摘要
为了能够解决在复杂光照背景条件下的变形绝缘子的缺陷诊断,提出了一种基于形状轮廓描述的缺陷分类诊断方法。采用了传统图分割方法分割出绝缘子的轮廓,对轮廓样本点进行采样,分析变形绝缘子轮廓的采样点,存在轮廓的样本点没有明显切线的情况,造成缺失了相应的形状上下文信息。由于变形绝缘子背景复杂,噪声点较多,未对图像轮廓进行处理直接使用形状上下文,会使得无法提取绝缘子的完整轮廓,采用图分割的方法改进了这一点。改进后的形状上下文算子以参考点与轮廓重心的连线为x轴,可以建立完整的极坐标系,点集形成相对向量,这些向量构成缺陷轮廓的形状上下文信息。改进后的算法可以很好提升缺陷识别的准确率,对输电线部件对应的缺陷特征来进行缺陷分类,经实验测试实现缺陷部件分类效果。
A fault classification diagnosis method based on shape contour description is proposed.The traditional image segmentation method is used to segment the outline of the insulator,sample the contour sample points,and analyze the sampling points of the deformed insulator profile.There is no obvious tangent to the reference point,which results in the missing shape context information.The improved shape context operator takes the line connecting the reference point and the center of gravity of the contour as the x-axis,and a complete polar coordinate system can be established.The point set forms a relative vector,and these vectors constitute the shape context information of the defect contour.
出处
《工业控制计算机》
2019年第6期37-38,40,共3页
Industrial Control Computer
关键词
图分割
形状轮廓
采样
相对向量
缺陷分类
graph segmentation
shape outline
samplingr
relative vector
defect classification
