【作者】 于淼；

【导师】 王文华；

【作者基本信息】 吉林大学 ， 工程硕士（专业学位）， 2024， 硕士

【摘要】 遥感技术因其能够提供广泛的覆盖范围、实现瞬时成像并高效获取数据,成为对地观测的重要工具,已经被广泛应用于气象、农业、军事和海洋等多个重要领域。因此,从遥感图像中提取有用信息是提高应用效率的关键。然而,目前可见光对地遥感观测技术面临的一个主要问题是白云遮挡导致大量遥感图像无效。除气象卫星外,高比例云覆盖的遥感图像应用价值低,不仅消耗了宝贵的卫星资源,同时也浪费地面处理资源。针对上述问题,本文创新性地提出一种卫星先导云检测遥感辅助方法,设计并开发一款低成本、低功耗的卫星先导辅助相机,按照任务需求,该相机需要在卫星飞临预定目标前获取飞行前方的遥感图像,进行云量检测并将可观测遥感的区域指向发送至卫星平台。使高分主相机(高分辨率主相机)更加有效、更加有针对性地对地遥感观测。本文设计的先导云检测相机可实现在轨实时云检测的功能,旨在提高遥感图像的质量和可用性。通过理论仿真和实际成像实验,验证了本文设计的先导辅助相机方案的可行性和有效性,具体研究内容如下:1)设计了一款卫星先导云检测辅助遥感相机,通过分析计算先导辅助相机幅宽、各通道输出条带图像的云占比、高分主相机启动时间以及卫星机动时间,先行判断卫星飞行前方云量,并将评估结果发送给卫星平台,由卫星平台控制高分主相机是否执行成像任务、卫星是否需要侧摆及侧摆角度的多少,进而提高遥感图像的质量和利用率。2)在轨实时云检测过程中,由于传统云检测算法对遥感相机波段数目和硬件资源要求高,难以在先导相机上应用,因此,本文引入一种适应先导辅助相机波段和硬件资源配置的云检测算法。该算法基于HSV色彩模型变换、形态学处理等数字图像处理理论,对先导辅助相机获取的遥感影像进行实时云检测研究,通过理论仿真和实际成像实验,该算法检测非雪区域的云占比检测误差小于2%,查全率和查准率都在90%以上,错误率小于4%。3)根据某高分主相机的参数指标设计先导云检测相机,地面分辨率为101m,焦距为27mm,视场角为23.6°,幅宽为209km,功耗小于2.5w,两个相机夹角为4.84°,先导辅助相机被配置为在预定成像区域前6s启动进行云检测。根据先导辅助相机成像系统的硬件框架,设计CMOS成像单元模组,完成先导辅助相机的研制工作;搭建云检测技术地面实验平台,开展云检测算法的地面实验,验证先导云检测方法的有效性,实验结果满足相机任务要求。综上所述,为解决云层覆盖导致高分主相机获取图像利用率低以及星上和地面资源浪费的问题,本文设计了一款新型先导云检测相机,作为高分主相机的辅助单机,实现在轨实时云检测。通过软件仿真分析和硬件实验验证了算法的可行性,达到在极短时间内完成云检测的目的。在航天遥感空间监测和在轨实时云检测领域具有一定的工程应用价值。更多还原

【Abstract】 Remote sensing technology has become an essential tool for Earth observation due to its ability to provide extensive coverage,achieve instantaneous imaging,and efficiently acquire data.It has been widely applied in key areas such as meteorology,agriculture,military,and oceanography.Consequently,extracting useful information from remote sensing images is crucial for enhancing application efficiency.However,a significant challenge faced by current visible light Earth observation remote sensing techniques is the obstruction caused by clouds,which renders many images unusable.Apart from meteorological satellites,remote sensing images with a high proportion of cloud cover have low application value,leading to the wastage of precious satellite resources and ground processing resources.To address the aforementioned challenges,this thesis innovatively proposes a satellite-based pilot cloud detection method for assisting remote sensing,designing and developing a low-cost,low-power consumption pilot assistance camera.According to mission requirements,this camera must acquire remote sensing imagery of the area ahead of the satellite’s flight path before it overflies the predetermined target.It performs cloud quantification detection and directs the observable remote sensing area to the satellite platform,enabling the high-resolution main camera to conduct Earth observation more effectively and purposefully.The pilot cloud detection camera designed in this thesis achieves on-orbit real-time cloud detection functionality,aimed at improving the quality and usability of remote sensing images.Through theoretical simulations and actual imaging experiments,the feasibility and effectiveness of the pilot assistance camera scheme are verified.The specific research contents are as follows:1)A satellite-based pilot cloud detection assistance remote sensing camera was designed.By analyzing and calculating the swath width of the pilot assistance camera,the cloud occupancy rate of the output strip images from each channel,the startup time of the high-resolution main camera,and the satellite maneuvering time,it preemptively judges the cloud amount ahead of the satellite flight path.The evaluation results are sent to the satellite platform,which controls whether the high-resolution main camera should execute the imaging task and whether the satellite needs to perform yaw and the extent of the yaw angle,thereby improving the quality and utilization rate of remote sensing images.2)Traditional cloud detection algorithms,due to their high requirements on the number of bands of remote sensing cameras and hardware resources,are challenging to apply on pilot cameras.Therefore,this thesis introduces a cloud detection algorithm adapted to the band and hardware resource configuration of the pilot assistance camera.Based on HSV color model transformation,morphological processing,and other digital image processing theories,this algorithm conducts real-time cloud detection research on remote sensing images acquired by the pilot assistance camera.Through theoretical simulation and actual imaging experiments,the algorithm achieves a cloud occupancy detection error of less than 2% in non-snow areas,with a recall and precision rate of over 90% and an error rate of less than 4%.3)The pilot cloud detection camera was designed according to the parameter specifications of a high-resolution main camera,with a ground resolution of 101 m,focal length of 27 mm,field of view of 23.6°,swath width of 209 km,and power consumption of less than 2.5w.The angle between the two cameras is 4.84°,and the pilot assistance camera is configured to start cloud detection 6 seconds before entering the predetermined imaging area.Based on the hardware framework of the pilot assistance camera imaging system,a CMOS imaging unit module was designed,completing the development work of the pilot assistance camera.A ground-based cloud detection experimental platform was established,and ground experiments on the cloud detection algorithm were conducted,verifying the effectiveness of the pilot cloud detection method.The experimental results meet the camera mission requirements.In summary,to address the issue of low utilization rate of high-resolution main camera images due to cloud cover and the wastage of satellite and ground resources,this thesis designs a novel pilot cloud detection camera as an auxiliary unit to the highresolution main camera.It achieves on-orbit real-time cloud detection and verifies the feasibility of the algorithm through software simulation analysis and hardware experiments,accomplishing cloud detection in a very short timeframe.This has significant engineering application value in the fields of aerospace remote sensing spatial monitoring and on-orbit real-time cloud detection.更多还原