中文摘要：

针对变压器散热器油路不通故障,提出一种结合多元信息的散热器红外故障自动识别方法.首先将彩色红外热像进行加权灰度化处理得到灰度图像,接着采用Canny算子进行边缘检测,再利用概率霍夫变换检测直线获得直线段集合.经过两级直线滤波后,对直线段进行分类与评价,再进行散热器轮廓的重构与图像分割,最后结合提出的图像沿重力方向的投影线积分,自动识别散热器红外故障.实验结果验证了所提方法的有效性.

英文摘要：

Combining multi-information,an automatic recognition method for infrared fault of trans-former radiator was proposed.The canny operation was first used to detect edges on weighted gray image of infrared thermography.Then the probabilistic Hough transform (PHT)was exploited to generate line segment set.After using a two-stage filter,the lines respectively were classified and e-valuated,and the contour of radiator was reconstructed.Then,the mask of radiator was generated by using image segmentation technology.Finally,the infrared fault of transformer radiator could be au-tomatic recognized by using the proposed image gray projecting line integral and the effectiveness of the proposed approach was verified.