Research And Implementation Of Anti-occlusion UAV Target Tracking Technology

The rapid development of UAV(Unmanned Aerial Vehicle)technology and the rapid improvement of computing processing capabilities provide a broad application prospect for UAV target tracking technology.In the real application scenario of UAV target tracking technology,the tracking target is easily blocked by obstacles and leads to tracking failure.Therefore,research on anti-occlusion,strong real-time and good accuracy target tracking algorithm for UAV target tracking Application will be significant.In this thesis,the related technologies of UAV target tracking algorithm and image-based visual controller are studied and analyzed in detail.The classic KCF(Kernel Correlation Filter)tracking algorithm has the advantages of fast tracking speed and real-time tracking of the target,but it does not perform well when the target is blocked;and the existing depth is introduced based on the KCF algorithm.The optimized learning algorithm improves the target tracking effect under occlusion,but it affects the real-time performance due to the high complexity.In this regard,this thesis intends to improve the anti-occlusion performance based on the KCF algorithm,and improve the anti-occlusion performance on the premise of ensuring real-time tracking.Based on the directional gradient histogram features used by the KCF algorithm,this thesis uses a multi-feature fusion mechanism and introduces color space features to jointly describe the target to improve the accuracy of the target description.Secondly,based on the KCF algorithm,tracking confidence is introduced to determine the severity of the target's occlusion,and this is used as a reference basis for position prediction of the occlusion target.When the target is heavily occluded,the position prediction value is used as the result of target tracking.The OTB2015 data set is used to compare the improved algorithm with the classic algorithm.The results show that the optimized algorithm improves the tracking accuracy and anti-occlusion performance,and because of the use of position prediction,the reliability of real-time target tracking is also improved.In the second half of this article,according to the requirements of the research and development project,based on the DJI M100 UAV platform,a UAV target tracking system is developed and designed for moving target tracking,and the improved anti-occlusion algorithm proposed on the ground control end is applied.In order to achieve effective control of UAV flight on the ground,the basic flight control commands are encapsulated,and a custom communication protocol is specially designed for the transmission of control commands.The results of outdoor flight control tests show that our UAV target tracking system can effectively achieve real-time and effective tracking of general moving targets and moving targets with different degrees of occlusion.This means that the research goal of the project is achieved.