Font Size: a A A

Design Of Airborne Ground Target Tracker Based On ZYNQ

Posted on:2019-08-18Degree:MasterType:Thesis
Country:ChinaCandidate:H R YangFull Text:PDF
GTID:2382330566498038Subject:Instrument Science and Technology
Abstract/Summary:PDF Full Text Request
Airborne ground target tracking is widely used in video aerial photography,outdoor rescue,traffic grooming,and military investigation,because it is not subject to signal transmission interference and does not rely on ground control.The airborne ground target tracking task has the characteristics of small target size,complex scene,target distance and scale change,etc.It puts forward higher requirements for the tracking algorithm's robustness.However,if the target tracking algorithm with good robustness is to realize real-time tracking on the airborne processing platform with limited power consumption and load capacity,the on-board computing platform should have greatly computing capabilities.In order to solve the contradiction between tracking robustness and real-time performance in airborne ground target tracking tasks,this project chooses the Kernelized Correlation Filter(KCF)tracking algorithm which has less complexity and improves it to promote its tracking robustness.Based on ZYNQ UltraScale+ Multiprocessor System-on-Chip(MPSoC),an airborne ground target tracking computing platform was designed to achieve real-time ground target tracking with high tracking robustness.Firstly,after comparing and analyzing the current mainstream tracking algorithms,this paper designed a KCF-based target tracking algorithm based on the premise of satisfying tracking precision and tracking success rate.For the problem that the KCF method cannot adapt to scale changes,a scale adaptive strategy is designed,and an improved KCF method is proposed to improve the adaptive tracking capability of the scale.Secondly,an airborne ground target tracking computing platform is designed based on ZYNQ UltraScale+ MPSoC.In order to meet the real-time requirements of the airborne tracker,the high level synthesis technology is applied,what's more,parallel optimization is performed to improve the calculation process of the KCF tracking algorithm.After that,the parallel optimized tracking algorithm is exported to an intellectual property core.Based on that,the on-chip system is designed and corresponding embedded applications and a human-computer interaction interface is developed.A full-featured airborne ground target tracker system was built to achieve a complete tracking process from video streaming access to tracking result output.Finally,several detailed test experiments were designed.Based on the ZUC102 evaluation board,the UAV123 data set was used to comprehensively test the tracker's trackingThe test results show that the airborne ground target tracker designed by this subject can realize the real-time and accurate tracking of the ground target,and at the same time adapt to the scale change of the tracking target well.The performance of the tracking device meets the design index requirements well.This subject has designed an airborne ground target tracker with strong robustness,good real-time performance and fast processing speed,which has clear practical value.At the same time,our subject is meaningful to other image processing tasks.
Keywords/Search Tags:UAV, Target tracking, Kernelized correlation filter, Parallel acceleration of algorithms
PDF Full Text Request
Related items