基于红外热像和权值直接确定神经网络的零值绝缘子识别方法

On-site Identification of Zero Resistance Insulator Based on Infrared Thermal Image and One-step Weights-determination of Neural Network

提出一种利用红外图像和神经网络权值直接确定法进行现场零值绝缘子识别的方法。识别的基本过程为:对红外图像进行图像增强、去噪、分割,在绝缘子串内截取矩形作为目标图像;鉴于红外热像的灰度值与物体表面温度的关系,直接提取了灰度标准偏差值、绝对偏差、四分差以及极差4个特征参数;将这4个参数作为权值直接确定法神经网络的输入来训练模型,并用于现场零值绝缘子的识别。该方法有效剔除了现场识别时输电线路的干扰,且能满足现场识别实时性要求。实验结果验证了该方法的可行性与有效性。

A method using infrared thermal images and one-step weights-determination of neural network to identify the zero resistance insulators on-site is presented. The basic procedures are as follows： the infrared thermal images are intensified, denoised, segmented, and a rectangular which was regarded as object is intercepted in the insulators chain; in view of the relationship between gray value of infrared thermal images and temperature of object surface, four parameters which stand for standard deviation, absolute deviation, quartile and range of gray value, are extracted directly; these four parameters are used as the input of the neural network of one-step weights-determination to train the model, which could be used to identify the zero resistance insulators after being trained. This method can effectively avoid the interference of transmission lines, and can meet the real-time requirement when identifying on-site. Experimental results verify the feasibility and effectiveness of this method.