采用对象分割和图像配准方法的变电站红外预警系统

Substation Infrared Early Warning System Using Object Segmentation and Image Registration

为了早期发现变电站事故隐患,建立了一个红外温度巡检与预警系统,该系统能够自动循环采集变电站电气设备的红外图像,通过提取目标设备的温度信息进行故障判别。为了解决机械传动造成的图像位置偏差导致系统难以对感兴趣设备进行定位监测的问题,提出了一种基于对象分割和图像配准的校正方法,首先对图像进行对象分割和组态定义,并将待测目标提取出来,然后采用基于相位相关和Harris角点匹配的图像配准方法对序列偏差图像进行配准。使用上述方法对某变电站中获取的15组电气设备红外遥视图像进行了实验,平均正确匹配率达到93.3%,并且配准精度较高,达到了亚像素级。实验表明该方法能够对偏差图像进行有效校正,使系统能够及时准确地获得设备部件的温度,保障监测系统的可靠运行。

To detect the defects of the electrical apparatus earlier ,an infrared temperature inspection and early-warning system was established. The system could automatically gather the infrared images of the electrical apparatus and detect the corresponding temperature information. In order to position deviation caused by mechanical transmission issue based on object segmentation and image registration correction algorithm,an image object segmentation and configuration definitions were extracted from the target under test.Based on phase correlation,Harris corner points matched image registration methods on sequence deviation image registration. Fifteen groups of infrared images of the electrical apparatus in a substation were used to test the proposed approach. The results show that the averaged correct matching rate has achieved 93.3%,and the matching precision reaches the Asian picture element level. The experiment results also indicate that the proposed approach is effective in the registration of the deviation images,enabling the system to accurately obtain the temperature of the desired parts of the electrical apparatus.