摘要
针对电力设备红外图像包含大量的噪声,且设备边缘较为模糊,传统图像分割方法无法有效提取红外图像中电力设备等问题,提出一种基于密度相似因子的电力设备红外图像分割方法。首先,对获取的电力设备红外图像分别进行R、G、B三通道光照不均匀性校正,其次,转换至Lab彩色空间并构造待分割的特征空间信息点集,然后采用最小距离原则分配信息点到最邻近的聚类中心,再通过平均连接代价和k-距离邻域的平均连接代价,计算出信息点的密度相似因子,最终实现电力设备红外图像滤除噪声分割。通过与K均值和模糊c均值对实际绝缘子红外图像的分割实验定量对比,实验结果表明,本文所提方法具有噪声滤除能力,改善了分割效果。
A density-similarity-fact0r-based segmentation method is proposed to solve the problem that power equipment cannot be distinguished effectively from noise and edge blur in infrared images by traditional image segmentation methods. First, the illumination non-uniformity of the R, G, and B channels of the original infrared images of the power equipment is corrected. Then, the R, G, and B are converted to lab color space, and the information point set of the feature space of the segmentation image is constructed. The information points are distributed to the nearest cluster center according to the principle of minimum distances. The density similarity factor of the information point is calculated from both the average connection cost, and the average connection cost of k-distance neighbors that is adopted to remove the noise points. The experimental results show that the proposed method eliminates the noise points and improves the effectiveness of segmentation in comparison with the traditional clustering algorithms that use K-means and fuzzy c-means.
出处
《红外技术》
CSCD
北大核心
2017年第12期1139-1143,共5页
Infrared Technology
关键词
故障诊断
红外图像分割
密度相似因子
fault diagnosis, segmentation of infrared images, density-similarity-factor
