摘要
为了提升变电站无人机巡检图像的边缘识别精度,构建了一种结合改进Canny算法与轻量化深度学习模型的混合检测方法。研究动态高斯核抑制金属反光噪声的作用,优化自适应阈值计算策略,并引入SE注意力机制增强特征提取能力。实验结果表明,该方法在边缘连续性、误检率及计算效率方面均优于传统算法,实现了高效、精准的设备缺陷识别,为智能巡检系统提供了高精度的图像分析支持。
In order to improve the edge recognition accuracy of unmanned aerial vehicle inspection images in substations,a hybrid detection method combining improved Canny algorithm and lightweight deep learning model is constructed.The effect of dynamic Gaussian kernel on suppressing metal reflection noise is studied,the adaptive threshold calculation strategy is optimized,and SE attention mechanism to enhance feature extraction capability is introduced.The experimental results show that this method is superior to traditional algorithms in terms of edge continuity,false detection rate,and computational efficiency,achieving efficient and accurate equipment defect recognition and providing high-precision image analysis support for intelligent inspection systems.
作者
李更达
LI Gengda(Nanning Monitoring Center of UHV Transmission Company of China Southern Power Grid Co.,Ltd.,Nanning,Guangxi 530000,China)
出处
《自动化应用》
2025年第18期23-25,共3页
Automation Application
关键词
无人机巡检
边缘检测
高斯核优化
深度学习
UAV inspection
edge detection
Gaussian kernel optimization
deep learning
