基于三种分割算法的高分辨率影像分割比较

Comparison of High Resolution Image Segmentation Based on Three Segmentation Algorithms

随着遥感影像分辨率的不断提高,基于高分辨率影像采用图像分割技术获取单木树冠成为一种重要方法。为了获取单木树冠,选取林区高分辨率的无人机影像为数据源,采用迭代式阈值法、最大类间方差法和分水岭分割法三种常见的阈值分割方法进行分割对比研究,并从图像处理时间和单个树冠匹配率两个方面进行精度评价,结果表明:分水岭分割法在处理时间上优于迭代式阈值分割法,在单个树冠匹配率上也显著高于其他两种方法;综合两个方面的评价指标表明分水岭的分割效果更好。

With the continuous improvement of remote sensing image resolution,it becomes an important method to acquire single canopy with image segmentation technology based on high resolution image. In order to obtain single canopy,this article selects in high resolution of unmanned aerial vehicle（ UAV） images as the data source,the iterative threshold method,the maximum variance method and the watershed segmentation method would be used for segmentation and comparative analysis. The accuracy was evaluated from image processing time and single canopy matching rate. The results show that the watershed segmentation method is superior to the iterative threshold segmentation method in processing time. The watershed segmentation method is significantly higher than the other two methods in single canopy matching rate. The evaluation indexes of the two aspects show that the segmentation effect of the watershed method is better.