基于虚实融合的低能见度下航拍图像地平线检测  

Virtual-real Fusion Based Horizon Detection Under Low Visibility Conditions

作　　者：刘长江[1,2,3] 张轶[1,2] 杨红雨[1,2] 

机构地区：[1]四川大学视觉合成图形图像技术国家重点学科实验室,四川成都610065 [2]四川大学计算机学院,四川成都610065 [3]四川理工学院理学院,四川自贡643000

出　　处：《四川大学学报（工程科学版）》2012年第4期110-114,共5页Journal of Sichuan University (Engineering Science Edition)

基　　金：国家重点基础研究发展计划项目子课题资助项目(2009CB320803);四川理工学院理学院科研项目(10LXYB04)

摘　　要：地平线检测是飞行器导航系统的控制和稳定性分析中重要的步骤。针对低能见度下低空航拍图像,提出了基于虚实融合的地平线检测算法。首先依据GPS/AHRS组合导航系统提供的航姿信息,得到地平线的准确斜率及参考位置,然后在该参考位置上方选取种子点区域,根据该区域像素的统计特性计算出阈值,采用区域生长法进行天地分割,最后根据最小误差准则确定地平线的准确位置。文中算法应用于实测数据,实验结果表明算法能有效地提取正确的地平线,能为飞行器在低能见度下进近着陆阶段提供必要的导航信息。Horizon detection plays a vital role in the control system of an aircraft during approach and landing. A new horizon detection algorithm based on virtual and actual information fusion was proposed. Firstly, position and pose data were extracted from GPS/AHRS guidance system, and then the exact slope and estimated equation of horizon were calculated with the help of virtual model. Further- more, the variance of the region above the horizon was acquired. Finally, precise equation of the horizon was determined using region growing and estimation algorithms. Experiment and visual measurement showed this algorithm can effectively detect horizon under low visibility conditions, and provide much auxiliary information to pilots during approach and landing.

关 键 词：虚实融合 低能见度 地平线检测 区域生长 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]