基于改进SSD的电力设备红外图像异常自动检测方法

Infrared Image Anomaly Automatic Detection Method for Power Equipment Based on Improved Single Shot Multi Box Detection

为实现各类巡检机器人、无人机等智能电力巡检设备所携红外热像仪采集的红外图像自动检测,该文提出基于改进SSD的电力设备红外图像异常自动检测方法。对收集的典型故障电力设备红外图像统一预处理;标注电力设备及异常区域并制作标准数据集;搭建检测网络;读入数据与预训练模型到搭建的网络进行微调训练验证,得到最终模型文件并测试。实验表明,该方法泛化性强,准确率较高,能达到实时自动检测红外图像下多类典型电力设备并定位异常发热区域的效果,将使现有电力巡检设备实现“智能+”。

In order to realize automatic detection of infrared images collected by infrared thermal imaging cameras carried by intelligent power inspection equipment such as various inspection robots and drones,an automatic detection method for infrared image anomalies of power equipment based on improved SSD is proposed.The infrared image of the typical faulty power equipment collected is uniformly preprocessed;the power equipment and the abnormal area are marked and a standard data set is created;the detection network is built;the data and the pre-training model are read into the established network for fine-tuning training verification,and the final model file is obtained.test.Experiments show that the method has high generalization and high accuracy;it can achieve the effect of real-time automatic detection of many types of typical power equipment under infrared images and locate abnormal heating areas,which will make the existing power inspection equipment“smart+”.