...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Planetary and space science >Automated detection of Martian water ice clouds using Support Vector Machine and simple feature vectors
【24h】

Automated detection of Martian water ice clouds using Support Vector Machine and simple feature vectors

机译：使用支持向量机和简单特征向量自动检测火星水冰云

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

We present a method for evaluating the presence of Martian water ice clouds using difference images and cross correlation distributions calculated from blue band images of the Valles Marineris obtained by the Mars Orbiter Camera onboard the Mars Global Surveyor (MGS/MOC). We derived one subtracted image and one cross correlation distribution from two reflectance images. The difference between the maximum and the average, variance, kurtosis, and skewness of the subtracted image were calculated. Those of the cross-correlation distribution were also calculated. These eight statistics were used as feature vectors for training Support Vector Machine because they were the simplest of features that was expected to be closely associated with the physical properties of water ice clouds. The generalization ability was tested using 10-fold cross-validation. F-measure and accuracy tended to be approximately 0.8 if the maximum in the normalized reflectance and the difference of the maximum and the average in the cross-correlation were selected as features. This result can be physically explained because the blue band as well as the red band is sensitive to water ice clouds. A simple and low dimensional feature vector enables us to understand the detected water ice clouds physically and presents the lower bound of the score that classifiers trained using more sophisticated feature vectors have to achieve.
机译：我们提出了一种方法，用于使用火星全球测量师（MGS / MOC）上的火星轨道摄像机获得的，由Valles Marineris蓝带图像计算的差异图像和互相关分布来评估火星水冰云的存在。我们从两个反射率图像中导出了一个减法图像和一个互相关分布。计算出相减图像的最大值和平均值，方差，峰度和偏度之间的差异。还计算了互相关分布的那些。这八个统计量被用作训练支持向量机的特征向量，因为它们是最简单的特征，被认为与水冰云的物理特性密切相关。使用10倍交叉验证测试泛化能力。如果选择归一化反射率的最大值和互相关的最大值与平均值之差作为特征，则F量度和精度往往接近0.8。由于蓝带和红带对水冰云敏感，因此可以从物理上解释此结果。一个简单的低维特征向量使我们能够物理上理解检测到的水冰云，并呈现使用更复杂的特征向量训练的分类器必须达到的分数下限。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号