摘要
为了提高电力设备检测效果、及时恢复供电,提出基于AI视觉技术的电力设备检测方法。从不同角度全面拍摄电力设备图像,采用边界特征融合法检测电力设备视觉图像边缘轮廓,结合特征分解和尺度模板匹配提取视觉图像的奇异特征,经模型训练和测试后,结合L-ReLU激活函数获取最佳电力设备异常检测结果。经实验验证,该方法对电力设备视觉图像奇异特征提取效果较好,电力设备检测精度较高,适应度波动小,收敛度高,受天气及噪声等环境影响小。
A power equipment detection method based on AI vision technology is proposed to improve the detection effect and facilitate the timely restoration of power supply.The power equipment images are taken comprehensively from different angles,the edge contour of the power equipment visual image is detected by the boundary feature fusion method,and the singular features of the visual image are extracted by combining the feature decomposition and scale template matching.After model training and testing,combined with the L-ReLU activation function,the best power equipment anomaly detection results are obtained.Experiments show that this method has good effect on singular feature extraction of visual image of power equipment,high detection accuracy of power equipment,small adaptability fluctuation,high convergence,and little influence by weather,noise and other environment.
作者
李杨
董元龙
林明晖
高明
岳衡
丁靖
LI Yang;DONG Yuanlong;LIN Minghui;GAO Ming;YUE Heng;DING Jing(State Grid Ningbo Electric Power Supply Company,Ningbo 315000,China)
出处
《微型电脑应用》
2023年第9期90-93,共4页
Microcomputer Applications
关键词
视觉技术
电力设备
深度神经网络
异常检测
视觉图像
激活函数
visual technology
power equipment
deep neural network
anomaly detection
visual image
activation function
