• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Symposium on the interface >Estimating Map Accuracy without a Spatially Representative Training Sample
【24h】

Estimating Map Accuracy without a Spatially Representative Training Sample

机译：在没有空间代表训练样本的情况下估算地图精度

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

A land cover map is constructed by partitioning a geographic area of interest into a finite set of map units and assigning a land cover class label to each unit. Land cover maps covering millions of acres consisting of millions of units are often constructed from satellite remotely-sensed data. A classification rule is constructed from a training sample of ground-truthed map units. Because of the expense of collecting a spatially representative training sample for such a large map, the training sample is often drawn from a variety of existing data collected for purposes other than mapping land cover. The spatial distribution of the training sample tends therefore to be highly irregular. It is crucial to estimate the accuracy of the resulting map both overall and on a smaller scale since accuracy may vary spatially and by land cover type. Traditional methods of assessing accuracy, such as cross-validation, may be biased because of the spatial irregularity of the training sample if the classification rule uses spatial information. To reduce bias, we suggest methods of estimating overall map accuracy and unit-by-unit accuracy by using calibrated estimates of the posterior probability of correct classification for each map unit.
机译：通过将感兴趣的地理区域分配成有限的地图单元并将陆地覆盖类标签分配给每个单元来构建陆地覆盖图。覆盖数百万英亩的土地覆盖地图包括数百万个单位通常由卫星远程感测数据构建。分类规则由地面判处地图单元的训练样本构建。由于为这么大地图收集空间代表性的训练样本的费用，训练样本通常从用于除映射陆盖以外的目的收集的各种现有数据中。因此，训练样品的空间分布趋于高度不规则。重要的是估计总体和较小的尺度的所得到的地图的准确性，因为精度可以在空间上和陆地覆盖类型变化。评估准确性的传统方法，例如交叉验证，可能是由于训练示例的空间不规则如果分类规则使用空间信息而偏置。为了减少偏差，我们建议通过使用每个地图单元的正确分类的后验概率的校准估计来估计整体地图精度和单位单位精度的方法。

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