【摘要】 大部分的电压致热型设备由于温升不明显，故障判断难度较大，并且红外图像受到背景复杂，尺寸多的影响，导致红外图像特征提取困难，另外，大多数现有的检测工作依靠人工判别，且存在判断分歧。目前，代替人工判别的研究中，许多只利用单一的特征来表征热图像中的故障特征。针对此类问题，笔者提出一种基于多特征融合的电压致热型设备故障提取方法。此方法将红外图像多故障目标检测算法和多特征融合算法相结合，建立动态决策准则。首先提取红外图像的三种典型特征，包括颜色、纹理和轮廓特征，继而通过融合算法(串行融合、并行融合和DCA)将三种不同属性的故障特征进行融合，充分地提取设备红外图像的故障特征，从而提高识别的准确性。验证表明，相对于某一低层特征故障提取，本研究的方法在电压致热型设备红外图像故障诊断特征提取方面，不仅保证时效性的同时，红外图像故障分类精确度得到了提高，分类更加可靠。更多还原

【Abstract】 Most of the voltage heating equipment is difficult to determine the fault due to the temperature rise is not obvious, and the infrared image is affected by the complex background and variable size, which leads to the low accuracy of fault identification in the infrared image. In addition, most of the existing work depends on manual judgment, and there are differences in judgment. In the research of replacing manual discrimination,it is only used a single feature to characterize the fault features in thermal images. To solve these problems, the author proposes a fault feature extraction method of Voltage heating equipment based on multi-feature fusion. It combines the infrared image multi-fault target detection algorithm with the multi-feature fusion algorithm to establish dynamic decision criteria. Firstly, the low-level visual features of infrared image of equipment are extracted, including color feature, texture feature and contour feature. Then, three fault features of different attributes are fused through fusion algorithms(Serial feature fusion, Parallel fusion and DCA) to fully extract fault features of infrared image of equipment, so as to improve the accuracy of recognition. Compared with a certain low-level feature fault extraction, the verification shows that the method presented in this paper not only ensures the timeliness but also improves the accuracy of infrared image fault classification and makes the classification more reliable.更多还原