论文部分内容阅读
基于月貌图像的撞击坑的检测需要采用合理的特征选择和机器学习策略,我们提出了一种基于区域局部灰度和梯度分布特征与机器学习方法相结合的撞击坑检测方法.这种方法将Haar特征与AdaBoost结合,使候选撞击坑区域的定位更加快捷,采用局部区域的塔式梯度方向直方图(PHOG)与高效的支持向量机学习算法相结合的方法用来精确地对撞击坑候选区域进行分类.考虑到Haar特征数的繁多而采用AdaBoost作为特征提取和分类方法,并由于PHOG特征的每一项都对分类起作用,将撞击坑区域统一预处理为不含阴阳面的各向梯度向量基本一致的圆形模糊边界,使圆形撞击坑的正样本特征具备更多的稳定性.文中还讨论了几种特征和分类方法的机理和集成,以及参数调整对撞击坑检测的效率分析.
The detection of craters based on lunar images requires reasonable feature selection and machine learning strategies, and we propose a craters detection method based on the combination of regional gray and gradient distribution features with machine learning methods The combination of Haar feature and AdaBoost can make the location of the candidate craters more rapid. The method of combining the local region gradient pyramid histogram (PHOG) and efficient support vector machine learning algorithm is used to accurately predict the location of the craters candidate region , AdaBoost was adopted as the feature extraction and classification method considering the variety of Haar features and each of the PHOG features was used to classify the impact pits into a uniform gradient without anisotropy The vectors have the same circular fuzzy boundary, which makes the positive sample features of circular craters have more stability. The paper also discusses the mechanism and integration of several features and classification methods, as well as the efficiency analysis of craters detection by parameter adjustment .