基于SLIC超像素分割和ICP配准的牵引变压器套管红外三维立体重构方法

Infrared-three-dimensional Reconstruction Method of Traction Transformer Bushing Based on SLIC Super-pixel Segmentation and ICP Registration

针对目前在电力设备运维采用的数字孪生应用中,三维模型仅能展示设备空间维度数据,不能反映设备内部温度信息等,易造成设备内部放电故障难以检测的问题,提出一种基于红外图像的牵引变压器套管三维立体重构方法。利用红外热成像技术采集变压器套管的二维红外图像,使用SLIC算法对目标设备进行超像素分割,采用激光雷达扫描仪获取变压器套管的三维点云数据,经点云配准和滤波后重构三维模型,实现牵引变压器套管红外图像和三维模型的融合。以武汉铁路局集团有限公司信阳供电段110 kV变压器套管为算例进行三维立体重构测试。结果表明,设置了标记规则的基于SLIC的超像素分割技术可有效提高Dice及Jaccard系数,基于ICP算法的点云处理方法增加了配准精度和三角网格曲面质量,验证了新三维立体重构方法的有效性。

In the current digital twin application in power equipment operation and maintenance,the 3 D model can only display the spatial dimension data of the equipment,rather than reflect the temperature and other information inside the equipment,which may cause the problem of difficult detection of internal discharge fault of the equipment.To deal with this problem,a three-dimensional reconstruction method of traction transformer bushing based on infrared image was proposed.The infrared thermal imaging technology was used to acquire the two-dimensional infrared image of transformer bushing.The SLIC algorithm was used to perform super-pixel segmentation of the target equipment.The three-dimensional point cloud data of transformer bushing was acquired by lidar scanner.After point cloud registration and filtering,the three-dimensional model was reconstructed to realize the fusion of infrared image and three-dimensional model of traction transformer bushing.Based on the case of the 110 kV transformer bushing in Xinyang power supply section of Wuhan Railway Bureau,the 3 D reconstruction test results show that the SLIC-based super-pixel segmentation technology can effectively improve Dice and Jaccard coefficients.The point cloud processing method based on ICP improves the registration accuracy and the quality of triangular mesh surface,verifying the effectiveness of the new 3 D reconstruction method to display temperature information.