CN-121981971-A - State detection method, system, equipment and medium for ammeter junction box

CN121981971ACN 121981971 ACN121981971 ACN 121981971ACN-121981971-A

Abstract

The application discloses a method, a system, equipment and a medium for detecting the state of an ammeter junction box, which are characterized by comprising the steps of obtaining image data of the ammeter junction box; inputting image data into a preset state detection model, carrying out feature extraction on the image data through an attention convolution module to obtain a first connecting sheet feature map and a first line number feature map, carrying out feature reconstruction on the first connecting sheet feature map and the first line number feature map through a space channel convolution module to obtain a second connecting sheet feature map and a second line number feature map, carrying out feature map splicing on the second connecting sheet feature map and the second line number feature map to obtain a fusion feature map, determining each connecting sheet detection result and each line number detection result in the ammeter junction box based on the fusion feature map, and detecting the state of the ammeter junction box based on the connecting sheet detection result and the line number detection result. The method and the device can improve the accuracy of identifying the abnormal conditions of the junction box of the ammeter.

Inventors

HOU ENZHEN
Huang anrui
WANG YAN
Mao Xinpeng
ZHANG LEI
CHEN GUOPEI
Dui Zixuan
ZHANG WANYING
NIU JIN
RONG QI

Assignees

中国能源建设集团广东省电力设计研究院有限公司

Dates

Publication Date: 20260505
Application Date: 20260105

Claims (10)

1. A method for detecting the status of an electric meter junction box, wherein the electric meter junction box comprises a plurality of connecting pieces and line numbers, the method comprising: acquiring image data of an ammeter junction box to be detected; Inputting the image data into a preset state detection model, carrying out feature extraction on the image data through an attention convolution module to obtain a first connecting sheet feature map and a first line number feature map, carrying out feature reconstruction on the first connecting sheet feature map and the first line number feature map through a space channel convolution module to obtain a second connecting sheet feature map and a second line number feature map, carrying out feature map splicing on the second connecting sheet feature map and the second line number feature map based on a feature fusion module to obtain a fusion feature map, and processing the fusion feature map through a detection head to output a detection result of each connecting sheet and a detection result of a line number in an ammeter junction box, wherein the state detection model comprises the attention convolution module, the space channel convolution module, the feature fusion module and the detection head; And detecting the state of the ammeter junction box based on the detection result of each connecting piece and the detection result of the line number.
2. The method for detecting the status of an electric meter junction box according to claim 1, further comprising, after said acquiring the image data of the electric meter junction box to be detected: Performing contour extraction processing on the image data to determine the boundary contour of the ammeter junction box in the image data; And correcting the boundary contour based on a perspective transformation algorithm to obtain corrected image data.
3. The method for detecting the state of an ammeter junction box according to claim 1, wherein said inputting said image data into a predetermined state detection model to perform feature extraction on said image data by an attention convolution module, to obtain a first connection tab feature map and a first line number feature map, comprises: inputting the image data into a backbone network of the state detection model to extract local features of the image data through a first convolution branch to obtain first branch features; Extracting semantic features from the image data through a second convolution branch to obtain a second branch feature, wherein the attention convolution module comprises the first convolution branch and the second convolution branch; performing feature fusion on the first branch feature and the second branch feature based on an attention mechanism to obtain an attention-enhancing feature map; And carrying out channel shuffling on the attention enhancement feature map to obtain a channel rearrangement feature map, and determining the first connecting piece feature map and the first line number feature map based on the channel rearrangement feature map.
4. The method for detecting a state of an electric meter junction box according to claim 1, wherein the performing, by a spatial channel convolution module, feature reconstruction on the first connection piece feature map and the first line number feature map to obtain a second connection piece feature map and a second line number feature map includes: Inputting the first connecting piece characteristic diagram and the first line number characteristic diagram to a space reconstruction unit of the space channel convolution module to perform space characteristic separation to obtain a first space separation characteristic branch and a second space separation characteristic branch; Performing space reconstruction on the first space separation characteristic branch and the second space separation characteristic branch to obtain a space reconstruction characteristic diagram; inputting the space reconstruction feature map to a channel reconstruction unit of the space channel convolution module to perform channel feature splitting to obtain a first channel feature branch and a second channel feature branch; and carrying out self-adaptive weighted fusion on the first channel characteristic branch and the second channel characteristic branch based on preset weights, and outputting the second connecting piece characteristic diagram and the second line number characteristic diagram.
5. The method for detecting the state of an ammeter junction box according to claim 1, wherein the feature-based fusion module performs feature-map stitching on the second connection piece feature map and the second line number feature map to obtain a fused feature map, and the method comprises: performing up-sampling treatment on the second connecting piece characteristic diagram to perform scale alignment on the second connecting piece characteristic diagram and the second line number characteristic diagram to obtain a target connecting piece characteristic diagram; and performing feature map splicing on the target connecting sheet feature map and the second line number feature map based on the feature fusion module to obtain a fusion feature map.
6. The method for detecting the state of an electric meter junction box according to claim 1, wherein the processing the fusion feature map by the detecting head to output the detection result of each connecting piece and the detection result of the line number in the electric meter junction box comprises: performing convolution operation on the fusion feature map based on a boundary box regression branch of the detection head to obtain boundary boxes of connecting pieces and line numbers in an ammeter junction box; performing convolution operation on the fusion feature map based on the classification branches of the detection head to obtain class probabilities of each connecting sheet and line number in the ammeter junction box; And performing non-maximum suppression processing on each bounding box and each class probability to obtain detection results of each connecting sheet and line number in the ammeter junction box.
7. The method for detecting the state of an electric meter junction box according to claim 1, wherein said detecting the state of the electric meter junction box based on each of said connection piece detection result and said line number detection result comprises: carrying out statistical analysis on the detection result of each connecting piece and the detection result of each line number to determine the spatial position relationship between each connecting piece and each line number; And carrying out matching analysis on the spatial position relation based on a preset dynamic judgment rule, and outputting a wiring state detection result of the ammeter junction box.
8. A status detection system for an electric meter junction box, the system comprising: The acquisition module is used for acquiring image data of the to-be-detected ammeter junction box; The identification module is used for inputting the image data into a preset state detection model, carrying out feature extraction on the image data through the attention convolution module to obtain a first connecting sheet feature map and a first line number feature map, carrying out feature reconstruction on the first connecting sheet feature map and the first line number feature map through the space channel convolution module to obtain a second connecting sheet feature map and a second line number feature map, carrying out feature map splicing on the second connecting sheet feature map and the second line number feature map based on the feature fusion module to obtain a fusion feature map, and processing the fusion feature map through a detection head to output each connecting sheet detection result and each line number detection result in an ammeter junction box, wherein the state detection model comprises the attention convolution module, the space channel convolution module, the feature fusion module and the detection head; And the detection module is used for detecting the state of the ammeter junction box based on the detection result of each connecting piece and the detection result of the line number.
9. A terminal device comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, the processor implementing a method for detecting the status of an electric meter junction box according to any one of claims 1-7 when the computer program is executed.
10. A computer readable storage medium comprising a stored computer program, wherein the computer program, when run, controls a device in which the computer readable storage medium resides to perform a method of detecting a status of an electric meter junction box according to any one of claims 1-7.

Description

State detection method, system, equipment and medium for ammeter junction box Technical Field The invention relates to the field of ammeter wiring identification, in particular to a method, a system, equipment and a medium for detecting the state of an ammeter junction box. Background In modern electric power systems, an electric energy meter is a core device for realizing electric energy metering and electric charge settlement, and the wiring states of phase lines, voltage lines, current lines and neutral lines in an electric energy meter junction box directly determine the accuracy of metering results and the operation safety of a power grid. If the junction box has the conditions of misconnection, missing connection, virtual connection or phase sequence error, etc., the voltage and current sampling relationship is abnormal, and active electric energy metering deviation is caused, so that the electric charge settlement is not public, the line loss is abnormally increased, and even the stable operation of the power distribution system is affected. The existing ammeter junction box detection technology still has obvious defects in the aspect of detection precision. The traditional manual detection mode highly depends on manual experience, and when the terminal is densely arranged, has similar appearance and has a shielded junction box structure, missed detection and misjudgment are easy to occur, and the consistency of detection results is poor. Although the method based on electrical parameters or phase analysis can realize preliminary screening, the method is limited by single rule dimension, the complicated wiring abnormality type is difficult to distinguish accurately, and the detection result can only stay on the coarse granularity judgment level of correct or error. When the detection method based on deep learning is used for facing small-scale targets, reflective metal parts and multi-specification junction box scenes, the problems of small target omission, positioning deviation and the like are prone to occurring, and the overall detection precision is difficult to meet the actual application requirements. Disclosure of Invention The invention provides a state detection method, a state detection system, state detection equipment and a state detection medium for an ammeter junction box, which can improve the accuracy of identifying the abnormal state of the ammeter junction box. In a first aspect, an embodiment of the present invention provides a method for detecting a state of an ammeter junction box, including: acquiring image data of an ammeter junction box to be detected; Inputting the image data into a preset state detection model, carrying out feature extraction on the image data through an attention convolution module to obtain a first connecting sheet feature map and a first line number feature map, carrying out feature reconstruction on the first connecting sheet feature map and the first line number feature map through a space channel convolution module to obtain a second connecting sheet feature map and a second line number feature map, carrying out feature map splicing on the second connecting sheet feature map and the second line number feature map based on a feature fusion module to obtain a fusion feature map, and processing the fusion feature map through a detection head to output a detection result of each connecting sheet and a detection result of a line number in an ammeter junction box, wherein the state detection model comprises the attention convolution module, the space channel convolution module, the feature fusion module and the detection head; And detecting the state of the ammeter junction box based on the detection result of each connecting piece and the detection result of the line number. The embodiment of the invention provides an original visual information foundation comprising a connecting sheet form, a line number and space distribution information thereof, prevents deviation of a detection result caused by data deficiency, ensures detection reliability from the source, performs characteristic extraction on image data through a attention convolution module, can adaptively enhance the significance of a key target area in an ammeter junction box under complex background and multiple interference factors, thereby improving the distinguishing capability of the key target in a characteristic extraction stage, improving subsequent identification precision, enhancing structural layout information of the connecting sheet and the line number in a space dimension, inhibiting redundant characteristics and enhancing effective semantic expression in a channel dimension, effectively reducing the influence of characteristic aliasing on the identification result, improving the accuracy of target positioning and category discrimination, fully fusing high-level semantic information with fine granularity space information, enabling a model to simultaneously integrate wiring topologica