Research On Inspection Of Optical Cable Lines And Construction Machinery Based On ARM

The rapid development of my country’s telecommunications industry,many related industries such as mobile payment,sharing economy are inseparable from optical cable.The optical cable trunk directly affects the economy,and even concerns national defense.Once these important optical cables are dug by external forces such as construction,the loss will be immeasurable.And a large number of such events occur every year.Optical cable inspection refers to confirming along the operator’s optical cable,whether the optical cable is being damaged by external force or the risk of being damaged by external force.The traditional inspection method is mainly manual inspection,and there are many optical cable trunks in the suburbs with undeveloped traffic.Many places need manual walking to confirm the safety of the optical cable.In many dangerous terrain inspections,the security problems of the inspectors are difficult to guarantee and cannot ensure Inspector’s attendance rate.Based on the fact that the vast majority of heavy construction machines can pose a threat to the buried optical cables,this article specifically proposes a smart inspection solution based on the Raspberry Pi and 4G networked industrial drone: training before PC A target detection model based on Yolo V3-tiny was developed,and then a Raspberry Pi with a recognition model deployed on a networked industrial drone to achieve real-time target detection.Identify the main categories of excavators,bulldozers and trucks,select these markers to select the photos of hidden dangers on the optical cable path,then manually confirm,and finally determine whether they are hidden danger points,and dispatch relevant personnel to coordinate construction to ensure Optical cable safety.The main work and research content of this article are: collecting data to making a data set;training and deploying models;comparing and comparing the performance of models obtained from the same data set in different device environments;writing auxiliary programs based on Python to achieve Raspberry Pi 4B Automation of target detection.After testing,it takes about 1s to detect a single image,and the recall rate of the target detection model is 67.62%.