【作者】 刘威；

【导师】 王文华；

【作者基本信息】 吉林大学 ， 电子信息（专业学位）， 2025， 硕士

【摘要】 随着空间技术的不断进步,光学遥感卫星不断朝着高空间分辨率、高时间分辨率、高光谱分辨率与多传感器融合等方向发展,急剧膨胀的遥感图像数据量使得卫星平台资源的有效利用变得尤为重要。然而,地球全年平均有超过66%的区域被云层所覆盖,被云层所覆盖的光学遥感图像几乎没有利用价值,且占用大量宝贵的星载固存资源与数传下行带宽。因此,针对光学遥感卫星由于地面云层覆盖而导致的资源浪费与成像利用率下降问题,本文提出了具有遥感任务规划能力的星载云判仪系统,基于在轨实时云检测技术预先获取的地表云层分布信息实现在轨智能任务决策,结合新型敏捷卫星的任意曲线非沿轨道动中成像能力为高分辨率主载荷遥感相机提供云占比优化的成像推扫路径,改善了云层覆盖条件下的遥感卫星资源利用效率。本文的具体研究内容如下:1)根据在轨实时云检测与遥感任务规划的应用需求,面向某型高分辨多光谱遥感相机设计了空间分辨率30m、成像幅宽240km的星载云判仪前视辅助相机方案,并与高分主载荷遥感相机构成5°的前置安装角,能够在主载荷目标成像区域前6s启动以进行在轨云检测与遥感任务优化;设计了基于拜尔阵列图像传感器与可编程逻辑器件的硬件方案,并实现了高速图像数据流采集与Camera Link Medium数传输出的FPGA接口驱动逻辑。2)在星载云判仪的硬件平台基础上,设计了基于多维度特征与最小距离分类器的在轨实时云检测算法,实现了遥感云图光谱特征与纹理特征的联合判别,云检测算法的精度达90%;设计了遥感推扫路径优化策略,采用侧摆成像、曲线推扫和相机停机三种策略对主载荷遥感相机进行任务优化,改善了在大量云层覆盖条件下的遥感相机推扫成像效率;设计了基于FPGA硬件流水线的算法硬件加速模块,提高了遥感影像在轨处理的速度。3)针对星载云判仪焦面电子学初样模组,设计实现了基于软核微处理器的FPGA固件系统,并制定实验方案对模组的功耗、数据流链路及算法功能进行测试。模组的整体功耗小于12W,逻辑时序裕量达20%以上,能够实现最快4帧每秒的全画幅图像数据采集,且整个算法的数据处理链路耗时小于100ms。综上所述,在遥感图像数据量急剧膨胀的背景下,本文结合敏捷卫星非沿轨动中成像技术与遥感影像在轨处理技术,设计并实现了具有遥感任务规划能力的星载云判仪系统。作为高分辨率主载荷遥感相机的辅助载荷,使用在轨实时云检测技术提前探测云层分布信息,并使用群智能优化算法对遥感卫星的成像推扫路径进行在轨优化处理,改善了云层覆盖条件下的遥感卫星资源利用效率,具有一定的实际意义与工程应用价值。更多还原

【Abstract】 With the continuous progress of space technology,optical remote sensing satellites continue to develop in the direction of high spatial resolution,high time resolution,hyperspectral and multi-sensor.The rapidly expanding amount of remote sensing image data makes the effective utilization of satellite platform resources become particularly important.However,on average,more than 66%of the earth is covered by clouds throughout the year.The optical remote sensing images covered by clouds are useless and occupy valuable spaceborne storage resources and data downlink bandwidth.Therefore,in view of the waste of resources and the decline of imaging utilization rate of optical remote sensing satellite due to cloud,this thesis proposes a spaceborne cloud detection system with remote sensing mission planning capability.Based on cloud distribution information obtained in advance by the real-time cloud detection technology,the intelligent task decision-making in orbit is realized.Combined with the non-along track imaging ability of the agile satellite in motion,it provides the high-resolution remote sensing camera with an optimized imaging track,which improves the resource utilization efficiency of remote sensing satellite under the condition of cloud cover.The specific research contents of this thesis are as follows:1)Based on the application requirements of real-time cloud detection and remote sensing mission planning in orbit,a forward-looking auxiliary camera for a high-resolution multispectral remote sensing camera with a spatial resolution of 30m and an imaging swath of 240km was designed.It forms a 5°installation angle with the high-resolution remote sensing camera and can be activated 6 seconds before the target imaging area for real-time cloud detection and imaging track optimization;Designed a hardware solution based on Bayer array image sensor and FPGA and implemented the interface logic for high-speed image data acquisition and Camera Link Medium data transmission output.2)Based on the hardware platform of the spaceborne cloud detector,a real-time cloud detection algorithm based on multi-dimensional features and minimum distance classifier was designed,which achieved joint classification of spectral and texture features of cloud images.The accuracy of the cloud detection algorithm reached 90%;Designed a remote sensing imaging track optimization strategy,using three strategies of lateral swing imaging,non-along track imaging,and camera shutdown to optimize the mission of the remote sensing camera,improving the efficiency of remote sensing camera imaging under conditions of cloud cover;Designed an algorithm hardware acceleration module based on FPGA hardware pipeline,which improves the speed of in orbit processing of remote sensing images.3)An FPGA firmware system based on a soft-core microprocessor was designed and implemented for the hardware prototype module of the spaceborne cloud detector.An experimental plan was developed to test the power consumption,data flow link,and algorithm functions of the module.The overall power consumption of the module is less than 12w,with a logic timing margin of over 20%.It can achieve full frame image data acquisition of up to 4 frames per second,and the data processing link of the entire algorithm takes less than 100ms.In summary,against the backdrop of the rapid expansion of remote sensing image data,this thesis combines agile satellite non-along track imaging technology and remote sensing image in orbit processing technology to design and implement a satellite cloud detection system with remote sensing mission planning capabilities.As an auxiliary payload for high-resolution remote sensing cameras,real-time cloud detection technology in orbit is used to detect cloud distribution information in advance,and swarm intelligence optimization algorithms are used to optimize the imaging track of remote sensing satellites in orbit,improving the resource utilization efficiency of remote sensing satellites under cloud cover conditions.This has certain practical significance and engineering application value.更多还原