基于Canny算子与决策树模型的水体信息提取研究

Extracting Water Body Information Based on Canny Algorithm and Decision Tree

选用白山市浑江区曲家营水库及其周围区域的TM影像作为基础数据。首先,对其进行主成分变换后再进行彩色合成处理,结合Canny算子的理论,对水体边缘进行了提取;第二,在参考了水体波谱特征以及前人对水体影像特征研究成果的基础上,利用ENVI 5.1软件建立了决策树水体遥感解译模型,进行了水体信息提取,经过栅矢数据转换,选用ArcGIS 10.1软件对碎斑进行处理,将其与边缘监测结果进行图像融合。通过利用Canny算子对彩色合成影像进行边缘提取能克服传统灰色图像的缺点,将其与决策树水体解译模型结果进行图像融合能达到较理想的效果。

The paper took the TM images of Qujiaying reservoir of Hunjiang district, Baishan City and its surrounding area as basic data. Firstly, extracting the edge of water - body by using the Canny algorithm theory after the principal component transform and color synthesis. Secondly, establishing decision tree to extract water information based on the spectral characteristics and other existing re- searches of water body by using the EnviS. 1 software. To chosen ArcGIS10.1 softerware processed porphyroclastic texture data after Raster - Vector transformation, and then carried out image fusion with the edge detection results . Experiment shows that the edge ex- traction of water body after color composite images can be avoided by using the Canny operator, and the results of image fusion can a- chieve the better results.