基于深度学习的输电线路防振锤故障检测

Vibration hammer fault detection for transmission lines based on deep learning

针对传统输电线路防震锤故障检测准确率低,检测的正检率和误检率效果不佳的问题,为此提出并设计基于深度学习的输电线路防振锤故障检测。通过构建深度学习的网络模型提取图像特征,对训练分类器进行识别图像处理,将处理结果输入到待测图像上所对应的感受野区域,得出防振锤的概率。对概率阈值进行设置,同时利用检测器判断该点对应的原始待测图像中的防振锤故障,当检测的图像回溯到原始图像上,防振锤故障被准确检测出来。测试结果表明:与传统的基于图像形态学的检测方法相比,基于深度学习的检测方法减少了冗余计算,正检率提高了19.1%,误检率降低了10.8%,说明基于深度学习的检测方法适合应用在输电线路防振锤故障检测中。

Aiming at the problems of low accuracy of traditional transmission line anti-vibration hammer fault detection and poor effect of positive and false detection,a fault detection method of transmission line anti-vibration hammer is proposed and designed based on deep learning.The image features are extracted by constructing a network model of in-depth learning,and the training classifier is processed by image recognition.The processing results are input into the corresponding receptive field area of the image to be tested,and the probability of anti-vibration hammer is obtained.The probabilistic threshold is set,and the detector is used to judge the hammer fault in the original image to be measured.When the detected image is traced back to the original image,the hammer fault is detected accurately.The test results show that compared with the traditional method based on image morphology,the method based on depth learning reduces the redundancy calculation,increases the positive detection rate by 19.1%,and reduces the false detection rate by 10.8%.It shows that the method based on depth learning is suitable for the fault detection of transmission line anti-vibration hammer.