Transmission line detection in aerial images: An instance segmentation approach based on multitask neural networks

作者：

Highlights：

• Introduction of instance segmentation framework to solve the camera-based transmission line detection problem.

• Assign correct cable ID for segmented cable pixels for the requirements of aircraft navigation.

• Dilated convolutional layers and spatial convolutional layers are configured in fully convolutional networks with two branches of outputs.

• Superior performance compared with traditional methods and high efficiency based on implementation on GPUs.

摘要

•Introduction of instance segmentation framework to solve the camera-based transmission line detection problem.•Assign correct cable ID for segmented cable pixels for the requirements of aircraft navigation.•Dilated convolutional layers and spatial convolutional layers are configured in fully convolutional networks with two branches of outputs.•Superior performance compared with traditional methods and high efficiency based on implementation on GPUs.

论文关键词：Transmission line detection,Deep neural networks,Convolutional neural networks

论文评审过程：Received 30 April 2020, Revised 13 January 2021, Accepted 30 March 2021, Available online 17 April 2021, Version of Record 21 April 2021.

论文官网地址：https://doi.org/10.1016/j.image.2021.116278