Intelligent Flight Control Of Highway Inspection UAV Based On Visual Navigation

With the rapid development of the highway traffic construction,the highway maintenance tasks have rapidly increased.In recent years,most of the UAV road inspection systems that have emerged are manual operating systems,which are still low in efficiency and cannot meet the increasing demand for highway inspections.In order to improve the inspection efficiency of the highway inspection UAV platform,this research studies the vision-based intelligent flight control algorithm of the highway inspection UAV,and builds an autonomous vision navigation UAV system for highway inspection.Nowadays,UAV systems based on vision navigation are still dominated by traditional multi-data information fusion.The object detection and tracking algorithms in traditional multi-data fusion systems have low efficiency in detecting ground objects,resulting in slower flight speeds of UAV systems,and even failures.The flight situation.Aiming at the main problems of the current navigation UAV inspection system.This research analyzes the design requirements of the autonomous visual navigation system of the highway inspection UAV platform based on the background of the highway inspection task,designs a new UAV platform object detection and tracking model.Introducing a video judgment box to judge and analyze the results of the highway center marking tracking,and obtain the corresponding instruction information to realize the intelligent control of the drone.The accuracy and speed of the proposed model is improved by optimizing the new network structure and loss function of YOLOv3 model.Deep-SORT(Deep Simple Online Real-time Tracking)algorithm is used to track the highway center marking on the detected information.The experimental results of object detection and tracking show that the detection efficiency is greatly improved compared with other models.Compared with several types of typical object tracking models,the processing speed is increased while the tracking accuracy is basically unchanged.Finally,the ground station APP is designed based on Android and DJI Mobile SDK platform,the platform and environment configuration process are introduced in detail,and the main functions of the ground station APP are designed.This research tests all functions of the UAV system one by one.The test result shows that the system can realize the visual navigation flight of the highway center markings according to the preset route.In summary,by integrating deep learning technology,object tracking technology,and data fusion concepts this research developed a new vision navigation highway inspection UAV intelligent control algorithm and ground station APP platform based on quadrotor UAV.This research improved the navigation UAV system problems such as poor recognition efficiency of ground target information,and initially realized the autonomous visual navigation flight of the highway inspection UAV.