CAJ | 학술논문

在无人机着舰的合作目标分割问题中，当目标占图像比例小于0．2％时，双Otsu法、最大熵法等经典的图像分割方法效果不佳。为了提高分割的效果，本文提出了自适应的区域生长法。首先，根据机载摄像机与着陆合作目标的估计距离，确定合作目标的像素点数Ⅳ；然后利用红外合作目标的特点，自适应地寻找出图像中Ⅳ个最亮点的几何中心作为生长核；最后根据灰度直方图的统计特性，确定生长终止条件。实验结果表明，本文提出的方法分割效果明显优于现有常用的双Otsu法、最大熵法等方法。
재무인궤착함적합작목표분할문제중，당목표점도상비례소우0．2％시，쌍Otsu법、최대적법등경전적도상분할방법효과불가。위료제고분할적효과，본문제출료자괄응적구역생장법。수선，근거궤재섭상궤여착륙합작목표적고계거리，학정합작목표적상소점수Ⅳ；연후이용홍외합작목표적특점，자괄응지심조출도상중Ⅳ개최량점적궤하중심작위생장핵；최후근거회도직방도적통계특성，학정생장종지조건。실험결과표명，본문제출적방법분할효과명현우우현유상용적쌍Otsu법、최대적법등방법。
In the problem of cooperative target segmentation for the landing of unmanned aerial vehicle （UAV）, the segmentation results of the double Otsu algorithm and maximum entropy algorithm are unsatisfactory when the target accounts for the image less than 0.2%. In order to promote the effects, an adaptive region growing method is proposed. Firstly, the pixel number of the cooperative target is set through the estimated distance between airborne camera and the landing target. Secondly, N brightest points of the image are found out automatically utilizing the characteristics of IR images. Then, the geo- metric center of the N brightest points is set to be the seed point. Finally, the terminal principle is determined by the statistical properties of histogram. The experimental results show that this method is better than the common segmentation methods, such as double Otsu and maximum entropy method.