基于分水岭与Krawtchouk不变矩相结合的改进方法在变电站巡检图像处理中的应用

Application of an Improved Algorithm Based on Watershed Combined With Krawtchouk Invariant Moment in Inspection Image Processing of Substations

变电站的巡检机器人在执行巡检任务时,由于位置的变动导致采集的现场设备图像内容出现差异。针对传统定点拍摄的图像处理方法不能满足变电站设备图像处理要求的问题,该文提出了一种基于标记分水岭模型与Krawtchouk不变矩相结合的改进算法,以实现巡检图像中的设备存在遮挡物及仿射变换情况下的目标分割与特征量提取,基于Hsim函数完成最终的目标识别。以松原500 k V变电站巡检机器人实际拍摄的图像为例,结果证明所提算法能够对设备图像进行有效的分割与分类,并通过特征量匹配实现对变电站设备现场状态的甄别,为变电站现场设备图像智能处理提供一种可行的方案。

This paper proposed an improved algorithm of marked Watershed model combined with the Krawtchouk invariant moment for automatic equipment monitoring through image segmentation and feature extraction, in the case of occurrence of obscured objects and affine transformation. Hsim function was used to complete the recognition and classification procedure. Actual images, obtained by a real substation inspection robot from Songyuan 500 kV substation, were adopted in the experiment of verifying the feasibility and applicability of this algorithm. The results indicate that this algorithm could work properly on equipment image segmentation and classification. And it could provide a practicable solution of smart image processing by monitoring the running states of equipment via object image feature extraction and comparison, and performing inspection tasks in different locations or perspectives by inspection robots used in power substations.