Research On Target Tracking Algorithm Against Cloud Occlusion Based On TLD

In recent years,the research and development of military unmanned aerial vehicle technology has maintained a high degree of popularity.Unmanned aerial vehicles have attracted extensive attention from military national defense of various countries due to their advantages of small size,fast speed and rich application scenes.At present,the technology of machine vision is also in a rapid development stage,and tracking algorithms are constantly updating.It is of great significance for the development of UAV technology to reasonably integrate these algorithms into UAV research.Considering that cloud cover often occurs when unmanned aerial vehicles(UAVs)track ground targets,based on TLD target tracking framework,this thesis has conducted in-depth research on the long-term tracking of targets under random cloud cover.The main work is as follows:This thises summarizes the research progress of target tracking algorithm at present,expounds the principle and basic flow of TLD algorithm in detail,and analyzes the advantages and disadvantages of TLD tracking algorithm and traditional tracking algorithm.Aiming at the tracking drift phenomenon in the early stage of cloud occlusion,a TLD tracking algorithm based on FAST feature point enhancement is proposed.Experimental results show that the improved algorithm effectively overcomes the problem of target tracking drift,and the stability and real-time performance of the tracking module are improved.Aiming at the defects that the target tracking algorithm cannot effectively estimate and output the target position when the clouds are blocking,and the running time of the detection module is huge,which leads to low real-time performance of the algorithm,an improved TLD algorithm fused with Kalman filter is proposed.The experimental results show that the improved algorithm can still stably estimate the target state after the target is completely blocked by the clouds,and the Kalman filter is used to optimize the detection module in the algorithm,thus greatly improving the real-time performance of the tracking algorithm.According to the actual application requirements,through the improvement of TLD algorithm,the improved TLD algorithm has better adaptability under random cloud coverscenes,and the overall stability and real-time performance of the algorithm are also improved.