视觉辅助的飞行校验关键技术研究

Flight inspection (FI) is one of the most basic prerequisites for opening one civil aviation airport and operating route. Also, it is important and necessary to ensure that the equipments in communication, navigation and radar meet the normal operation requirements of the flight. In the current aviation safety inspection, one hot research issue is that how to achieve the changing of our country from the big one to the power one in the FI field. In this propose, we focus on the difficult FI problems with low visibility in night-time or airflow instability at highland airports. To propose the key technologies of vision-aid flight inspection system, our research includes the following components. Firstly, based on the analysis of airport structure, we establish a correlation expression model of visual features among multiple source images. Driven by the structural characteristics of special target, the image processing algorithms including image enhancement, image fusion and target detection are developed, which can be used to estimate the actual trajectory of inspection aircraft as the test benchmark for the instrument landing system in FI. According to human visual characteristics, a visual change detection and analysis model is studied for small targets, which is used to intelligently monitor the runway navigation lights. Owing to the real-time interaction between the airborne FI system and the visual aid system, the man-made risk can be reduced. Moreover, it can support the FI operation under complex conditions, and solve the basic theory and key technical problems for the development of FI system on UAV.

飞行校验是民航机场开放和航路运行的最基本前提之一，也是保证通信、导航、雷达等设备符合航班正常运营要求的必要手段。如何实现将我国从飞行校验领域大国到强国的转变，是目前航空安全检查的热点问题之一。本项目针对我国夜间低能见度、高原机场气流不稳定等飞行校验难点问题，研究视觉辅助的飞行校验关键技术。主要内容包括深入分析机场结构特性，建立多源图像视觉特征相关性表达模型；以此为基础，研究目标结构特性驱动的图像增强、图像融合、目标检测等图像处理算法；通过坐标反演计算实现对校验飞机的实际轨迹估计，为仪表着陆系统校验提供测试基准；以人眼视觉特性为依据，研究高效的小目标视觉变化检测分析模型，实现跑道助航灯光的智能监测。本项目的机载校验系统与视觉辅助系统可以进行实时交互，降低飞行校验过程中的人为风险，为复杂条件下的飞行校验提供技术支撑，从而解决研制无人机飞行校验装备所需的基础性理论和关键技术难题。

近 5 年来，我国自主研制的 CFIS 系统满足了常规飞行校验的需求。但随着我国航空工业的迅猛发展，飞行校验飞机和设备还存在很大缺口，特别是与未来高精度智能化飞行校验的需求还有一定差距。. 本项目系统地研究高动态环境下的图像处理，以及 ILS 校验、进近辅助灯光自动监测的相关理论和技术，在深入分析目标成像特性的前提下，完成了图像特征提取与融合分析方面的研究，根据细粒度目标识别任务进行可分离性特征提取模型构建，设计自主学习的融合策略，构建了基于深度学习的目标分类识别模型，为视觉信息处理提供良好的技术支撑。关于视觉定位误差分析方面，实现基于参照点几何关系的快速透视变化矩阵求解算法，研究基于合作参考点、基于固定位置相机、基于运动相机的不同情况下视觉定位方案，较好地实现静态目标、移动目标的定位误差分析。以人眼视觉特性为依据，构建了高效的机场斑马线视觉检测分析方法，实现飞行校验的 PAPI 跑道助航灯光自动监测。针对无人机飞行校验系统（UFIS）的飞行校验问题，开发三维视景监视系统、设计无人机飞行校验平台下飞行指引综显指示器，并进一步在北航东营研究院展开的无人机飞行校验系统科研试飞中进行系统测试验证。. 本项目的研究为建立视觉辅助的飞行校验系统提供理论基础，为未来搭建无人机飞行校验平台提供技术支撑。