恶劣天气下中高空无人机可见光图像退化及补偿研究

Mid-high altitude Unmanned Aerial Vehicle (UAV) are important platforms to obtain information in the air-space-ground integrated earth observation network. In bad weather such as haze, floating dust and smoke, visible image of mid-high altitude UAV degrades seriously. The acquisition of high quality image is a challenging problem to be solved. At home and abroad, most of the researches on optical image degradation and compensation under the influence of atmosphere are concentrated in the fields of low altitude outdoor image and satellite image, which cannot perform well to solve the problem of UAV image because of different atmosphere environment and imaging conditions. In order to solve the new problem of visible image degradation and compensation in bad weather for mid-high altitude UAV, this work is carried out from three aspects: (1) The degradation mechanism of optical image of UAV in bad weather is analyzed, and the image degradation model is established. (2) A new method based on the combination of global image restoration and local image enhancement is proposed, and the key techniques of ill-posed image restoration model and video compensation are broken through. (3) Based on engineering requirement of haze removal for UAV image and video, the verification experiments of degradation model and compensation method are taken, utilizing the real flight data and application system. The expected accomplishment can significantly improve the visible image quality to meet the urgent application requirements of mid-high altitude UAV for earth observation, disaster monitoring and emergency rescue and battlefield detection. Furthermore, it can also promote the development of the theory and method for mid-high altitude optical image degradation and compensation.

中高空无人机是空天地一体化对地观测网络中重要的信息获取平台。在雾霾、浮尘、烟雾等恶劣天气下，中高空无人机可见光图像退化严重，高质量图像获取成为亟待解决的难点问题。国内外大气影响下光学图像退化及补偿研究大多集中在低空户外图像和卫星图像领域，由于大气环境和成像条件不同，不能较好地解决无人机图像问题。针对恶劣天气下中高空无人机可见光图像退化及补偿这一新问题，本项目从三方面开展研究：（1）分析恶劣天气下中高空无人机光学图像退化机理，建立图像退化模型；（2）提出“全局图像复原+局部图像增强”的补偿方法，突破图像复原模型病态问题求解、视频补偿等关键技术；（3）基于无人机图像/视频去雾霾的工程需求，利用实飞数据和已有应用系统，完成退化模型及补偿方法的验证。预期成果可提高可见光图像质量，满足中高空无人机常态对地观测、灾害监测与应急救援、战场探测等迫切应用需求；推进中高空光学图像退化及补偿理论和方法的发展。

本项目针对雾霾、浮尘、烟雾等恶劣天气下中高空无人机可见光图像降质问题，开展光学图像退化模型及可见光图像/视频去雾霾应用研究，包括恶劣天气下中高空无人机光学图像退化机理、 中高空无人机可见光图像/视频补偿方法和中高空无人机可见光图像/视频去雾霾应用验证。.取得成果包括：（1）在机理层，①针对低能见度大气分层条件，建立了中高空无人机光学图像分层退化模型，并提出复原图像求解方法。②针对非均匀多散射问题，提出了两种散射模型，分析得到了不同环境约束条件下模型的有效性以及模型之间的关系。（2）在方法层，①从图像增强角度出发，提出了基于图像融合的去雾算法，通过设置全局融合权重和自动色阶得到增强图像，大大提高处理效率。②针对视频处理，提出了一种基于时空一致性优化的无人机视频去雾方法，解决了当帧间场景变化较大时，复原视频出现帧间闪烁的现象。③提出了一种大气能见度估计方法，建立了能见度与图像复原模型的关系，通过有雾图像估计得到能见度定量数值。④提出了一种基于亮色区域填充的无人机图像去雾方法，改进了暗通道先验算法在亮色区域复原图像失真的现象。⑤提出了一种元数据辅助的无人机图像多线程清晰化方法，在保证了图像清晰化效果的同时，实现了对无人机图像的实时处理。（3）在应用层，①以中高空无人机数百小时实飞图像数据和遥测数据为数据源，建立不同能见度和高度的可见光图像数据集。②建立了用于去雾前后图像质量评价的综合评价指标。③实现4种图像去雾方法与本项目提出方法对比。④完成图像复原、图像增强2类算法和软件的设计编码和实验验证。.本项目研究有助于推动中高空光学图像退化及补偿理论和方法的发展。研究成果可用于雾霾、浮尘、烟雾等恶劣天气下高质量可见光图像/视频的获取，为无人机可见光对地观测应用提供关键技术支撑，使可见光载荷应用在一定程度上突破恶劣天气限制，提升无人机全天候执行常态观测、灾情监控和战场探测的能力。