多维信息在回路动态像移自配准时间延迟积分成像技术研究

Image motion compensation is the basic guarantee for TDI remote sensing imaging to obtain high quality images. The pursuit of high precision, real time and simple image motion compensation method has always been the goal of the world. Based on the flexible advantages of TDI operation in digital domain, a new idea is proposed in this work, which integrates dynamic image motion compensation with the new TDI algorithm in digital domain. First, the image motion is calculated through the image, which is in line with the latest development trend. Then the dynamic image motion is compensated with the new TDI algorithm in digital domain, eliminating the complex mechanical drift adjusting mechanism, reducing the image motion control accuracy. The project team has carried out preliminary research on TDI in digital domain and the image-based adjusting method with fixed drift angle. On this basis, the project will conduct deep research on the high-precision image acquisition with low brightness and limited area images, and the new TDI algorithm integrated with dynamic image motion. In order to eliminate the influence of low luminance and limited image area, we will improve the feature point image registration algorithm by developing multi-dimensional information to describe the feature local information, and then build a high-accuracy algorithm for image motion calculation. Then, we use vector decomposition and deduction analysis method to reconstruct the TDI algorithm based on the principle of registration pixel accumulation, so that the algorithm can adapt to the dynamic image motion and achieve the time delay integration with automatic image motion compensation.

像移补偿是TDI遥感成像获取高分辨率清晰图像的基本保证。寻求高精度、实时性强而又简单易行的像移补偿方法一直是世界各国不断追求的目标。结合数字域TDI操作灵活的优势，项目提出将像移补偿集成于TDI的新思路，首先基于图像构建像移计算算法，符合了图像式像移测量的最新发展趋势；然后将动态像移补偿与新型数字域TDI算法结合在一起，省去了复杂的机械调偏流机构，降低了像移控制精度要求。项目组在数字域TDI和固定偏流角的图像调偏流方面开展了研究初探，在此基础上本项目将针对低亮度有限区域图像高精度像移量获取和动态变化像移情况下的数字域TDI开展深入研究。拟采用挖掘多维信息用于特征局部信息描述改进点特征图像匹配算法的方法，消除亮度低和图像区域有限的影响，构建高精度像移计算算法。然后采用向量分解和推演解析的方法，依据配准像素累加原则重构TDI算法，使算法能适应动态变化像移情况，实现自动像移补偿的时间延迟积分。

像移补偿是TDI遥感成像获取高分辨率清晰图像的基本保证。寻求高精度、实时性强而又简单易行的像移补偿方法一直是世界各国不断追求的目标。针对低亮度有限区域图像序列的高精度像移计算，本项目首先提出一种新型BM3D算法，实现低亮度遥感图像降噪，为高精度像移计算奠定基础，然后提出一种互功率谱图像配准方法用于计算像移偏差量，在像素级初步估计精度基础上进一步采用二次曲面插值，得到亚像素级最终估计。本项目设计的高精度像移计算方法，不仅适用于真实空间遥感成像仿射变换下像移估计，而且具有理论上的高精度，时间复杂度低，同时对亮度差异和频率相关的噪声具有鲁棒性。针对动态像移时间延迟积分，提出一种像素矩阵修正算法，实现亚像素级图像插值修正，并采用推演解析方法构建了时间延迟积分模型。最后项目开展了计算机仿真实验，并搭建成像实验系统开展了实际推扫成像实验。实验结果表明当96级积分的总像移量大于像元尺寸时，传统TDI方法扫描图像发生严重混叠，成像质量急剧下降，而本项目方法利用匹配像素叠加可有效控制成像质量不会急剧恶化，像移传函和相似相关性测度相较传统TDI方法均至少提高0.11。本研究是对我国现有像移补偿机制的革新，对提高我国的高性能遥感成像水平、推动遥感卫星的智能化发展和促进我国的多模态敏捷成像应用有重要意义。

内容获取失败，请点击重试