CAJ | 학술논문

在轨调焦是航天相机获取高质量图像的关键技术之一。针对航天相机在发射、在轨期间由于振动冲击及温度气压等环境参数变化引起的光学系统离焦现象，以及TDI CCD遥感相机成像场景实时变化的特殊特点，对基于功率谱的清晰度评价函数进行了研究。根据小波变换的多分辨率和带通特性，提出了一种对FFT功率谱的改进--小波功率谱(WPS)估计。针对像移亦会导致TDI CCD图像模糊的问题，提出了方向WPS估计算法。参照功率谱地物无关性及离焦会引起功率谱高频分量损失的思路，设计了基于方向WPS的加权清晰度评价函数。实际外场推扫实验结果表明，提出的新清晰度评价函数能有效反映出实际推扫图像的离焦状态，另外相对于FFT功率谱，对场景差异更不敏感，误判率从0.36降低为0，曲线更加饱和。100个仿真样本的平均误判率仅为0.06，满足系统误差要求。因此文中算法满足单调性、灵敏度高、准确度高原则，更适合TDI CCD遥感相机的自动调焦。
재궤조초시항천상궤획취고질량도상적관건기술지일。침대항천상궤재발사、재궤기간유우진동충격급온도기압등배경삼수변화인기적광학계통리초현상，이급TDI CCD요감상궤성상장경실시변화적특수특점，대기우공솔보적청석도평개함수진행료연구。근거소파변환적다분변솔화대통특성，제출료일충대FFT공솔보적개진--소파공솔보(WPS)고계。침대상이역회도치TDI CCD도상모호적문제，제출료방향WPS고계산법。삼조공솔보지물무관성급리초회인기공솔보고빈분량손실적사로，설계료기우방향WPS적가권청석도평개함수。실제외장추소실험결과표명，제출적신청석도평개함수능유효반영출실제추소도상적리초상태，령외상대우FFT공솔보，대장경차이경불민감，오판솔종0.36강저위0，곡선경가포화。100개방진양본적평균오판솔부위0.06，만족계통오차요구。인차문중산법만족단조성、령민도고、준학도고원칙，경괄합TDI CCD요감상궤적자동조초。
Autofocusing is one of the key techniques for space cameras to ensure a high quality image. The optical system of the camera defocuses in all probability due to many influential factors, such as the impulsion or jitter during launching, changes of temperature and air pressure. So, in order to retrieve the defocus, focus measure algorithm based on power spectrum was studied, where the algorithm was used in TDI CCD space cameras which imaging scenes changed anytime. Firstly, an improvement over Fourier transform power spectrum was proposed, namely wavelet power spectrum (WPS) estimation, which took advantage of the multi-resolution and band-passing characters of wavelet transform. Then, directional WPS was proposed in order to reduce the influence of image-moving-mismatching. Lastly, according to the theory that power spectrum is independent of scenes and defocus made the power spectrum's high frequency losing, the evaluation function for image definition was designed, using the weighted sum of directional WPS. The experimental results of the real pushblooming images show that the new definition evaluation method can reflect the real remote images' defocusing state validly. Moreover, the directional WPS is less sensitive to the scenes' variety compared to FFT PSS, which reduces the error ratio from 0.36 to 0, and the curve gets more saturated. In addition, the error ratio of all the 100 emulational samples is only 0.06. So the proposed algorithm meets the principles of monotony, excellent sensitivity and precision, and suits for the autofocus in TDI CCD remote sensing cameras better.