摘要
针对现在输电通道巡视问题,本文搭建了可以垂直起飞降落的固定翼无人机平台,用于解决普通四旋翼续航短,普通固定翼起降场地要求严格的问题。并针对输电通道巡视照片的分类问题提出了一种改进型的卷积神经网络,通过对数据的预处理以及网络模型的优化,使图像分类性能得到了明显提升。
In view of the current inspection problem of power line corridor, a fixed-wing UAV platform capable of vertical take-off and landing is set up in this paper to solve the problems of short endurance of ordinary quadrotor and strict requirements for the landing of ordinary fixed-wing. An improved convolutional neural network is proposed to classify the photos of power line corridor. By preprocessing the data and optimizing the network model, the image classification performance is significantly improved.
出处
《人工智能与机器人研究》
2020年第2期55-63,共9页
Artificial Intelligence and Robotics Research
