基于方差分析的绝缘子红外热像特征选择方法

利用红外成像技术检测绝缘子污秽等级的关键在于获取分类性能优异的红外热像特征，文章提出了基于方差分析的红外热像特征选择方法。利用直方图均衡化增强原始热像图对比度，人工截取绝缘子盘面图像区域；通过平滑后的图像直方图包络线提取分割阈值，对阈值分割后的二值图像进行形态学滤波，得到绝缘子盘面图像和背景图像，提取2者的最高温度、最低温度、平均温度、温度分布方差以及盘面相对于背景的最大温升和平均温升共10个红外热像特征；应用单因素方差分析甄别特征优劣，实现特征选择。瓷绝缘子人工污秽试验结果表明：文中提出的红外热像特征选择方法和图像分割算法简单有效。

Feature Selection Method of Insulator Infrared Thermal Image Based on Variance Analysis

As for detecting the insulator contamination grades by infrared thermal imaging, it is the key problem to obtain infrared thermal image features with excellent classification performance. For this purpose, the authors propose an infrared thermal image feature selection method based on variance analysis. By means of equalizing the histogram, the contrast of original thermal image is enhanced; then image region of the disc surface of insulator is intercepted manually; through the envelope of the smoothened histogram of thermal image the segmentation threshold is extracted and the morphological filtering is applied to the threshold- segmented binarized image to obtain insulator disc surface image and background image, and then 10 infrared image features of insulator disc surface image and background image, such as the highest temperature, the lowest temperature, mean temperature, variance of temperature distribution of both kinds of images, as well as maximum temperature rise and mean temperature rise of insulator disc surface relative to background, are extracted; finally, the good and bad features are discriminated by single factor variance analysis, thus the feature selection is implemented. The results of artificial pollution test of ceramic insulators show that the proposed selection method for infrared thermal image and the image segmentation algorithm are simple and effective.