利用无人机多光谱影像提取树冠信息

树冠作为树木主要组成部分之一,是树木长势监测、树种识别等内容的重要参数,对森林资源调查和生态研究具有重要意义。与传统的实地调查相比,运用无人机遥感技术提取树冠信息具有高效、便捷等优势。本文基于无人机多光谱影像提取树冠信息,在树冠点探测上结合局部最大值法与Mean Shift优化策略,较原始局部最大法探测精度提升约10%。此外,提出了一种新的树冠边界提取算法,运用动态规划思想进行全局最优边界提取。与以往分水岭分割算法相比,本文算法在较密集林区和稀疏林区均有更好的提取效果,在试验样区稀疏林区F测度提升12%,较密集区F测度提升28%。

Tree canopy delineation using UAV multispectral imagery

As one of the main components of trees,the canopy is an important parameter for tree growth and tree species identification, which is of great significance to forest resource survey and ecological research. Compared with traditional field surveys, UAV remote sensing technology is more efficient and convenient. This paper is based on UAV multi-spectral image for canopy extraction. We use local maximum algorithm and Mean Shift optimization for tree detection, whose detection accuracy is about 10% higher than the local maximum method. In addition, we design a new tree canopy delineation algorithm,which use dynamic programming algorithm to extract the global optimal boundary. Compared with the watershed segmentation algorithm, the proposed method has better results in sparse or denser forest. The F-score is increased by 12% in the sparse, and the F-score is increased by 28% in the desne.