应用聚类和证据理论实现变电站保护压板状态校核

Application of Clustering and Evidence Theory to Checking the Status of Substation Protection Platen

针对目前人工校核保护压板效率低、图像识别易受光照条件影响的问题,提出了一种基于聚类和D-S证据理论的变电站保护压板状态校核方法。首先对压板图像进行聚类分割,按行检查压板图像像素量和边界点,重新分割异常压板。根据压板在投、退状态下的形态特征,构造压板状态指标作为证据,分析了光照、透视变形因素对压板识别的影响。最后,将多个状态指标依据证据理论进行融合,减少个体指标的不确定性影响。结果表明,所提方法在光照条件不良情况下仍能正确识别压板状态,性能优于单一判据及多判据表决方法。

Aiming at the problems of low efficiency of manual checking protection platen and the vulnerability of image recognition to illumination conditions,a method of checking the status of substation protection platen based on clustering and D-S evidence theory is proposed.Firstly,the image of platen is clustered and segmented,and the pixel size of protection platen image and boundary nodes are checked by row,and the abnormal platen is re-segmented.Secondly,according to the morphological character of the platen in the state of on or off,the state index of the platen is constructed as evidence,and the effects of illumination and perspective deformation factors on the recognition of the platen are analyzed.Finally,the multiple state indicators are fused according to evidence theory to reduce the uncertainty of individual indicators.The results show that the proposed method can correctly identify the state of the platen under poor illumination conditions,and its performance is better than that of single criterion and multi-criterion voting methods.