top detection model is obtained.
It can be understood that when the hardware rust detection in the power distribution pole top image is performed, the power distribution pole top image is shot by an unmanned plane or other equipment for acquiring the image, and the position of the hardware rust in the image is not obvious.
And S320, carrying out standard convolution processing on the top image of the distribution pole tower through a second standard convolution check to obtain third characteristic data.
S330, performing feature fusion processing on the third feature data through the second depth separable convolution check to obtain fourth feature data.
S240, outputting a hardware fitting area image by the tower top detection model based on the third characteristic data and the fourth characteristic data.
Optionally, the hardware area image is output through the tower top detection model and is used as input for feature extraction of the hardware corrosion detection model.
In some embodiments, the hardware rust detection model further comprises a backbone network structure, the neck network structure being connected to the backbone network structure; the neck network structure includes a plurality of network branches; the network branch is constructed by adopting a first general convolution kernel and a first fusion convolution kernel; feature extraction is performed on the hardware region image through the hardware rust detection model to obtain feature data to be processed of the hardware region image, and the method can comprise the following steps:
And S310, extracting features of the hardware region image through a backbone network structure to obtain backbone feature data.
S320, inputting the trunk feature data into a neck network structure, and carrying out standard convolution processing on the trunk feature data through a first general convolution check of network branches to obtain standard convolution feature data; the standard convolution characteristic data is used as the characteristic data to be processed.
Specifically, feature extraction is performed on the hardware region image through a main network of the hardware corrosion detection model, so that main feature data are obtained, the main feature data are input into a neck network of the hardware corrosion detection model, standard convolution processing is performed on the main feature network through a first general convolution check, and the main feature data are used as feature data to be processed of the neck network of the hardware corrosion detection model.
Fig. 4 is a flow chart of a method for detecting rust in a top hardware of a power distribution tower according to a second embodiment of the present invention.
It should be noted that, the network branch includes a first general convolution kernel, a first branch convolution kernel, and a second branch convolution kernel, which are sequentially connected; the first branch convolution kernel and the second branch convolution kernel adopt a first fusion convolution kernel.
Correspondingly, after the main characteristic data is subjected to standard convolution processing through the first general convolution check of the network branches to obtain the standard convolution characteristic data, the method further comprises the following steps:
S410, sequentially processing the feature data to be processed through a first standard convolution kernel and a first depth separable convolution kernel in the first branch convolution kernel to obtain second feature data; the second characteristic data obtained based on the first branch convolution kernel corresponds to first intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the first branch convolution kernel.
It can be understood that the first branch convolution kernel adopts a structure of a first fusion convolution kernel, that is, the first branch convolution kernel comprises a first standard convolution kernel and a first depth separable convolution, and the feature data to be processed is processed through the first standard convolution kernel and the first depth separable convolution to obtain first intermediate fusion feature data.
S420, sequentially processing the first intermediate fusion characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in a second branch convolution kernel of the network branch to obtain target fusion convolution characteristic data of the network branch.
The second branch convolution kernel adopts a structure of a first fusion convolution kernel and is connected with the first branch convolution kernel, so that the first intermediate fusion characteristic data is processed through a first standard convolution kernel and a first depth separable convolution in the second branch convolution kernel, and target fusion convolution characteristic data is obtained.
Fig. 5 is a schematic structural diagram of a network branch according to an embodiment of the present invention, where, as shown in fig. 5, the structure of the network branch may include a first general convolution kernel Conv1, a first branch convolution kernel GSConv1, and a second branch convolution kernel GSConv2, where, as shown in fig. 2, specific structures of the first branch convolution kernel GSConv1 and the second branch convolution kernel GSConv2 sequentially obtain, through the first general convolution kernel Conv1, the first branch convolution kernel GSConv1, and the second branch convolution kernel GSConv2, feature images and feature data to be processed obtained by the first general convolution kernel Conv1 correspond to target fusion convolution feature data.
In some embodiments, the plurality of network branches includes a first network branch; the first network branch comprises a third branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, wherein the third branch convolution kernel adopts a first fusion convolution kernel; correspondingly, before the standard convolution processing is performed on the trunk feature data through the first general convolution check of the network branch to obtain the standard convolution feature data, the method further comprises: sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the third branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the third branch convolution kernel corresponds to the second intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the third branch convolution kernel.
Correspondingly, the main characteristic data is subjected to standard convolution processing through the first general convolution check of the network branches to obtain standard convolution characteristic data, which comprises the following steps: and carrying out standard convolution processing on the second intermediate fusion characteristic data through a first general convolution check of the first network branch to obtain standard convolution characteristic data.
Fig. 6 is a schematic structural diagram of a first network branch according to an embodiment of the present invention, where the first network branch includes, as shown in fig. 6, a third branch convolution kernel GSConv3, a first general convolution kernel Conv1, a first branch convolution kernel GSConv1, and a second branch convolution kernel GSConv2, and the third branch convolution kernel GSConv3 of the first network branch is used to process backbone feature data output by a hardware rust detection model backbone network to obtain second intermediate fusion feature data, and the first general convolution kernel Conv1 of the first network branch is used to process the second intermediate fusion feature data to obtain standard convolution feature data, that is, feature data to be processed.
In some embodiments, the plurality of network branches includes a second network branch; the second network branch comprises a first general convolution kernel, a first branch convolution kernel, a second branch convolution kernel, a fourth branch convolution kernel and a fifth branch convolution kernel which are sequentially connected, wherein the first general convolution kernel is adopted by the fourth branch convolution kernel, and the first fusion convolution kernel is adopted by the fifth branch convolution kernel;
Correspondingly, the main characteristic data is subjected to standard convolution processing in the first general convolution check through the network branch to obtain standard convolution characteristic data, and the method further comprises the following steps: and carrying out standard convolution processing on the trunk characteristic data through the first general convolution check of the second network branch to obtain standard convolution characteristic data.
Correspondingly, after the standard convolution processing is performed on the trunk feature data through the first general convolution check of the network branch to obtain the standard convolution feature data, the method further comprises the following steps: sequentially processing the standard convolution characteristic data through a first branch convolution kernel and a second branch convolution kernel in the second network branch to obtain target fusion convolution characteristic data of the second network branch; performing standard convolution processing on target fusion convolution characteristic data of the second network branch through a fourth branch convolution check to obtain first intermediate standard convolution characteristic data; sequentially processing the first intermediate standard convolution characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the fifth branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the fifth branch convolution kernel corresponds to the third intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the fifth branch convolution kernel.
Fig. 7 is a schematic structural diagram of a second network branch according to an embodiment of the present invention, as shown in fig. 7, where the second network branch includes a first general convolution kernel Conv1, a first branch convolution kernel GSConv1, a second branch convolution kernel GSConv2, a fourth branch convolution kernel Conv4, and a fifth branch convolution kernel GSConv5, and standard convolution processing is performed on backbone feature data through the first general convolution kernel Conv1, so as to obtain standard convolution feature data, that is, feature data to be processed. Further, after the standard convolution characteristic data of the second network branch is processed through the first branch convolution kernel GSConv1 and the second branch convolution kernel GSConv2, target fusion convolution characteristic data of the second branch network are obtained, and third intermediate fusion characteristic data are output through the fourth branch convolution kernel GSConv4 and the fifth branch convolution kernel GSConv5 of the second network branch.
In some embodiments, the plurality of network branches further includes a third network branch; the third network branch comprises a sixth branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, wherein the sixth branch convolution kernel adopts a first fusion convolution kernel;
correspondingly, before the main characteristic data is subjected to standard convolution processing through the first general convolution check of the network branches to obtain the standard convolution characteristic data, the method further comprises the steps of: sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in a sixth branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the sixth branch convolution kernel corresponds to fourth intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the sixth branch convolution kernel;
Correspondingly, the main characteristic data is subjected to standard convolution processing through the first general convolution check of the network branches to obtain standard convolution characteristic data, which comprises the following steps: and carrying out standard convolution processing on the fourth intermediate fusion characteristic data through the first general convolution check of the third network branch to obtain standard convolution characteristic data.
Illustratively, fig. 8 is a schematic structural diagram of a third network branch according to an embodiment of the present invention, and as shown in fig. 8, the third network branch includes a sixth branch convolution kernel GSConv6, a first general convolution kernel Conv1, a first branch convolution kernel GSConv1, and a second branch convolution kernel GSConv2, which are sequentially connected. And processing the main characteristic data output by the main network in the hardware rust detection model through a sixth branch convolution kernel GSConv6 of the third network branch to obtain fourth intermediate fusion characteristic data, and processing the fourth intermediate fusion characteristic data through a first general convolution kernel Conv1 of the third network branch to obtain standard convolution characteristic data, namely the characteristic data to be processed.
In some embodiments, the plurality of network branches includes a first intermediate convolution kernel disposed between the second network branch and the third network branch, a second intermediate convolution kernel disposed between the first network branch and the second network branch, the first intermediate convolution kernel and the second intermediate convolution kernel each employing a first fused convolution kernel; the second intermediate convolution kernel is used for processing the output of the first network branch to obtain fifth intermediate fusion characteristic data, and the fifth intermediate fusion characteristic data and the output of the second network branch are subjected to fusion processing to obtain sixth intermediate fusion characteristic data input into the first intermediate convolution kernel.
Correspondingly, before the standard convolution processing is performed on the fourth intermediate fusion characteristic data through the first general convolution check of the third network branch to obtain the standard convolution characteristic data, the method further comprises: sequentially processing sixth intermediate fusion characteristic data through a first general convolution kernel and a first depth separable convolution kernel in the first intermediate convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the first intermediate convolution kernel corresponds to the seventh intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the first intermediate convolution kernel; carrying out fusion processing on the seventh intermediate fusion characteristic data and the fourth intermediate fusion characteristic data to obtain eighth intermediate fusion characteristic data; the fourth intermediate fusion characteristic data and the seventh intermediate fusion characteristic data are checked through the first general convolution of the third network branch to be subjected to standard convolution processing, so that standard convolution characteristic data are obtained, and the method comprises the following steps: and carrying out standard convolution processing on the eighth intermediate fusion characteristic data through the first general convolution check of the third network branch to obtain standard convolution characteristic data.
It can be understood that the first network branch and the second network branch are connected through the first middle convolution kernel, and the second network branch is connected through the second middle convolution kernel, so that the bottom-up characteristic fusion and the top-down characteristic fusion are performed, three output results of the hardware rust detection model neck detection network are obtained, and the three output results are respectively input into the hardware rust detection model prediction layer to perform prediction and regression targets.
As an example, fig. 9 is a schematic structural diagram of a neck network of a hardware rust detection model according to an embodiment of the present invention.
As shown in fig. 9, in this embodiment, after the detected image of the top of the distribution pole or the image of the region of the hardware is preprocessed, the detected image of the top of the distribution pole or the image of the region of the hardware is first input to an input layer of a hardware corrosion detection model, then is subjected to multi-layer feature extraction through a main network, and then is processed through a first branch network, a second branch network and a third branch network, wherein the first branch network and the second branch network are connected through a second middle convolution kernel GSConv8, and the second network branch is connected through a first middle convolution kernel GSConv7 and a third network branch, so that the features of the image are sampled from bottom to top and from top to bottom through the first branch network, the second branch network and the third branch network, the feature data are combined and mixed, and then are transmitted to a prediction layer of the model, that is, the prediction layer of the top of the tower top detection model or the hardware corrosion detection model processes a first target detection head1, a second target detection head2 and a third target detection head3 obtained through the neck network, that is processed to obtain three feature maps with different sizes, and the head portions of the model includes three detection layers, that each target detection head is respectively and three target detection head feature regression feature is set, and each target regression feature is set to obtain different aspect ratio.
It should be noted that, the sizes of the corresponding standard convolution layers and the fusion convolution layers may be set according to the number of the required channels, and the number of the required output channels may be set according to actual needs, which is not particularly limited in the present invention.
As a possible implementation manner, based on the fusion result of the first feature data and the second feature data, the method may further include: and multiplying the first characteristic data by the second characteristic data, and performing information fusion on the multiplied result of the first characteristic data and the second characteristic data through a shuffle function to obtain a fusion result.
It can be understood that the depth separable convolution belongs to sparse convolution, and consists of two parts, namely channel-by-channel convolution and point-by-point convolution, and characteristic values are disassembled according to channels, so that required output is obtained through two-step processing. For the channel-by-channel convolution, one convolution kernel thereof is responsible for one channel, and one channel is only convolved by one convolution kernel, and the number of the convolution kernels is the same as that of the channels of the previous layer. The number of the feature images after the channel-by-channel convolution is the same as the number of channels of the input layer, the feature images cannot be expanded, and the feature images are combined to generate a new feature image through the point-by-point convolution.
It should be noted that, in this embodiment, the structure of the first fusion convolution kernel GSConv1 may be as shown in fig. 2, and in this embodiment, first characteristic data of half the required channels is obtained by first standard convolution Conv, and then second characteristic data of half the required channels is obtained by depth separable convolution DSC. And multiplying the first characteristic data and the second characteristic data to obtain a characteristic diagram of the required output channel number.
Further, information of the first characteristic data and information of the second characteristic data are fused through shuffle fusion, and model generalization capability is improved.
For example, for a 5×5 pixel image, three channels (shape is 5×5×3), a convolution layer with 3×3 convolution kernels (assuming output channel number is 4, convolution kernel shape is 3×3×4, and finally 4 feature maps are output, the convolution layer has a total of 4 filters, each Filter contains 3 convolution kernels, the size of each convolution kernel is 3 x 3, so the number of parameters of the convolution layer is calculated as n_std=4 x 3 x 3=108, the calculated amount of which is c_std=3 x 3 x (5-2) × (5-2) x3 x 4=972, i.e. convolution kernel W (wide) xconvolution kernel H (high) × (picture W-convolution kernel w+1) × (picture H-convolution kernel h+1) xnumber of input channels x number of output channels for channel-by-channel convolution in the depth separable convolution, the number of convolution kernels is identical to the number of channels of the previous layer, so that a three-channel image generates 3 feature maps after operation, wherein one Filter only comprises a convolution kernel with the size of 3 x 3, the number of convolution layers is n_depthwise=3 x 3 x 3=27, the calculated amount is c_depthwise=3 x 3 x (5-2) x 3=243, i.e. c_depthwise = convolution kernel W x convolution kernel H x (picture W-convolution kernel w+1) × (picture H-convolution kernel h+1) ×input channel number. After the channel-by-channel convolution, the feature map of the previous step is weighted and combined in the depth direction through the point-by-point convolution, and a new feature map is generated. The operation of point-by-point convolution is very similar to the conventional convolution operation, and the size of the convolution kernel is 1×1×3,3 being the number of channels of the upper layer. Since a 1 x 1 convolution is adopted, the number of parameters involved in the convolution in this step is n_pointwise=1 x 1 x 3 x 4=12, the calculated amount is c_pointwise=1×1×3×3×3×4=108, c_pointwise=1×1×1×feature layer W x feature layer H x number of input channels x number of output channels (as exemplified above, the number of channels is 4). Thus, for the entire depth separable convolution, the total parameter amount is n_parallel=n_depthwise27+n_poiintwise12=39, and the total parameter amount is c_parallel=c_depthwise243+c_poiintwise108=351. From the above analysis, it can be seen that the depth separable convolution reduces the number of parameters and the amount of computation compared to the conventional standard convolution algorithm. However, conventional dense convolution algorithms can maximally preserve hidden connections between each channel, while sparse volume integration algorithms completely cut off these connections, while feature extraction and fusion capabilities are greatly reduced despite the increased speed. In this embodiment, the standard convolution kernel and the depth separable convolution are combined for use, that is, the standard convolution kernel is first used to obtain a feature map of half the required output channels, then the feature map is used to obtain a feature map of half the required output channels through the first depth separable convolution kernel, the two feature maps are multiplied to obtain a feature map of the required output channels, and then the information obtained by the standard convolution kernel and the information obtained by the depth separable convolution kernel are fused together through the buffer, so that the operation amount of the parameter kernel is greatly reduced, and the features of the image are reserved to a great extent.
In some embodiments, the tower top detection model and the hardware rust detection model employ EIoU Loss as candidate box regression Loss functions.
Specifically, the Loss functions of the tower top detection model and the hardware rust detection model are divided into three parts, including candidate frame regression Loss, foreground background classification Loss and multi-class classification Loss, and in this embodiment, the candidate frame regression Loss function CIoU is replaced by an EIoU Loss function. The expression of the EIoU Loss function is shown as follows:
wherein A and B respectively represent the area of a real boundary box and the area of a prediction boundary box, and rho tableEuclidean distances, b and b, showing the centers of the true and predicted bounding boxes gt Representing the center points, w and w, of the real and prediction bounding boxes, respectively gt Representing the width of the real bounding box and the width of the prediction bounding box, respectively, d represents the Euclidean distance of the diagonal vertices of the closed box, h and h gt High, C representing the true bounding box and the high prediction bounding box, respectively w And C h Representing the area of the real bounding box and the width and height of the smallest bounding rectangle of the prediction bounding box, respectively.
It will be appreciated that CIoU loss function is the most widely used loss function in current anchor-based detectors, but CIoU loss function still suffers from drawbacks. Its two variables (width and height) can only be updated in the same direction, increasing or decreasing at the same time, and the scene adaptation is weak. In this embodiment, by replacing the CIoU loss function with the EIoU loss function, the EIoU loss function directly uses the width and the height of the prediction bounding box as penalty terms, instead of the ratio of the width to the height, so as to further improve the precision of the trained model and improve the accuracy of detecting the rust part of the hardware.
As an example, the tower top detection model and the hardware rust detection model use an exponential sliding average method for parameter weight correction.
The expression of the exponential moving average method is as follows:
v t =β·v t-1 +(1-β)·θ t
wherein, beta is E [0,1 ]]When β=0, it means that the larger the value of β, the more relevant the value of θ obtained by the moving average and the history value of θ, is not used t The value of the variable at time t is indicated.
It will be appreciated that the conventional correction of the parameter weights corresponds to a constant cumulative update of the gradient of the entire training process, and that the parameter weights using the exponential moving average method correspond to a weighted average of the gradient of the training process. Because the training is unstable at the beginning, the obtained gradient gives a smaller weight value more reasonably, and therefore, the exponential moving average method is effective.
According to the method for detecting the rust of the hardware fitting at the top of the power distribution rod tower, provided by the embodiment of the invention, the standard convolution of the part of the neck of the model is replaced by the fusion convolution block, and the index sliding average method which is more in line with the loss function EIoU of the rust scene of the hardware fitting and improves the robustness of the model is applied to the detection model at the top of the tower and the rust detection model of the hardware fitting, so that the memory consumption of the model is reduced, the detection precision is improved, the rust of the hardware fitting at the top of the power distribution rod can be automatically and accurately monitored in real time, and the hidden danger of the rust is timely treated when the rust condition occurs to the hardware fitting at the top of the power distribution rod, and the potential safety hazard is further reduced.
In order to achieve the above embodiments, the present invention further provides a computer readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method for detecting rust on top hardware of a power distribution pole tower according to the above embodiments of the present invention.
Fig. 10 is a schematic structural view of an edge terminal device according to an embodiment of the present invention.
In order to implement the above embodiment, the present invention further provides an edge terminal device 1000, which includes a memory 1001, a processor 1002, and a power distribution pole tower top hardware rust detection program 1003 stored in the memory and capable of running on the processor, where the processor 1002 implements the power distribution pole tower top hardware rust detection method according to the above embodiment of the present invention when executing the power distribution pole tower top hardware rust detection program.
It should be noted that the logic and/or steps represented in the flowcharts or otherwise described herein, for example, may be considered as a ordered listing of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium may even be paper or other suitable medium on which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings are merely for convenience in describing the present invention and simplifying the description, and do not indicate or imply that the device or element being referred to must have a specific orientation, be configured and operated in a specific orientation, and therefore should not be construed as limiting the present invention.
Furthermore, the terms "first," "second," and the like, as used in embodiments of the present invention, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or as implying any particular number of features in the present embodiment. Thus, a feature of an embodiment of the invention that is defined by terms such as "first," "second," etc., may explicitly or implicitly indicate that at least one such feature is included in the embodiment. In the description of the present invention, the word "plurality" means at least two or more, for example, two, three, four, etc., unless explicitly defined otherwise in the embodiments.
In the present invention, unless explicitly stated or limited otherwise in the examples, the terms "mounted," "connected," and "fixed" as used in the examples should be interpreted broadly, e.g., the connection may be a fixed connection, may be a removable connection, or may be integral, and it may be understood that the connection may also be a mechanical connection, an electrical connection, etc.; of course, it may be directly connected, or indirectly connected through an intermediate medium, or may be in communication with each other, or in interaction with each other. The specific meaning of the above terms in the present invention can be understood by those of ordinary skill in the art according to specific embodiments.
In the present invention, unless expressly stated or limited otherwise, a first feature "up" or "down" a second feature may be the first and second features in direct contact, or the first and second features in indirect contact via an intervening medium. Moreover, a first feature being "above," "over" and "on" a second feature may be a first feature being directly above or obliquely above the second feature, or simply indicating that the first feature is level higher than the second feature. The first feature being "under", "below" and "beneath" the second feature may be the first feature being directly under or obliquely below the second feature, or simply indicating that the first feature is less level than the second feature.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.
Claims (11)
1. The utility model provides a distribution pole tower top gold utensil corrosion detection method which characterized in that, the method includes:
extracting a hardware fitting area image from the top image of the distribution rod tower;
extracting features of the hardware fitting area image through a hardware fitting corrosion detection model to obtain feature data to be processed of the hardware fitting area image; the hardware rust detection model comprises a hardware rust detection model, a hardware rust detection model and a hardware rust detection model, wherein the neck network structure of the hardware rust detection model is built by adopting a first general convolution kernel and a first fusion convolution kernel, and the first fusion convolution kernel is built by adopting a first standard convolution kernel and a first depth separable convolution kernel;
the feature data to be processed is subjected to standard convolution processing through the first standard convolution check to obtain first feature data;
performing feature fusion processing on the first feature data through the first depth separable convolution check to obtain second feature data;
Generating a hardware rust detection result based on the fusion result of the first characteristic data and the second characteristic data, wherein the hardware rust detection model further comprises a main network structure, and the neck network structure is connected with the main network structure; the neck network structure includes a plurality of network branches; the plurality of network branches includes a first network branch, a second network branch, and a third network branch; the third network branch comprises a sixth branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, and the sixth branch convolution kernel adopts a first fusion convolution kernel; correspondingly, before the first general convolution check trunk feature data passing through the network branch performs standard convolution processing to obtain standard convolution feature data, the method further includes:
sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the sixth branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the sixth branch convolution kernel corresponds to fourth intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the sixth branch convolution kernel, and the main characteristic data is obtained by extracting characteristics of the hardware fitting area image through the main network structure;
Correspondingly, the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, including:
the fourth intermediate fusion characteristic data is checked through the first general convolution of the third network branch to be subjected to standard convolution processing, and the standard convolution characteristic data is obtained;
the plurality of network branches comprise a first intermediate convolution kernel arranged between the second network branch and the third network branch and a second intermediate convolution kernel arranged between the first network branch and the second network branch, and the first intermediate convolution kernel and the second intermediate convolution kernel respectively adopt a first fusion convolution kernel; the second intermediate convolution kernel is used for processing the output of the first network branch to obtain fifth intermediate fusion characteristic data, and the fifth intermediate fusion characteristic data and the output of the second network branch are subjected to fusion processing to obtain sixth intermediate fusion characteristic data input to the first intermediate convolution kernel; correspondingly, before the first general convolution check passing through the third network branch performs standard convolution processing on the fourth intermediate fusion feature data to obtain the standard convolution feature data, the method further includes:
Sequentially processing the sixth intermediate fusion characteristic data through a first general convolution kernel and a first depth separable convolution kernel in the first intermediate convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the first intermediate convolution kernel corresponds to seventh intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the first intermediate convolution kernel;
performing fusion processing on the seventh intermediate fusion characteristic data and the fourth intermediate fusion characteristic data to obtain eighth intermediate fusion characteristic data;
the standard convolution processing is performed on the fourth intermediate fusion characteristic data and the seventh intermediate fusion characteristic data through the first general convolution check of the third network branch, so as to obtain the standard convolution characteristic data, which comprises the following steps:
and carrying out standard convolution processing on the eighth intermediate fusion characteristic data through a first general convolution check of the third network branch to obtain the standard convolution characteristic data.
2. The method for detecting corrosion of a power distribution pole tower top hardware according to claim 1, wherein the extracting the hardware area image from the power distribution pole tower top image comprises:
Inputting the power distribution pole tower top image into a tower pole top detection model; the neck network structure of the tower top detection model comprises a second general convolution kernel and a second fusion convolution kernel, wherein the second fusion convolution kernel comprises a second standard convolution kernel and a second depth separable convolution kernel;
performing standard convolution processing on the top image of the distribution pole tower through the second standard convolution check to obtain third characteristic data;
performing feature fusion processing on the third feature data through the second depth separable convolution check to obtain fourth feature data;
and outputting the hardware fitting area image by the tower top detection model based on the third characteristic data and the fourth characteristic data.
3. The method for detecting corrosion of a power distribution pole tower top hardware according to claim 1, wherein the network branches are constructed by using the first general convolution kernel and the first fusion convolution kernel; the feature extraction is performed on the hardware fitting area image through the hardware fitting corrosion detection model to obtain feature data to be processed of the hardware fitting area image, and the feature data to be processed comprises the following steps:
extracting features of the hardware fitting area image through the backbone network structure to obtain backbone feature data;
Inputting the trunk feature data into the neck network structure, and carrying out standard convolution processing on the trunk feature data through the first general convolution check of the network branches to obtain standard convolution feature data; and the standard convolution characteristic data is used as the characteristic data to be processed.
4. The method for detecting corrosion of a power distribution pole tower top hardware according to claim 3, wherein the network branch comprises a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence; the first branch convolution kernel and the second branch convolution kernel adopt a first fusion convolution kernel; correspondingly, after the first general convolution check passing through the network branches performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method further comprises:
sequentially processing the feature data to be processed through a first standard convolution kernel and a first depth separable convolution kernel in the first branch convolution kernel to obtain second feature data; the second characteristic data obtained based on the first branch convolution kernel corresponds to first intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the first branch convolution kernel;
And sequentially processing the first intermediate fusion characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in a second branch convolution kernel of the network branch to obtain target fusion convolution characteristic data of the network branch.
5. The method of claim 4, wherein the plurality of network branches comprises a first network branch; the first network branch comprises a third branch convolution kernel, a first general convolution kernel, a first branch convolution kernel and a second branch convolution kernel which are connected in sequence, and the third branch convolution kernel adopts a first fusion convolution kernel; correspondingly, before the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method further comprises:
sequentially processing the trunk characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the third branch convolution kernel to obtain second characteristic data; the second characteristic data obtained based on the third branch convolution kernel corresponds to second intermediate fusion characteristic data with first characteristic data output by a first standard convolution kernel in the third branch convolution kernel;
Correspondingly, the first general convolution check passing through the network branch performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, including:
and carrying out standard convolution processing on the second intermediate fusion characteristic data through a first general convolution check of the first network branch to obtain the standard convolution characteristic data.
6. The method of claim 5, wherein the plurality of network branches includes a second network branch; the second network branch comprises a first general convolution kernel, a first branch convolution kernel, a second branch convolution kernel, a fourth branch convolution kernel and a fifth branch convolution kernel which are sequentially connected, wherein the fourth branch convolution kernel adopts the first general convolution kernel, and the fifth branch convolution kernel adopts the first fusion convolution kernel; correspondingly, the main characteristic data is subjected to standard convolution processing in the first general convolution check passing through the network branch to obtain standard convolution characteristic data, and the method further comprises the following steps:
the main characteristic data is checked through a first general convolution of the second network branch to carry out standard convolution processing, and the standard convolution characteristic data is obtained;
Correspondingly, after the first general convolution check passing through the network branches performs standard convolution processing on the trunk feature data to obtain standard convolution feature data, the method comprises the following steps:
sequentially processing the standard convolution characteristic data through a first branch convolution kernel and a second branch convolution kernel in the second network branch to obtain target fusion convolution characteristic data of the second network branch;
performing standard convolution processing on target fusion convolution characteristic data of the second network branch through the fourth branch convolution check to obtain first intermediate standard convolution characteristic data;
sequentially processing the first intermediate standard convolution characteristic data through a first standard convolution kernel and a first depth separable convolution kernel in the fifth branch convolution kernel to obtain second characteristic data; and the second characteristic data obtained based on the fifth branch convolution kernel corresponds to third intermediate fusion characteristic data with the first characteristic data output by the first standard convolution kernel in the fifth branch convolution kernel.
7. The method for detecting rust on top of a power distribution pole tower according to claim 1, wherein based on the fusion result of the first characteristic data and the second characteristic data, comprising:
Multiplying the first characteristic data by the second characteristic data, and performing information fusion on the multiplied result of the first characteristic data and the second characteristic data through a shuffle function to obtain the fusion result.
8. The method for detecting corrosion of a power distribution pole tower top fitting according to any one of claims 2 to 7, wherein the tower top detection model and the fitting corrosion detection model use EIoU Loss as a candidate frame regression Loss function.
9. The power distribution pole tower top hardware rust detection method according to any one of claims 2-7, wherein the tower top detection model and the hardware rust detection model are subjected to parameter weight correction by an exponential sliding average method.
10. A computer readable storage medium having stored thereon a computer program, which when executed by a processor implements a method of detecting rust in a power distribution pole tower top fitting according to any of claims 1-9.
11. An edge terminal device, comprising a memory, a processor, and a power distribution tower top hardware rust detection program stored in the memory and operable on the processor, wherein the processor implements the power distribution tower top hardware rust detection method according to any one of claims 1-9 when executing the power distribution tower top hardware rust detection program.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211491468.6A CN115908999B (en) | 2022-11-25 | 2022-11-25 | Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211491468.6A CN115908999B (en) | 2022-11-25 | 2022-11-25 | Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115908999A CN115908999A (en) | 2023-04-04 |
CN115908999B true CN115908999B (en) | 2023-07-28 |
Family
ID=86492991
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211491468.6A Active CN115908999B (en) | 2022-11-25 | 2022-11-25 | Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115908999B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116385952B (en) * | 2023-06-01 | 2023-09-01 | 华雁智能科技(集团)股份有限公司 | Distribution network line small target defect detection method, device, equipment and storage medium |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108830211A (en) * | 2018-06-11 | 2018-11-16 | 厦门中控智慧信息技术有限公司 | Face identification method and Related product based on deep learning |
CN110599452B (en) * | 2019-08-07 | 2022-02-22 | 全球能源互联网研究院有限公司 | Rust detection system, method, computer device and readable storage medium |
CN112232116A (en) * | 2020-09-08 | 2021-01-15 | 深圳微步信息股份有限公司 | Facial expression recognition method and device and storage medium |
CN114638969A (en) * | 2022-02-25 | 2022-06-17 | 高新兴科技集团股份有限公司 | Vehicle body multi-attribute detection method, electronic equipment and storage medium |
CN114998757A (en) * | 2022-05-19 | 2022-09-02 | 天津大学 | Target detection method for unmanned aerial vehicle aerial image analysis |
CN115100549A (en) * | 2022-06-24 | 2022-09-23 | 国网山西省电力公司超高压输电分公司 | Transmission line hardware detection method based on improved YOLOv5 |
-
2022
- 2022-11-25 CN CN202211491468.6A patent/CN115908999B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN115908999A (en) | 2023-04-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108489996B (en) | Insulator defect detection method and system and terminal equipment | |
CN111291826B (en) | Pixel-by-pixel classification method of multisource remote sensing image based on correlation fusion network | |
CN110346699B (en) | Insulator discharge information extraction method and device based on ultraviolet image processing technology | |
CN112164086B (en) | Method, system and electronic equipment for determining refined image edge information | |
CN115546630A (en) | Construction site extraction method and system based on remote sensing image characteristic target detection | |
CN110866926B (en) | Infrared remote sensing image rapid and fine sea-land segmentation method | |
CN111860255A (en) | Training and using method, device, equipment and medium of driving detection model | |
CN115908999B (en) | Method for detecting rust of top hardware fitting of distribution pole tower, medium and edge terminal equipment | |
CN112580600A (en) | Dust concentration detection method and device, computer equipment and storage medium | |
CN112101138A (en) | Bridge inhaul cable surface defect real-time identification system and method based on deep learning | |
CN110503049B (en) | Satellite video vehicle number estimation method based on generation countermeasure network | |
CN115760762A (en) | Corrosion detection method, detection device and storage medium | |
CN117409083B (en) | Cable terminal identification method and device based on infrared image and improved YOLOV5 | |
CN111915558A (en) | Pin state detection method for high-voltage transmission line | |
CN115239646A (en) | Defect detection method and device for power transmission line, electronic equipment and storage medium | |
CN117952960A (en) | Electric power iron tower part defect detection method based on artificial intelligence | |
CN113963314A (en) | Rainfall monitoring method and device, computer equipment and storage medium | |
CN112597995A (en) | License plate detection model training method, device, equipment and medium | |
CN110738229B (en) | Fine-grained image classification method and device and electronic equipment | |
CN111914766B (en) | Method for detecting business trip behavior of city management service | |
CN112949641A (en) | Image segmentation method, electronic device and computer-readable storage medium | |
CN113298759A (en) | Water area detection method and device, electronic equipment and storage medium | |
CN116188998B (en) | Method, device, equipment and storage medium for identifying defects of overhead transmission line | |
CN116993743B (en) | Method, device, equipment and storage medium for detecting galloping amplitude of power transmission line | |
CN117173385B (en) | Detection method, device, medium and equipment of transformer substation |
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 |