摘要
针对月面巡视探测器自主导航中的障碍识别问题,为了采集清晰有效的图形,使用一种基于图像增强的方法结合尺度不变的特征点提取和匹配新的算法来识别障碍物。使用图像增强方法预处理图像,用多尺度特征极值点检测的SIFT方法,提取特征,进行左右双目图像的特征匹配。与传统视觉算法相比,可以解决仿真试验场较差的光源环境,并提高对不同光照环境图像的特征提取和匹配的鲁棒性。在仿真试验场的双目视觉图像匹配中,仿真实验取得较好的效果。
Presenting a new feature points extracting and matching algorithm based on image enhancement for the obstacle recognition in navigating systems of lunar rover. We enhanced original simulative lunar image at first and then used Scale Invariant Feature Transform(SIFT) to get the feature for matching. This method can solve the weak lighting condition of simulative experimental enviorment and improve the stability for lightness invariant of environment and detection accuracy in the later matching phase compared with traditional methods. We use this algorithm in simulative environment experiments and have obtained good result.
出处
《计算机仿真》
CSCD
北大核心
2010年第5期27-30,共4页
Computer Simulation
关键词
图像增强
特征提取
特征匹配
Image enhancement
Feature extracting
Feature matching