输电线路是电力系统中电能传输的关键,随着无人机技术和机器视觉的快速发展,为电力线路绝缘子识别提供了便捷.基于无人机拍摄的输电线路破损绝缘子图像识别,能有效降低人工巡检的成本和安全隐患.为了提高绝缘子识别精确度,提出一种基于图像灰度均衡化和超分辨率卷积神经网络的图像增强算法.灰度均衡化算法能够改善图像的对比度,提高图像清晰度;超分辨率卷积神经网络图像增强算法实现了图像的超分辨率重建,从而使图像清晰度更高、真实度更强.通过使用VGG16作为特征提取和分类器,结合迁移学习实现模型的训练,对增强后的彩色图像数据集进行训练,进一步提高了识别精度.在权威数据集CPLID上验证了所提出算法的有效性与高识别精度,并比较了不同算法模型的精确度,所提出的深度迁移学习算法在实际应用中能有效降低人工巡检的成本和安全隐患.
Power transmission lines play a key component in the power system for transmitting electricity.With the rapid development of UAV technology and machine vision,identifying damaged insulators in power transmission lines has become more convenient.The identification of defective insulator image captured by UAV in power transmission lines can effectively reduce the cost and safety hazards of manual inspections.In order to improve the accuracy of insulator identification,an image enhancement algorithm based on image grayscale equalization and super-resolution convolutional neural network(SRCNN)is proposed.Grayscale equalization algorithm can improve the contrast and sharpness of the image.The image enhancement algorithm based on SRCNN can realize the super-resolution reconstruction of the image,which makes higher sharpness and authenticity.By using VGG16 as the feature extractor and classifier and incorporating transfer learning for model training,the accuracy of the identification is further improved by training the enhanced color image dataset.The effectiveness and high recognition accuracy of the proposed algorithm are validated on the authoritative CPLID(Chinese power line insulator dataset),and the accuracy of different algorithm models is compared.The deep transfer learning algorithm proposed can effectively reduce the cost and safety risks of manual inspections in practical applications.