摘要
提出一种新的基于彩色可见光图像的高压绝缘子污秽等级判别方法。对深圳变电局所属多个变电站进行现场拍摄获取污秽绝缘子可见光图像,并实验获取其对应等值附盐密度,经图像灰度化、图像增强、滤波后,用两次最大类间方差法进行分割,得到盘面积污区域。提取积污区域的RGB、HSV空间共36个特征分量,并运用Fisher判别法进行特征量筛选。用筛选的特征量训练BP神经网络,建立可见光图像污秽等级判别网络。试验结果表明可见光污秽等级判别法具有较高准确率,是一种检验高压绝缘子污秽等级的可行方法。
A new method is presented, using digital images of high voltage insulators to detect the contamination grades. Firstly, we obtain the sample digital images from the transformer substations in Shenzhen City by taking photos of insulators with different contamination grades and their ESDD （equivalent salt deposit density） respectively. Secondly, after graying, enhancing and denoising, we obtain the polluted area of an insulator with two times of OTSU image threshold segmentation method. Thirdly, we visit 36 characteristics from RGB and HSV space, and with the help of Fisher criterion, we extract the features that can reflect the characteristics of insulator contamination grades. Finally, we train a BP neural network with the extracted features, whose accuracy rate is 88% in detecting the contamination grades of sample images. The experimental results show that this method is feasible for detecting the contamination grades of high voltage insulators, which may provide some engineering significance for insulator cleaning and flashover prevention.
出处
《电工电能新技术》
CSCD
北大核心
2015年第9期70-74,80,共6页
Advanced Technology of Electrical Engineering and Energy
基金
国家自然科学基金(51177109)
电力设备电气绝缘国家重点实验室(EIPE14211)资助项目
