一种基于太赫兹成像的复杂地形自适应定高方法

An adaptive height measuring method in complex terrain based on terahertz imaging

摘要:

针对现有近感探测器难以适应复杂地形情况下的精确定高问题，提出一种基于成像的图像域自适应对地高度估计方法。该方法将多普勒锐化成像原理应用到前下视场景，利用太赫兹频段分辨力高、孔径合成长度短的优势，采用太赫兹前下视成像实现地形感知，将传统近感探测器的一维距离测量扩展为二维成像测量，实现对运动轨迹前下方地形地貌的精确感知。在获得地面目标图像后，利用地面图像特征，采用图像域特征拟合方法实现对地高度估计。对前下视多普勒锐化成像的参数设计和成像算法进行了理论分析和仿真验证，开发了一套工作于220 GHz频段的太赫兹近感探测系统，并开展了无人机挂载试验。试验表明，该方法能有效感知地表目标环境，排除角反、树木等其他目标对地面高度测量的干扰，测高精确度达0.5 m，证明了该基于太赫兹成像的对地高度估计方法的可行性，为基于图像域特征提升近感探测器对复杂地形适应能力和感知能力奠定了工作基础。

Abstract:

In response to the issue of precise altitude measurement in complex terrains where existing proximity sensors struggle to adapt, a novel imaging-based adaptive ground altitude estimation method in the image domain is proposed. This method applies the principle of Doppler sharpening imaging to the forward-looking view scene, taking advantage of the high resolution in the terahertz frequency band and the short synthetic aperture length. It utilizes terahertz forward-looking imaging for terrain perception, transforming the one-dimensional distance measurement of traditional proximity sensors into two-dimensional imaging measurement, thus achieving precise perception of the terrain and topography directly below the motion trajectory. After obtaining the ground target image, the altitude estimation is realized by fitting image domain features using the characteristics of the ground image. Theoretical analysis and simulation verification were conducted on the parameter design and imaging algorithm of the forward-looking Doppler sharpening imaging. A terahertz proximity detection system operating at 220 GHz frequency band was developed and unmanned aerial vehicle-mounted tests were carried out. The tests show that this method can effectively perceive the ground environment, eliminate interference from corner reflectors, trees, and other targets on ground altitude measurement, and achieve an altitude measurement accuracy of 0.5 m. This demonstrates the feasibility of the terahertz imaging-based ground altitude estimation method and lays a foundation for improving the adaptability and perception capability of proximity sensors to complex terrains based on image domain features.