低光照条件下基于图像增强的视觉惯性组合导航方法  

Visual-inertial Integrated Navigation Method Based on Image Enhancement under Low Illumination Conditions

作　　者：李雪强 李建胜[1] 王安成[1] 郭雨岩 付飞 赵刚 LI Xueqiang;LI Jiansheng;WANG Ancheng;GUO Yuyan;FU Fei;ZHAO Gang(Information Engineering Unitersity,Zhengzhou 450001,China;61618 Troops,Beijing 100080,China;Xi'an Satellite Control Center,Xi'an 710043,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]61618部队,北京100080 [3]西安卫星测控中心,陕西西安710043

出　　处：《测绘科学技术学报》2025年第1期21-26,共6页Journal of Geomatics Science and Technology

基　　金：国防科技重点实验室基金项目(6142403210201);国家自然科学基金重点项目(42330113)。

摘　　要：针对视觉惯性SLAM在光照变化剧烈和低光照场景下容易产生漂移甚至系统失效的情况,提出了一种基于图像增强的单目视觉惯性组合导航算法。首先对图像进行改进的自适应伽马变换和CLAHE处理,增强图像亮度和对比度;之后采用基于角度一致性的改进RANSAC算法对误匹配点进行剔除;最后将视觉信息与IMU预积分进行紧耦合,得到稳定的位姿结果。EuRoC数据集中4个光照变化大和低光照的困难序列验证结果表明,与开源的VINS_Mono算法比较,本文算法定位精度提升13.2%~25.8%不等。A monocular visual-inertial integrated navigation algorithm based on image enhancement is proposed to solve the problem that visual-inertial SLAM is prone to drift and even system failure in the scene of intense illumination change and low illumination.Firstly,improved adaptive Gamma transform and CLAHE processing are used to enhance image brightness and contrast.Then,an improved RANSAC algorithm based on angle consistency is used to eliminate the mismatched points.Finally,the visual measurements and IMU pre-integral are tightly coupled to obtain stable pose results.Four difficult sequences with large illumination variation and low illumination in EuRoC dataset were used for verification.Compared with that of the open source VINS_Mono algorithm,the positioning accuracy of the proposed algorithm is improved by 13.15%~25.81%.

关 键 词：低光照 图像增强 角度一致性 视觉惯性 组合导航 

分 类 号：P228[天文地球—大地测量学与测量工程]