摘要
由于雾天、光线较暗等恶劣现场条件下采集到的绝缘子图像清晰度与可读性较低,绝缘子目标及盘面区域色彩特征的提取较难,导致现有的可见光图像污秽检测方法不具备通用性,为此提出了一种基于图像增强的瓷质绝缘子灰密程度检测方法。先用改进的带颜色恢复的多尺度Retinex(MSRCR)算法对采集到的绝缘子图像进行增强,提高图像的清晰度和对比度;然后,采用二维最小误差法结合形态学滤波分割提取出绝缘子盘面区域,分别提取6个通道的均值、最大值、最小值等7个特征量并用Fisher准则函数筛选出分类能力较强的特征Smean,Smax,Svar作为灰密程度判别特征;最后,用思维进化算法(MEA)优化反向传播(BP)神经网络进行仿真预测。实验结果表明,概率神经网络和粒子群优化算法优化BP神经网络的判别准确率分别为88.00%和93.00%,而所提方法的准确率可达95.00%,可以准确判别恶劣条件下的绝缘子灰密程度。
It is difficult to gather the high-definition and readable images of insulators in bad conditions such as foggy day and dim light.The available contamination detection methods of visible images does not have generality due to the difficulty of extracting insulator objects and color features of surface region.Therefore,this paper proposes a image enhancement based detection method of non-soluble deposit density（NSDD）levels of porcelain insulators.Firstly,the improved multiple scale Retinex with color restoration（MSRCR）algorithm is used to enhance the clarity and contrast of the collected insulator images.Secondly,the two-dimensional minimum error algorithm and the morphological filter algorithm are combined to segment and extract the surface region of insulators.And seven characteristic values in six channels are extracted,such as mean value,maximum value,minimum value.Then the Smean,Smaxand Svar with highly classification ability are selected as the identify features of NSDD levels by using the Fisher criterion function.Finally,the back propagation（BP）neural network optimized by the mind evolutionary algorithm（MEA）is used for simulation and forecast.The experiment results show that the recognition accuracy rates of the probabilistic neural network algorithm and the BP neural network optimized by particle swarm optimization（PSO）algorithm are 88.00% and 93.00%,respectively.In comparison,the accuracy rate of the proposed method is 95.00%,which shows that it can accurately identify the NSDD levels of insulators in bad conditions.
作者
黄新波
杨璐雅
张烨
曹雯
李立浧
HUANG Xinbo;YANG Luya;ZHANG Ye;CAO Wen;LI Licheng(School of Electronics and Information, Xi＇an Polytechnic University, Xi＇an 710048, China;School of Electric Power, South China University of Technology, Guangzhou 510641, China)
出处
《电力系统自动化》
EI
CSCD
北大核心
2018年第14期151-157,共7页
Automation of Electric Power Systems
基金
陕西省重点科技创新团队计划资助项目(2014KCT-16)
陕西省工业科技攻关项目(2016GY-052)
国家自然科学基金资助项目(51707141)~~
关键词
绝缘子
带颜色恢复的多尺度Retinex算法
特征提取
FISHER准则
反向传播神经网络
灰密程度检测
insulator
multiple scale Retinex with color restoration（MSRCR）algorithm
feature extraction
Fisher criterion
back propagation（BP）neural network
non-soluble deposit density（NSDD）level detection
