基于低秩与稀疏分解的VideoSAR散射关键帧提取

VideoSAR scattering key-frame extraction based on low-rank plus sparse decomposition

视频合成孔径雷达(video synthetic aperture radar,VideoSAR)的超长相干孔径观测使得区域动态信息的快速浏览极其困难。为以机器视觉方式自动捕捉地物散射消失-瞬态持续-消失-瞬态持续-消失的关键帧变化全过程,提出了一种子孔径能量梯度(subaperture energy gradient,SEG)和低秩与稀疏分解(low-rank plus sparse decomposition,LRSD)相结合的VideoSAR关键帧提取器。提取器为系列性通用架构,适用于任何SEG和LRSD系列方法相结合的形式。所提技术首要针对同时单通道、单波段、单航迹等有限信息条件的解决途径,有助于打破应急响应场景中难以采集多通道、多波段、多航迹或多传感器数据的应用局限性。基于实测数据处理和多种先进LRSD算法进行了对比验证,其代表性散射信息的充分提取可促进未来快速地理解并浓缩区域动态。

The super long coherent aperture observation of Video Synthetic Aperture Radar(VideoSAR)makes it extremely difficult to quickly browse the regional dynamic information.In order to automatically capture the whole process of key-frame change of ground object“disappearance to transient-persistence to disappearance to transient-persistence to disappearance”by machine vision,a VideoSAR key frame extractor via combining subaperture energy gradient(SEG)and Low-Rank plus Sparse Decomposition(LRSD)is proposed.The extractor is a serial general architecture,which is applicable to any combination of SEG and LRSD series methods.The proposed technique mainly provides solutions to limited information conditions on simultaneously single-channel,single-band and single-pass configuration,which can be conducive to break the application limitation of emergency response not permitting the acquisitions of multi-channel,multi-band,multi-pass,or multi-sensor data.Based on the processing of measured data and various advanced LRSD algorithms,the comparative verifications are performed.The thorough extraction of representative scattering information can promote the rapid understanding and summarization of regional dynamics in the future.