基于单像素成像体制的制导信息提取方法研究

Precision guidance ensures that the aircraft hits a target and key parts of the body with astounding accuracy in the complex battlefield environment. The core of precise guidance is the precisely seeker subsystem and control subsystem. The guidance information is the key technology of precise guide system, and is outputed to control system as an instruction by seeker system. Therefore, aim at guidance information extraction which is the problem to be solved urgently in infrared imaging guidance, this project will focuse on the extraction of guidance information based on the single-pixel imaging system. Firstly, we will study the sparse representation of the infrared image, and design the strategy of sampling the sparse signal fastly from the infrared image according to the measurement matrix. The sensing sparse signals are relatively few, and capture most of the target information which is important for the recognition.Then we will put an emphasis on target recoginition and guidance information extraction from the sensing relatively few sparse signal. The process of recoginition and extraction is similar to the process of signal sampling. The way of "sampling sparse signal - extracting useful information" will break out the traditional way of "sampling the whole image- processing image- extracting guidance information". The target recognition and guidance information extraction will be realized synchronozing at the stage of signal detection and sampling. The proposed new guidance system in the project will obtain and sense the effective information of the infrared image according to its sparsity, and the sensing stage will acquire and process the highly importance image imformation. Compared with the traditional focal plane array infrared imaging guidance system, the new one will be advantage at overcoming the non-uniformity, high propotion of useful information, fastly processing speed. In the foreseeable future, the main contributions and innovation points of this project will play the important role of presenting an theoretical significance and application value for developing new intelligent imaging guidance system and guidance information extraction.

精确制导能确保飞行器在复杂战场环境中精确命中目标乃至要害部位，其核心是精确导引和控制，而制导信息是导引系统给控制系统的指令，是实现精确制导的关键。因此本项目针对红外成像制导领域亟待解决的制导信息提取问题，研究单像素成像体制下的制导信息提取方法。研究红外图像的稀疏表示与观测矩阵，设计快速采样策略，采集含目标重要信息的稀疏信号；从采样的少量信号中以信号分析的角度进行目标识别与制导信息提取。这种"采样稀疏信号－提取有用信息"的方式突破了焦平面成像制导"采样整幅图像－图像处理－提取制导信息"的传统方式，在探测与采样阶段同步实现目标识别与制导信息提取。所提出的新型制导体系根据图像稀疏性有针对性地获取目标有效信息，增加重要性高的图像信息的采集与处理，能够克服传统焦平面阵器件的非均匀性，具有有用信息比例高、处理速度快等优势；对发展新型智能探测成像体制和制导信息提取方法具有重要的理论意义与应用参考价值。

本项目针对单像素成像体制下的目标识别与制导信息提取等关键技术开展了研究。项目的研究首先将单像素成像体制作为具体的分析对象，分析与研究了红外图像的稀疏表示，其目的是用尽可能少的非零系数来表示图像信号的主要信息；其次研究了观测矩阵的设计，观测矩阵在获取测量向量和信号重建的过程中起着关键性作用，设计高效的观测矩阵就是设计一个能捕捉稀疏信号中有用信息的观测协议，进而把该稀疏信号压缩成少量数据，因此测量向量的获取并不是通过直接采集稀疏信号，而是通过观测底层信号（underlying signal）的线性预测。通过理论分析研究设计了单像素成像的光学扫描算法，为单像素成像制导走向工程化提供了理论依据。在单像素成像体制下，借鉴传统成像体制下模版匹配的思想，以历史帧数据已识别的目标作为测试样本，用当前帧采样数据所代表的图像信息形成训练样本，设计相应的目标识别方法，以区分目标与干扰，同时设计相应的目标跟踪与目标搜索策略，实现边识别边跟踪的目标识别与跟踪策略。针对单像素成像体制下的制导信息提取问题，主要研究了从稀疏采样信号中提取目标姿态信息的方法，首先建立了目标的姿态数据库，在当前帧目标发生姿态变化时，以当前帧图像数据为测试样本，姿态数据库作为训练样本，通过特征匹配确定目标的姿态，完成姿态制导信息的提取。通过大量的仿真实验，验证了本项目研究的有效性，并为后续的研究提出了相关建议。本项目从采样的少量信号中以信号分析的角度进行目标识别与信息提取，以“采样稀疏信号－提取有用信息”的方式突破了焦平面成像制导“采样整幅图像－图像处理－提取制导信息”的传统方式，在探测与采样阶段同步实现目标识别与制导信息提取，本项目研究成果为发展新型智能探测成像体制和制导信息提取方法具有重要的理论意义与应用参考价值。