基于深度学习的变电设备红外热像识别

Infrared Thermal Image Recognition of Substation Equipment Based on Deep Learning

在变电设备状态监测与故障诊断中,可自动识别变电设备的红外热像识别技术是关键技术之一。为解决目前在变电设备的红外热像识别中存在的背景温度过度集中、对比度低、缺乏智能方法等问题,提出了一种使用RetinexNet算法对图像进行增强的方法,为红外热像的精准识别创造条件;使用YOLOX-Darknet53算法对增强后的图像进行目标检测。在试验中,使用该方法对红外热像进行识别,不仅每张图像的识别时长可以达到6.88 ms,且8种变电设备识别的平均精确率可以达到96.51%。实验数据验证了,所提方法的高效性和精准度,可以满足监测变电设备状态的需求。

In the process of condition monitoring and fault diagnosis of substation equipment,the infrared thermal image recognition technology that can automatically identify substation equipment is one of the key technologies.In order to solve the problems of excessive concentration of background temperature,low contrast and lack of intelligent methods in the current infrared thermal image recognition of substation equipment,a method of image enhancement using RetinexNet algorithm is proposed to create conditions for accurate recognition of infrared thermal images.Object detection is performed on the enhanced image using YOLOX-darknet53 algorithm.In the experiment,the infrared thermal images are recognized by used the proposed method.Not only the recognition time can reach 6.88 ms each piece of image,but also the average accuracy of 8 kinds of substation equipment can reach 96.51%.The experimental data show that the method is efficient and accurate,and can meet the needs of monitoring the status of substation equipment.