基于优化列文伯格-马夸尔特法的SLAM图优化  

Optimization of SLAM Graph Based on Optimization Levenberg-Marquardt Method

作　　者：戴天虹[1] 李志成 DAI Tian-hong;LI Zhi-cheng(School of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150036, China)

机构地区：[1]东北林业大学机电工程学院,哈尔滨150036

出　　处：《哈尔滨理工大学学报》2021年第2期68-74,共7页Journal of Harbin University of Science and Technology

基　　金：中央高校基本科研业务费专项资金(2572019CP17);黑龙江省自然科学基金(C201414);哈尔滨市科技创新人才项目(2014RFXXJ086).

摘　　要：针对目前的视觉SLAM技术中存在的非线性优化方法过程复杂、优化速度慢、优化精度低等缺点,在广泛应用的BA非线性优化方法的框架基础之上,对其核心下降策略列文伯格-马夸尔特法进行优化,以便改善传统的列文伯格-马夸尔特法在BA非线性优化中的不足之处。首先,初始化待优化点处的信赖区域半径;其次,拟定一个扩大倍数,并设定阈值;最后,通过拟定的近似范围限定每次迭代的信赖区间,以达到非线性优化的目的。通过设置对比实验和分析实验的仿真结果,可以得出经过优化后的列文伯格-马夸尔特法下降策略能够加快优化速度,具有优化建图的效果。Aimed at the present situation of the visualSimultaneous Localization and Mapping technology of complex nonlinear optimization method of process,optimization of slow speed and low accuracy of optimization shortcomings,in the wide application of nonlinear optimization method based on the framework of Bundle Adjustment,drops to its core strategy Levenberg-Marquardt method is optimized,so as to improve the traditional Levenberg-Marquardt deficiency in Bundle Adjustment nonlinear optimization.Firstly,the trust region radius at the optimization point is initialized.Secondly,draw up an expansion multiple and set the threshold.Finally,the proposed approximate range is used to limit the confidence interval of each iteration,so as to achieve the purpose of nonlinear optimization.By setting up the simulation results of comparative experiments and analysis experiments,it can be concluded that the optimized Levenberg-Marquardt descent strategy can speed up the optimization speed and has the effect of optimizing the mapping.

关 键 词：BA非线性优化 列文伯格-马夸尔特法下降策略 信赖区域 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]