基于P-N学习的高分遥感影像道路半自动提取方法

基于模板匹配的道路跟踪是半自动提取道路的主要方法。然而场景中地物干扰和道路宽度的变化降低了模板匹配的稳定性；另外，道路跟踪失败后缺乏重检测机制，使得道路提取过程中人机交互频繁。针对以上问题，提出了一种基于P-N（positive-negative）学习的高分遥感影像道路半自动提取方法。该方法由道路跟踪、检测和学习构成，关键是采用了P-N学习的策略迭代的训练分类器，通过纠正违反结构约束的样本分类结果来提高分类器性能。实验使用了不同场景下的城区高分遥感影像，与经典的模板匹配和在线学习的道路跟踪方法进行了比较。实验结果表明该方法在道路提取的精度和稳定性方面均有提升。

Semi-automatic Road Extraction Method from High Resolution Remote Sensing Images Based on P-N Learning

The road tracking method based on template matching is one major semi-automatic road extraction method. However, template matching is sensitive to complexity of road scenes and variance in road width. In addition, road extraction requires frequent human-computer interaction while road tracking encounters failure without a mechanism for re-detection. To solve these problems, one semi-automatic road extraction method using high resolution remote sensing image based on P-N learning is proposed. It consists of road tracking, detecting and learning. In order to improve the stability of road detection, we train a classifier with an iterative P-N learning strategy. The performance of classifier is improved by correcting sample labeling under structural constraints. In experiments, the proposed method and three classical methods are tested on high-resolution remote sensing images of different scenes. Comparitive results show proposed method' improves precision and stability of road extraction.