Research On Security Warning Model Of Intelligent Optical Cable Based On YOLOv3-tiny

Abstract

In order to ensure the unbolcked communication of optical cable lines need to be regularly inspected.Most of the existing UAV(Unmanned Aerial Vehicle) patrol systems work by collecting images from the aircraft and sending them back to the ground station for processing.It will take a long time and may be affected by the wireless channel, which makes it impossible to stop and intervene in time. This paper proposes an intelligent security warning system for optical cable line based on YOLOv3-tiny.By enlarging the pre-processed images of engineering vehicles, the data set is expanded and the generalization ability of the model is improved. The target box of the data set is reset by k-means clustering algorithm, which improves the parameter performance of YOLOv3-tiny. Finally, the model system was tested on the Nvidia jetson tx2 platform, and the results showed that the system could quickly carry out real-time detection of construction vehicles, with the Mean Average Precision (mAP) up to 0.33 and the detection speed up to 24fps.