结合航空影像纹理和光谱特征的单木冠幅提取
Extracting individual tree crown by combining spectral and texture features from aerial images
-
摘要: 随着航空摄影测量技术的不断发展与进步,为提高森林资源调查的工作效率,航空影像已经成功应用到林业资源监测中,但在单木冠幅提取上,研究多考虑影像光谱信息,使得分类结果存在偏差。本文提出同时结合航空影像的纹理及光谱特征,利用面向对象的影像分割方法,通过多次实验对比结果确定最优分割尺度,同时在结合正态分布法确定各光谱及纹理特征信息范围的基础上,提取单木冠幅信息。以2012年鹫峰国家森林公园航空像片为数据源,以ENVI5.0为数据处理平台,对影像进行面向对象的分割,提取试验区域内32株树木的冠幅,并结合传统外业实测数据以及立体像对观测数据进行精度分析。试验结果表明:文章所提出的方法试验精度达到90.05%,与传统立体像对量测方法精度相近,但数据获取速度快,满足林业调查基本需求。Abstract: Aero photogrammetry has been used in forestry successfully for improving the efficiency in forest resources survey with the development and progress of it. However, the results in most studies in single tree crown extraction were biased because the merely image spectral information was considered. Individual tree crown can be extracted by jointly combining the texture and spectral features of the aerial images, then determining the optimal segmentation scale while using object-oriented image segmentation method and multiple comparing the experimental results, and defining the information scope of spectral and texture features with normal distribution method. Using aerial images of Jiufeng National Forest Park of Beijing in 2012 as the data source, and using ENVI5.0 as the data processing platform to segment the images, then extract 32 trees’ crown in the test area, and analyze accuracy combined with the traditional field measured data and stereo image measured data. The results showed that the accuracy could achieve 90.05% by the proposed method, it is close to the accuracy by traditional stereo image measure, and data acquisition is quick, so it meets the basic needs of the forestry survey.