基于语义分割与连通区域标记的隔离开关状态识别方法

State Recognition Method for Disconnector Based on Semantic Segmentation and Connected Component Labeling

提出一种基于语义分割与连通区域标记的隔离开关状态识别方法。首先,提出基于语义分割算法的隔离开关像素提取方法,将图像按照语义类别进行像素分类,实现对开关臂所在像素的提取。其次,提出基于区域生长算法的语义分割图像标记方法,实现对隔离开关连通区域的标记,并提出面积排序统计方法优化面积阈值和消除非开关臂区域。最后,根据开关臂连通区域个数判断隔离开关状态。通过引入风格迁移算法生成隔离开关风格化图像增强训练集,提升污损图像中隔离开关定位、分割的准确性。实验结果表明所提方法能够准确定位、分割隔离开关,并识别开关状态。

A state recognition method for disconnectors based on semantic segmentation and connected component labeling is proposed. Firstly, a method of disconnector pixel extraction based on semantic segmentation algorithm is proposed. The images are classified according to the semantic category, and the pixels of the disconnector arm are extracted. Secondly, a semantic segmentation image labeling method based on the region growing algorithm is proposed to label the connected component of disconnectors, and an area-sorting statistical method is proposed to optimize the area threshold and eliminate the non disconnector arm region. Finally, the state of the disconnector is judged according to the number of connected components of the disconnector arm. In order to improve the accuracy of disconnector location and segmentation in defaced image, the style transfer algorithm is introduced to generate the stylized images of disconnectors to enhance the training set. Experimental results show that the proposed method can accurately locate and segment disconnectors, and identify the state of disconnectors.