CN111754737A - On-line identification and evaluation device and evaluation method for installation and acceptance of metering device - Google Patents

On-line identification and evaluation device and evaluation method for installation and acceptance of metering device Download PDF

Info

Publication number
CN111754737A
CN111754737A CN202010690116.8A CN202010690116A CN111754737A CN 111754737 A CN111754737 A CN 111754737A CN 202010690116 A CN202010690116 A CN 202010690116A CN 111754737 A CN111754737 A CN 111754737A
Authority
CN
China
Prior art keywords
module
image
evaluation
scoring
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010690116.8A
Other languages
Chinese (zh)
Inventor
王晖南
韩迎军
刘佳易
索思远
张建民
高强
孙晋凯
武文萍
谭沛然
杨兆忠
赵园
董力群
杨艳芳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Marketing Service Center of State Grid Shanxi Electric Power Co Ltd
Original Assignee
Marketing Service Center of State Grid Shanxi Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Marketing Service Center of State Grid Shanxi Electric Power Co Ltd filed Critical Marketing Service Center of State Grid Shanxi Electric Power Co Ltd
Priority to CN202010690116.8A priority Critical patent/CN111754737A/en
Publication of CN111754737A publication Critical patent/CN111754737A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G08SIGNALLING
    • G08BSIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18Status alarms
    • G08B21/185Electrical failure alarms
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/70Denoising; Smoothing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/40Scenes; Scene-specific elements in video content
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00Cameras or camera modules comprising electronic image sensors; Control thereof

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Business, Economics & Management (AREA)
  • Emergency Management (AREA)
  • Image Analysis (AREA)

Abstract

本发明涉及一种计量装置安装验收在线识别评价装置及评价方法,目的是解决现有计量装置安装验收的评价受外界环境及主观因素影响大、投入人力物力大、作业过程不能实时监控且不能实现自动评价及告警功能的技术问题,本发明的技术方案为:一种计量装置安装验收在线识别评价装置,包括图像采集预处理模块、图像识别分析模块、评价打分模块、处理器、存储模块、定位模块、人机交互模块和报警模块。该评价方法采用前端采集、智能识别和深度学习框架技术对图像进行处理,可以实现图像识别,实时评价打分,智能化地管理计量装置,降低人力物力成本。

Figure 202010690116

The invention relates to an online identification and evaluation device and an evaluation method for the installation and acceptance of a metering device, and aims to solve the problem that the evaluation of the installation and acceptance of the existing metering device is greatly affected by the external environment and subjective factors, the input of manpower and material resources is large, and the operation process cannot be monitored in real time and cannot be realized. The technical problem of automatic evaluation and alarm function, the technical solution of the present invention is: an online identification and evaluation device for installation and acceptance of a metering device, including an image acquisition preprocessing module, an image recognition analysis module, an evaluation scoring module, a processor, a storage module, a positioning module module, human-computer interaction module and alarm module. The evaluation method uses front-end acquisition, intelligent recognition and deep learning framework technology to process images, which can realize image recognition, real-time evaluation and scoring, intelligent management of metering devices, and reduce labor and material costs.

Figure 202010690116

Description

一种计量装置安装验收在线识别评价装置及评价方法On-line identification and evaluation device and evaluation method for installation and acceptance of metering device

技术领域technical field

本发明属于电力计量领域,具体涉及一种计量装置安装验收在线识别评价装置及评价方法。The invention belongs to the field of electric power metering, and in particular relates to an online identification and evaluation device and an evaluation method for the installation and acceptance of a metering device.

背景技术Background technique

随着电网技术复杂程度的不断升高,影响电网安全的因素越来越多,安全生产形势日益严峻。虽然电网、电力企业已制定了完备的电力安全生产规程,但是现阶段依然存在很多因电力安全管理松懈、作业人员安全意识薄弱等原因导致的违章行为频发的问题。现代电力安全生产对安监评价系统提出了作业过程可监控、结果可追踪、智能告警的要求。With the increasing complexity of power grid technology, there are more and more factors affecting power grid security, and the safety production situation is becoming increasingly severe. Although power grids and power companies have formulated complete power safety production regulations, there are still many problems of frequent violations caused by lax power safety management and weak safety awareness of operators. Modern electric power safety production puts forward the requirements of safety monitoring and evaluation system that the operation process can be monitored, the results can be traced and intelligent alarming.

电力行业目前对计量装置安装验收的评价均是通过飞行器进行检查,再经人工目测来评价优劣,需投入大量的人力、物力,且飞行检查易受各种不确定因素的影响,人工目测受主观因素影响较大,无法及时客观全方位地准确掌握设备的安装状态,不能实现自动评价功能。At present, the evaluation of the installation and acceptance of the metering device in the electric power industry is based on the inspection of the aircraft, and then the evaluation of the pros and cons through manual visual inspection, which requires a lot of manpower and material resources. Subjective factors have a great influence, and it is impossible to accurately grasp the installation status of the equipment in a timely, objective and comprehensive manner, and cannot realize the automatic evaluation function.

发明内容SUMMARY OF THE INVENTION

本发明的目的是解决现有计量装置安装验收的评价受外界环境及主观因素影响大、投入人力物力大、作业过程不能实时监控且不能实现自动评价及告警功能的技术问题,提供一种计量装置安装验收在线识别评价装置及评价方法。The purpose of the present invention is to solve the technical problems that the evaluation of the installation and acceptance of the existing metering device is greatly affected by the external environment and subjective factors, the input of manpower and material resources is large, the operation process cannot be monitored in real time, and the automatic evaluation and alarm functions cannot be realized, and a metering device is provided. Installation and acceptance online identification and evaluation device and evaluation method.

为实现上述目的,本发明采用的技术方案是:For achieving the above object, the technical scheme adopted in the present invention is:

一种计量装置安装验收在线识别评价装置,包括图像采集预处理模块、图像识别分析模块、评价打分模块、处理器、存储模块、定位模块、人机交互模块和报警模块;An online identification and evaluation device for installation and acceptance of a metering device, comprising an image acquisition preprocessing module, an image identification analysis module, an evaluation scoring module, a processor, a storage module, a positioning module, a human-computer interaction module and an alarm module;

所述图像采集预处理模块由图像传感器、采集模块和图像降噪处理模块组成,所述图像传感器对现场待评价电能计量装置进行图像和/或视频的拍摄,并将拍摄的视频文件逐帧转换为RGB图像;所述采集模块实时无损地对图像传感器所拍摄的图像和/或视频进行采集;所述图像降噪处理模块对采集模块发送的图片和/或视频中的噪声进行消除;The image acquisition preprocessing module is composed of an image sensor, an acquisition module and an image noise reduction processing module. The image sensor captures images and/or videos of the on-site electric energy metering device to be evaluated, and converts the captured video files frame by frame. It is an RGB image; the acquisition module collects images and/or videos captured by the image sensor in real time without loss; the image noise reduction processing module removes noise in the pictures and/or videos sent by the acquisition module;

所述图像识别分析模块由图像训练模块和图像预测模块组成,所述图像训练模块,接收图像采集预处理模块发送的图像信息,对要识别图像进行标注,然后使用深度学习Gaussian_yolo框架进行训练,获得模型文件并保存;所述图像预测模块,加载训练好的模型文件,对采集的目标图像依次进行训练过程的步骤,多层次下的采样、卷积、池化和全连接层,最后输出图像中每个目标的类别、位置和置信度信息;The image recognition analysis module is composed of an image training module and an image prediction module. The image training module receives the image information sent by the image acquisition preprocessing module, marks the image to be recognized, and then uses the deep learning Gaussian_yolo framework for training to obtain model file and save it; the image prediction module loads the trained model file, sequentially performs the steps of the training process on the collected target image, multi-level sampling, convolution, pooling and full connection layer, and finally outputs the image in the Category, location and confidence information for each target;

所述评价打分模块由评分规则库和量化打分模块组成,所述评分规则库根据产品的状态特征和产品状态特征对产品的影响程度,规定产品评分规则;所述量化打分模块,根据图像识别分析模块输出图像中每个目标的类别,按照评分规则进行评价打分;The evaluation scoring module is composed of a scoring rule base and a quantitative scoring module. The scoring rule base specifies product scoring rules according to the state characteristics of the product and the degree of influence of the product state characteristics on the product; the quantitative scoring module is based on image recognition and analysis. The module outputs the category of each target in the image, and evaluates and scores according to the scoring rules;

所述处理器对所述图像采集预处理模块、图像识别分析模块、评价打分模块和定位模块发送的数据进行处理和存储,并对综合得分低于得分阈值的待评价电能计量装置发出报警指令,通过所述人机交互模块完成操作人员与识别评价装置的交互;The processor processes and stores the data sent by the image acquisition preprocessing module, the image recognition and analysis module, the evaluation scoring module and the positioning module, and issues an alarm instruction to the electric energy metering device to be evaluated whose comprehensive score is lower than the score threshold, The interaction between the operator and the identification evaluation device is completed through the human-computer interaction module;

所述存储模块存储待测电能计量装置的历史综合得分和各类别的得分及目标图像的图像识别分析数据;The storage module stores historical comprehensive scores and scores of various categories and image recognition analysis data of the target image of the electric energy metering device to be tested;

所述定位模块对待测电能计量装置所在的实时位置进行定位;The positioning module locates the real-time position where the electric energy metering device to be measured is located;

所述人机交互模块:通过界面完成操作人员与识别评价装置的交互对话;The human-computer interaction module: completes the interactive dialogue between the operator and the identification and evaluation device through the interface;

所述报警模块接收所述处理器的报警信号并完成报警。The alarm module receives the alarm signal of the processor and completes the alarm.

一种计量装置安装验收在线识别评价装置的评价方法,包括以下步骤:An evaluation method for an online identification and evaluation device for installation and acceptance of a metering device, comprising the following steps:

1)评分规则库的建立:根据产品的状态特征和产品状态特征对产品的影响程度,制定产品评分规则并入库;1) Establishment of scoring rule base: formulate product scoring rules and incorporate them into the library according to the status characteristics of the product and the degree of influence of the product status characteristics on the product;

2)图像的采集和预处理:对现场待评价的电能计量装置进行图像和/或视频的拍摄,并将拍摄的视频文件逐帧转换为RGB图像,实时无损地对所拍摄的图像和/或视频进行采集,然后并对采集的图片和/或视频进行降噪处理;2) Image collection and preprocessing: image and/or video are captured on the electric energy metering device to be evaluated on site, and the captured video files are converted into RGB images frame by frame, and the captured images and/or Capture video, and then perform noise reduction processing on the captured pictures and/or videos;

3)图像的识别和分析:对预处理后的目标图像使用标注工具进行标注,然后使用深度学习Gaussian_yolo框架进行训练,获得模型文件并保存,接着加载训练好的模型文件,对采集的目标图像依次进行训练过程的步骤,多层次下的采样、卷积、池化和全连接层,获得并输出图像中每个目标的类别、位置和置信度信息;3) Image recognition and analysis: Label the preprocessed target image with the labeling tool, and then use the deep learning Gaussian_yolo framework for training, obtain the model file and save it, then load the trained model file, and sequence the collected target images. Steps in the training process, multi-level sampling, convolution, pooling and fully connected layers, to obtain and output the category, location and confidence information of each target in the image;

4)图像的评价打分:根据图像中每个目标的类别、位置和置信度信息以及建立的评分规则,进行打分;4) Image evaluation and scoring: score according to the category, location and confidence information of each target in the image and the established scoring rules;

5)报警:根据打分给出的目标综合得分判别是否报警,当得分低于预设的得分阈值时,发出报警信号并进行报警;5) Alarm: according to the comprehensive score of the target given by the score to determine whether to alarm, when the score is lower than the preset score threshold, an alarm signal is issued and an alarm is performed;

6)数据存储:对目标图像的图像识别分析数据以及历史综合得分数据进行存储。6) Data storage: the image recognition analysis data of the target image and the historical comprehensive score data are stored.

进一步的,所述步骤1)中所述产品评分规则为对表箱工艺检测、表箱安装偏正、表箱表面检测、检测表箱表面二维码提取、电表检测、电表安装偏正、电表铅封、电表入线、安全检测、漏线检测和接线工艺评分的规定。Further, the product scoring rules described in the step 1) are the process detection of the meter box, the installation of the meter box, the surface detection of the meter box, the extraction of the two-dimensional code on the surface of the detection meter box, the detection of the electricity meter, the installation of the electricity meter Provisions for lead sealing, meter entry, safety testing, leak detection and wiring workmanship scoring.

进一步的,所述步骤3)中的标注为对图像中的电表、接线、损坏和露线区域进行特征标注。Further, the labeling in the step 3) is to perform feature labeling on the electric meter, wiring, damage and exposed line areas in the image.

进一步的,所述步骤3)中训练的步骤包括:Further, the step of training in the step 3) includes:

(1)配置训练文件并对图像文件进行分批;(1) Configure training files and batch image files;

(2)根据配置的训练文件对每批次每张图像进行尺寸剪裁,缩放到同一尺寸;(2) Cut the size of each image in each batch according to the configured training file and scale it to the same size;

(3)对每张图像按照训练文件进行多层次下的采样、卷积、池化和全连接层进行特征提取;(3) Perform feature extraction on each image through multi-level sampling, convolution, pooling and fully connected layers according to the training file;

(4)对提取到的特征和标注的特征区域进行回归分析;(4) Perform regression analysis on the extracted features and marked feature regions;

(5)更新每个层次神经元的值,训练N次后,得到模型文件并保存。(5) Update the value of each level of neurons, after training N times, get the model file and save it.

进一步的,所述步骤4)中的评价规则按照百分制进行,评价规则具体为:类别得分=(总的检测目标数﹣异常的目标数)/总的检测目标数×该类别的分值分配×1/2Further, the evaluation rules in the step 4) are carried out according to the percentage system, and the evaluation rules are specifically: category score = (total number of detection targets - number of abnormal targets)/total number of detection targets × score allocation of this category × 1/2

当异常目标数为零时,则该类别为满分,各个类别的类别得分之和为综合得分。When the number of abnormal targets is zero, the category is full score, and the sum of the category scores of each category is the comprehensive score.

进一步的,所述训练步骤1)中的训练文件包括处理图像的尺寸、进行多层次下的采样、卷积、池化和全连接,以及每个处理的参数。Further, the training file in the training step 1) includes the size of the processed image, multi-level sampling, convolution, pooling and full connection, as well as the parameters of each processing.

本发明的有益效果是:The beneficial effects of the present invention are:

1、本发明提供的计量装置安装验收在线识别评价装置及评价方法利用图像识别和AI技术对表箱工艺、电表构件和线路安全等全方位进行前端采集和智能识别,智能化的图像识别算法可以区分不同的类型的电表箱外观,内部结构以及安装工艺,并针对识别结果进行综合评价打分,可最大限度的消除人工主观因素的影响,提升识别效率,方便快捷。1. The metering device installation acceptance online identification evaluation device and evaluation method provided by the present invention use image recognition and AI technology to carry out front-end collection and intelligent identification of meter box technology, meter components and line safety in an all-round way. The intelligent image recognition algorithm can Distinguish the appearance, internal structure and installation process of different types of meter boxes, and comprehensively evaluate and score the identification results, which can eliminate the influence of artificial subjective factors to the greatest extent, and improve the identification efficiency, which is convenient and fast.

2、本发明提供的计量装置安装验收在线识别评价装置具有便于携带、实时监测识别、评价客观等特点,工作人员可以随身携带此设备通过图像识别的方法在现场对安装后的表箱工艺、电表构件和线路安全进行实时采集,实时评价打分,智能化地管理计量装置,降低人力物力成本。2. The online identification and evaluation device for installation and acceptance of the metering device provided by the present invention has the characteristics of being easy to carry, real-time monitoring and identification, objective evaluation, etc. The staff can carry this device with them on the spot through the method of image recognition to check the meter box process and electricity meter after installation. Component and line safety are collected in real time, evaluated and scored in real time, and the metering device is intelligently managed to reduce labor and material costs.

附图说明Description of drawings

图1为本发明计量装置安装验收在线识别评价装置的模块结构示意图;1 is a schematic diagram of the module structure of the online identification and evaluation device for installation and acceptance of a metering device according to the present invention;

图2为本发明评价方法的流程图。Fig. 2 is a flow chart of the evaluation method of the present invention.

具体实施方式Detailed ways

下面结合附图和实施例对本发明作进一步的描述。The present invention will be further described below with reference to the accompanying drawings and embodiments.

如图1所示,本实施例中的一种计量装置安装验收在线识别评价装置,包括图像采集预处理模块、图像识别分析模块、评价打分模块、处理器、存储模块、定位模块、人机交互模块和报警模块;As shown in FIG. 1 , an online identification and evaluation device for installation and acceptance of a metering device in this embodiment includes an image acquisition preprocessing module, an image recognition analysis module, an evaluation scoring module, a processor, a storage module, a positioning module, and a human-computer interaction module. modules and alarm modules;

所述图像采集预处理模块由图像传感器、采集模块和图像降噪处理模块组成,所述图像传感器对现场待评价电能计量装置进行图像和/或视频的拍摄,并将拍摄的视频文件逐帧转换为RGB图像;所述采集模块实时无损地对图像传感器所拍摄的图像和/或视频进行采集;所述图像降噪处理模块对采集模块发送的图片和/或视频中的噪声进行消除;The image acquisition preprocessing module is composed of an image sensor, an acquisition module and an image noise reduction processing module. The image sensor captures images and/or videos of the on-site electric energy metering device to be evaluated, and converts the captured video files frame by frame. It is an RGB image; the acquisition module collects images and/or videos captured by the image sensor in real time without loss; the image noise reduction processing module removes noise in the pictures and/or videos sent by the acquisition module;

所述图像识别分析模块由图像训练模块和图像预测模块组成,所述图像训练模块,接收图像采集预处理模块发送的图像信息,对要识别图像进行标注,然后使用深度学习Gaussian_yolo框架进行训练,获得模型文件并保存;所述图像预测模块,加载训练好的模型文件,对采集的目标图像依次进行训练过程的步骤,多层次下的采样、卷积、池化和全连接层,最后输出图像中每个目标的类别、位置和置信度信息;The image recognition analysis module is composed of an image training module and an image prediction module. The image training module receives the image information sent by the image acquisition preprocessing module, marks the image to be recognized, and then uses the deep learning Gaussian_yolo framework for training to obtain model file and save it; the image prediction module loads the trained model file, sequentially performs the steps of the training process on the collected target image, multi-level sampling, convolution, pooling and full connection layer, and finally outputs the image in the Category, location and confidence information for each target;

所述评价打分模块由评分规则库和量化打分模块组成,所述评分规则库根据产品的状态特征和产品状态特征对产品的影响程度,规定产品评分规则;所述量化打分模块,根据图像识别分析模块输出图像中每个目标的类别,按照评分规则进行评价打分;The evaluation scoring module is composed of a scoring rule base and a quantitative scoring module. The scoring rule base specifies product scoring rules according to the state characteristics of the product and the degree of influence of the product state characteristics on the product; the quantitative scoring module is based on image recognition and analysis. The module outputs the category of each target in the image, and evaluates and scores according to the scoring rules;

所述处理器对所述图像采集预处理模块、图像识别分析模块、评价打分模块和定位模块发送的数据进行处理和存储,并对综合得分低于得分阈值的待评价电能计量装置发出报警指令,通过所述人机交互模块完成操作人员与识别评价装置的交互;The processor processes and stores the data sent by the image acquisition preprocessing module, the image recognition and analysis module, the evaluation scoring module and the positioning module, and issues an alarm instruction to the electric energy metering device to be evaluated whose comprehensive score is lower than the score threshold, The interaction between the operator and the identification evaluation device is completed through the human-computer interaction module;

所述存储模块存储待测电能计量装置的历史综合得分和各类别的得分及目标图像的图像识别分析数据;The storage module stores historical comprehensive scores and scores of various categories and image recognition analysis data of the target image of the electric energy metering device to be tested;

所述定位模块对待测电能计量装置所在的实时位置进行定位;The positioning module locates the real-time position where the electric energy metering device to be measured is located;

所述人机交互模块通过界面完成操作人员与识别评价装置的交互对话;The human-computer interaction module completes the interactive dialogue between the operator and the identification and evaluation device through the interface;

所述报警模块接收所述处理器的报警信号并完成报警。The alarm module receives the alarm signal of the processor and completes the alarm.

上述一种计量装置安装验收在线识别评价装置的评价方法,包括以下步骤:The above-mentioned evaluation method for an online identification and evaluation device for installation and acceptance of a metering device includes the following steps:

1)评分规则库的建立:根据产品的状态特征和产品状态特征对产品的影响程度,制定产品评分规则并入库;所述的产品评分规则为对表箱工艺检测、表箱安装偏正、表箱表面检测、检测表箱表面二维码提取、电表检测、电表安装偏正、电表铅封、电表入线、安全检测、漏线检测和接线工艺评分的规定,具体为:表箱工艺检测30分,表箱安装偏正10分,表箱表面检测10分,检测表箱表面二维码提取10分;电表检测30分,电表安装偏正10分,电表铅封10分,电表入线10分;安全检测40分,漏线检测20分,接线工艺20分;1) Establishment of the scoring rule base: According to the product status characteristics and the degree of influence of the product status characteristics on the product, formulate product scoring rules and incorporate them into the warehouse; Regulations on meter box surface inspection, detection of QR code extraction on meter box surface, meter inspection, meter installation bias, meter lead sealing, meter entry, safety inspection, leak detection and wiring process scoring, specifically: meter box process inspection 30 points, 10 points for meter box installation, 10 points for meter box surface inspection, 10 points for detection of QR code extraction on the surface of meter box; 30 points for meter inspection, 10 points for meter installation, 10 points for meter lead sealing, meter entry 10 points; 40 points for safety inspection, 20 points for leak detection, and 20 points for wiring process;

2)图像的采集和预处理:对现场待评价的电能计量装置进行图像和/或视频的拍摄,并将拍摄的视频文件逐帧转换为RGB图像,实时无损地对所拍摄的图像和/或视频进行采集,然后并对采集的图片和/或视频进行降噪处理;2) Image collection and preprocessing: image and/or video are captured on the electric energy metering device to be evaluated on site, and the captured video files are converted into RGB images frame by frame, and the captured images and/or Capture video, and then perform noise reduction processing on the captured pictures and/or videos;

3)图像的识别和分析:对预处理后的目标图像中的电表、接线、损坏和露线区域使用标注工具进行特征标注,然后使用深度学习Gaussian_yolo框架进行训练,获得模型文件并保存,接着加载训练好的模型文件,对采集的目标图像依次进行训练过程的步骤,多层次下的采样、卷积、池化和全连接层,获得并输出图像中每个目标的类别、位置和置信度信息;其中,所述训练的步骤包括:3) Image recognition and analysis: Use annotation tools to label the electricity meter, wiring, damage and exposed areas in the preprocessed target image, and then use the deep learning Gaussian_yolo framework for training, obtain the model file and save it, and then load it After the trained model file, perform the steps of the training process on the collected target images in turn, including multi-level sampling, convolution, pooling and full connection layers, to obtain and output the category, location and confidence information of each target in the image. ; Wherein, the step of described training comprises:

(1)配置训练文件并对图像文件进行分批;所述训练文件包括处理图像的尺寸、进行多层次下的采样、卷积、池化和全连接,以及每个处理的参数;(1) configure the training file and batch the image files; the training file includes the size of the processed image, multi-level sampling, convolution, pooling and full connection, and the parameters of each processing;

(2)根据配置的训练文件对每批次每张图像进行尺寸剪裁,缩放到同一尺寸;(2) Cut the size of each image in each batch according to the configured training file and scale it to the same size;

(3)对每张图像按照训练文件进行多层次下的采样、卷积、池化和全连接层进行特征提取;(3) Perform feature extraction on each image through multi-level sampling, convolution, pooling and fully connected layers according to the training file;

(4)对提取到的特征和标注的特征区域进行回归分析;(4) Perform regression analysis on the extracted features and marked feature regions;

(5)更新每个层次神经元的值,训练N次后,得到模型文件并保存;(5) Update the value of each level of neurons, after training N times, get the model file and save it;

4)图像的评价打分:根据图像中每个目标的类别、位置和置信度信息以及建立的评分规则,进行打分;所述评价规则按照百分制进行,满分100分,评价规则具体为:4) Image evaluation and scoring: according to the category, location and confidence information of each target in the image, and the established scoring rules, scoring; the evaluation rules are carried out according to the percentage system, with a full score of 100 points, and the evaluation rules are specifically:

类别得分=(总的检测目标数﹣异常的目标数)/总的检测目标数×该类别的分值分配×1/2Category score = (total number of detected targets - number of abnormal targets)/total number of detected targets × score distribution of this category × 1/2

当异常目标数为零时,则该类别为满分,各个类别的类别得分之和为综合得分;When the number of abnormal targets is zero, the category is full score, and the sum of the category scores of each category is the comprehensive score;

5)报警:根据打分给出的目标综合得分判别是否报警,当得分低于预设的得分阈值时,发出报警信号并进行报警;5) Alarm: according to the comprehensive score of the target given by the score to determine whether to alarm, when the score is lower than the preset score threshold, an alarm signal is issued and an alarm is performed;

6)数据存储:对目标图像的图像识别分析数据以及历史综合得分数据进行存储。6) Data storage: the image recognition analysis data of the target image and the historical comprehensive score data are stored.

Claims (7)

1. The utility model provides a metering device installation is checked and is accepted on-line identification evaluation device which characterized in that: the system comprises an image acquisition preprocessing module, an image recognition analysis module, an evaluation scoring module, a processor, a storage module, a positioning module, a human-computer interaction module and an alarm module;
the image acquisition preprocessing module consists of an image sensor, an acquisition module and an image noise reduction processing module, wherein the image sensor shoots images and/or videos of the electric energy metering device to be evaluated on site and converts the shot video files into RGB images frame by frame; the acquisition module acquires images and/or videos shot by the image sensor in a real-time and lossless manner; the image noise reduction processing module is used for eliminating noise in the picture and/or the video sent by the acquisition module;
the image recognition analysis module consists of an image training module and an image prediction module, wherein the image training module receives image information sent by the image acquisition preprocessing module, labels an image to be recognized, and then trains by using a deep learning Gaussian-yolo framework to obtain and store a model file; the image prediction module loads a trained model file, sequentially performs a training process on the collected target image, samples, convolutions, pooling and full-connection layers at multiple levels, and finally outputs the category, position and confidence information of each target in the image;
the evaluation scoring module consists of a scoring rule base and a quantification scoring module, wherein the scoring rule base specifies a product scoring rule according to the state characteristics of the product and the influence degree of the state characteristics of the product on the product; the quantitative scoring module is used for evaluating and scoring according to a scoring rule according to the category of each target in the image output by the image recognition and analysis module;
the processor processes and stores data sent by the image acquisition preprocessing module, the image recognition analysis module, the evaluation scoring module and the positioning module, sends an alarm instruction to the electric energy metering device to be evaluated, the comprehensive score of which is lower than a score threshold value, and completes the interaction between an operator and the recognition and evaluation device through the man-machine interaction module;
the storage module stores historical comprehensive scores and scores of various categories of the electric energy metering device to be tested and image recognition analysis data of the target image;
the positioning module is used for positioning the real-time position of the electric energy metering device to be measured;
the human-computer interaction module finishes interaction dialogue between an operator and the recognition and evaluation device through an interface;
and the alarm module receives the alarm signal of the processor and completes the alarm.
2. The evaluation method of the online identification and evaluation device for the installation acceptance of the metering device according to claim 1, characterized in that: the method comprises the following steps:
1) establishing a scoring rule base: according to the state characteristics of the product and the influence degree of the state characteristics of the product on the product, making a product scoring rule and warehousing the product scoring rule;
2) image acquisition and preprocessing: shooting images and/or videos of the electric energy metering device to be evaluated on site, converting the shot video files into RGB images frame by frame, collecting the shot images and/or videos in a real-time and lossless manner, and then carrying out noise reduction processing on the collected images and/or videos;
3) identification and analysis of images: marking the preprocessed target image by using a marking tool, then training by using a deep learning Gaussian _ yolo frame to obtain and store a model file, then loading the trained model file, sequentially carrying out the steps of a training process on the acquired target image, and obtaining and outputting the category, position and confidence information of each target in the image through sampling, convolution, pooling and full connection layers at multiple levels;
4) evaluation of the images was scored: scoring according to the category, position and confidence information of each target in the image and the established scoring rule;
5) and (4) alarming: judging whether to alarm or not according to the target comprehensive score given by the scoring, and sending an alarm signal and alarming when the score is lower than a preset score threshold value;
6) data storage: and storing the image identification analysis data and the historical comprehensive score data of the target image.
3. The evaluation method according to claim 2, characterized in that: the product scoring rule in the step 1) is a rule for detecting meter box processes, installing a meter box rightly, detecting the surface of the meter box, extracting two-dimensional codes on the surface of the detected meter box, detecting an electric meter, installing the electric meter rightly, sealing the lead of the electric meter, entering the electric meter, detecting safety, detecting missing lines and scoring a wiring process.
4. The evaluation method according to claim 2, characterized in that: and the marking in the step 3) is to mark the characteristics of the ammeter, the wiring, the damaged area and the exposed area in the image.
5. The evaluation method according to claim 2, wherein the step of training in step 3) comprises:
(1) configuring a training file and batching image files;
(2) cutting the size of each image of each batch according to the configured training file, and zooming to the same size;
(3) sampling, convolving, pooling and carrying out feature extraction on each image in a multi-level manner according to a training file;
(4) carrying out regression analysis on the extracted features and the marked feature regions;
(5) and updating the value of each layer of neuron, and obtaining and storing a model file after training for N times.
6. The evaluation method according to claim 2, characterized in that: the evaluation rule in the step 4) is carried out according to a percentage system, and the evaluation rule specifically comprises the following steps:
category score (total number of detected targets-number of abnormal targets)/total number of detected targets × score assignment of the category × 1/2
And when the number of the abnormal targets is zero, the category is full, and the sum of the category scores of all the categories is the comprehensive score.
7. The evaluation method according to claim 5, characterized in that: the training file in the training step 1) comprises the size of the processed image, the sampling, convolution, pooling and full connection under multiple levels and parameters of each processing.
CN202010690116.8A 2020-07-17 2020-07-17 On-line identification and evaluation device and evaluation method for installation and acceptance of metering device Pending CN111754737A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010690116.8A CN111754737A (en) 2020-07-17 2020-07-17 On-line identification and evaluation device and evaluation method for installation and acceptance of metering device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010690116.8A CN111754737A (en) 2020-07-17 2020-07-17 On-line identification and evaluation device and evaluation method for installation and acceptance of metering device

Publications (1)

Publication Number Publication Date
CN111754737A true CN111754737A (en) 2020-10-09

Family

ID=72710409

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010690116.8A Pending CN111754737A (en) 2020-07-17 2020-07-17 On-line identification and evaluation device and evaluation method for installation and acceptance of metering device

Country Status (1)

Country Link
CN (1) CN111754737A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113610490A (en) * 2021-07-13 2021-11-05 云南电网有限责任公司 Auxiliary acceptance platform and method for electric energy metering device
CN113988206A (en) * 2021-11-09 2022-01-28 刘帮奇 Image scoring generation system based on intelligent image recognition
CN114638970A (en) * 2022-03-14 2022-06-17 国网上海市电力公司 A normative judgment system for power facilities based on machine learning and edge detection
CN114677765A (en) * 2022-04-24 2022-06-28 东云睿连(武汉)计算技术有限公司 A system and method for comprehensive recognition and evaluation of interactive video actions
CN116228166A (en) * 2023-04-21 2023-06-06 广州智算信息技术有限公司 Water heater installation acceptance system based on artificial intelligence
CN116893386A (en) * 2023-07-17 2023-10-17 浙江涵普电力科技有限公司 Electric energy meter mounting process detection device and method based on deep learning image recognition

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107316131A (en) * 2017-06-09 2017-11-03 国网重庆市电力公司电力科学研究院 A kind of electric power meter mounting process quality detecting system based on image recognition
CN108647677A (en) * 2018-05-14 2018-10-12 山东师范大学 The ammeter appearance and performance intelligent detecting method and device that view-based access control model calculates
CN109447157A (en) * 2018-10-30 2019-03-08 国网四川省电力公司广元供电公司 A mobile model electric network operation cruising inspection system
CN109784348A (en) * 2018-12-17 2019-05-21 中国科学院深圳先进技术研究院 A kind of infrared power equipment identification and inline diagnosis method and its system
US20190154463A1 (en) * 2017-11-21 2019-05-23 International Business Machines Corporation Flow meter reading with image recognition secured with mask and software connected by mobile device
CN111222541A (en) * 2019-12-02 2020-06-02 国网浙江省电力有限公司 A method for identifying outer box types based on deep convolutional neural network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107316131A (en) * 2017-06-09 2017-11-03 国网重庆市电力公司电力科学研究院 A kind of electric power meter mounting process quality detecting system based on image recognition
US20190154463A1 (en) * 2017-11-21 2019-05-23 International Business Machines Corporation Flow meter reading with image recognition secured with mask and software connected by mobile device
CN108647677A (en) * 2018-05-14 2018-10-12 山东师范大学 The ammeter appearance and performance intelligent detecting method and device that view-based access control model calculates
CN109447157A (en) * 2018-10-30 2019-03-08 国网四川省电力公司广元供电公司 A mobile model electric network operation cruising inspection system
CN109784348A (en) * 2018-12-17 2019-05-21 中国科学院深圳先进技术研究院 A kind of infrared power equipment identification and inline diagnosis method and its system
CN111222541A (en) * 2019-12-02 2020-06-02 国网浙江省电力有限公司 A method for identifying outer box types based on deep convolutional neural network

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113610490A (en) * 2021-07-13 2021-11-05 云南电网有限责任公司 Auxiliary acceptance platform and method for electric energy metering device
CN113988206A (en) * 2021-11-09 2022-01-28 刘帮奇 Image scoring generation system based on intelligent image recognition
CN114638970A (en) * 2022-03-14 2022-06-17 国网上海市电力公司 A normative judgment system for power facilities based on machine learning and edge detection
CN114677765A (en) * 2022-04-24 2022-06-28 东云睿连(武汉)计算技术有限公司 A system and method for comprehensive recognition and evaluation of interactive video actions
CN116228166A (en) * 2023-04-21 2023-06-06 广州智算信息技术有限公司 Water heater installation acceptance system based on artificial intelligence
CN116228166B (en) * 2023-04-21 2023-08-04 广州智算信息技术有限公司 An artificial intelligence-based installation and acceptance system for water heaters
CN116893386A (en) * 2023-07-17 2023-10-17 浙江涵普电力科技有限公司 Electric energy meter mounting process detection device and method based on deep learning image recognition

Similar Documents

Publication Publication Date Title
CN111754737A (en) On-line identification and evaluation device and evaluation method for installation and acceptance of metering device
CN112233091B (en) An image damage detection and localization method for wind turbine blades
WO2017084186A1 (en) System and method for automatic monitoring and intelligent analysis of flexible circuit board manufacturing process
CN112858473A (en) Turnout switch blade damage state monitoring method based on feature fusion
CN112115927B (en) Intelligent machine room equipment identification method and system based on deep learning
CN109242439A (en) Feature extraction recognition methods based on substation equipment associated data
CN118504810B (en) Intelligent inspection optimizing system for power transmission line
CN111723774A (en) A target recognition method for power transmission equipment based on UAV inspection
CN118334563B (en) Power equipment abnormality detection analysis method and system based on monitoring environment
CN117975372B (en) Construction site safety detection system and method based on YOLOv and transducer encoder
CN111724290B (en) Environmental protection equipment identification method and system based on deep hierarchical fuzzy algorithm
CN113095189B (en) Machine vision and data analysis-based intelligent monitoring and management system for operating environment safety of power system transformer substation
CN118885941A (en) A fault prediction method based on 360° dynamic image detection system
CN117635905A (en) Intelligent monitoring method for electric energy meter installation quality based on image recognition algorithm
CN115511367B (en) Intelligent quality management system of production line
CN118014552A (en) Power plant coal conveying system inspection method and device and electronic equipment
CN118566235A (en) A method and system for visually detecting surface defects of fasteners
CN117314829A (en) Industrial part quality inspection method and system based on computer vision
CN107358297A (en) A kind of fast indexing method of state-detection data
CN119004377B (en) A harmful gas leakage detection method and system based on industrial multimodal data fusion
CN106774054A (en) GIS device analysis system and method based on the identification of complicated unstructured data
CN117114420B (en) Image recognition-based industrial and trade safety accident risk management and control system and method
CN113469938A (en) Pipe gallery video analysis method and system based on embedded front-end processing server
CN113724247A (en) Intelligent substation inspection method based on image discrimination technology
CN118447594A (en) Construction safety inspection method for farmland power station

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20201009

RJ01 Rejection of invention patent application after publication