基于改进FAST特征检测的ORB?SLAM方法

2018-03-13 20:49:14公维思周绍磊吴修振刘刚
现代电子技术 2018年6期

公维思+周绍磊+吴修振+刘刚

摘 要: 针对视觉ORB?SLAM中的特征点提取过程,提出一种改进的FAST特征点检测方法。在ORB?SLAM过程中,由于光照和周围环境信息的不确定性,当环境对比度下降时,易造成特征点数量急剧减少从而造成SLAM过程的失败。针对上述问题,基于图像的对比度信息,提出一种自适应阈值的FAST特征点检测方法,利用图像的对比度自适应调整特征点检测时的阈值,当图像对比度下降时依然能够保证特征点检测的数量,大大提高了ORB?SLAM的鲁棒性和适应性。对提出的方法进行了实验验证,结果表明该方法切实可行,具有较高的实用价值。

关键词: ORB?SLAM; 特征点检测; 图像对比度; 自适应阈值; SLAM过程; 特征点提取

中图分类号: TN964?34; V249.329 文献标识码: A 文章编号: 1004?373X(2018)06?0053?04

Abstract: Aiming at the feature point extraction process in visual ORB?SLAM, an improved FAST feature point detection method is proposed. In ORB?SLAM process, when the environment contrast degree decreases, the number of feature points tend to fall sharply due to the uncertainty of illumination and surrounding environment information, which leads to the failure of the SLAM process. To solve the above problem, an adaptive threshold based FAST feature point detection method is proposed based on image contrast degree information. Image contrast degree is used for adaptive adjustment of feature point detection thresholds so that the number of feature points for detection can be guaranteed when image contrast degree decreases, and the robustness and adaptability of ORB?SLAM can be improved greatly. An experiment was carried out to verify the proposed method. The results show that the method is feasible and has a high practical value.

Keywords: ORB?SLAM; feature point detection; image contrast degree; adaptive threshold; SLAM process; feature point extraction

0 引 言

近年来,随着机器人技术的发展,基于视觉的同时定位与地图构建技术(Visual Simultaneous Localization and Mapping, V?SLAM)也得到了越来越高的重视。V?SLAM是指机器人在未知环境中通过自身携带的视觉传感器信息确定当前空间位置,并建立所处空间的环境模型,是实现机器人任务规划、导航、控制等自主运动的前提之一[1]。

V?SLAM的典型应用就是在无人机中,无人机可以携带视觉传感器,通过视觉传感器获取当前空间的信息,实时向地面传输数据,进行当前空间的地图构建。V?SLAM还可以应用在导航系统中,与惯导系统结合,解决卫星导航系统信号失效场合下的载体导航问题。

如今最典型的视觉导航为ORB?VSLAM,本文基于ORB?SLAM特征点,提出一种具有自适应技术的特征点检测算法。……

登录APP查看全文