基于深度学习的绝缘子串检测算法研究

Research on Insulator String Detection Algorithm Based on Deep Learning

因绝缘子串在输电线路中作用重要、发生故障频率高等特点,绝缘子串检测已成为架空输电线路巡检任务的主要内容之一。为了解决传统检修方法效率低、成本高的缺点,文章提出了一种使用关键点对绝缘子串进行检测的算法。该算法使用自主标注的绝缘子串数据集,采用预训练模型VGG19、旋转、放缩等数据增强方法、梯度更新方法优化、中间监督等策略对模型性能进行多次改进优化。实验结果表明,最终模型精度值达到59.2%,比初始模型提高了15%,精准度有了显著提高。

Due to the important role of insulator string in transmission line and high fault frequency, insulator string detection has become one of the main tasks of overhead transmission line inspection. In order to solve the shortcomings of traditional maintenance methods, such as low efficiency and high cost, this paper proposes an algorithm to detect insulator strings using keypoints. The algorithm uses the self-labeled insulator string data set, uses the pre-training model vgg19, data enhancement methods such as rotation, expansion and contraction, gradient update method optimization, intermediate supervision and other strategies to improve the performance of the model. The experimental results show that the accuracy of the final model reaches59.2%, which is 15% higher than that of the initial model, and the accuracy is significantly improved.