基于虚实融合的低能见度下航拍图像地平线检测

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.