基于无人机可见光影像和面向对象的树种分类研究

针对目前林业部门人工调查树种存在效率低、成本高等问题，采用无人机遥感技术进行树种分类识别，提高树种调查效率，辅助林业管理部门进行林木种植结构分析、病虫害防治等工作。利用无人机获取矮冬青、三球悬铃木、马尾松和紫叶李的冠层红绿蓝（red-green-blue,RGB）可见光影像，进行数字表面模型（digital surface model，DSM）特征图像提取，通过色彩空间转换提高树种间颜色差异；应用最优尺度分割，以纹理特征、颜色特征及几何特征为分类特征参数，优选最佳分类特征集，以期实现无人机可见光影像的树种分类。结果表明，DSM与RGB特征融合图像提取树种的精度较高，可见光影像分类总精度为91.58%，Kappa系数为0.89；特征融合图像分类总精度为98.27%，Kappa系数为0.98。研究提出的特征融合图像结合面向对象分类方法实现了可见光影像的树种分类，为实现树种计数、统计、分类提供数据参考。

Classifications of Tree Species Based on UAV's Visible Light Images and Object-Oriented Method

Aiming at solving the problems of low efficiency and high cost in the artificial investigation of tree species in the forestry sector,the remote sensing technology of unmanned aerial vehicle(UAV)was used to classify and identify tree species to improve the investigation efficiency and to assist the forest management department in the analysis of tree planting structure,pest control,etc.Red-green-blue(RGB)visible light images of the canopy of Ilex lohfauensis,Platanus orientalis,Pinus massoniana,and Prunus cerasifera were obtained by UAV,and feature images of digital surface model(DSM)were extracted,the color differences among tree species were improved by color space conversion.The best classification feature set was optimized by using optimal scale segmentation and taking texture feature,color feature and geometric feature as classification feature parameters to realize the classification of tree species in UAV visible light images.The results showed that the accuracy of extracting tree species from DSM and RGB feature fusion images was high.The classification accuracy of the visible image achieved 91.58%,and the Kappa coefficient was 0.89;the total accuracy of feature fusion image classification was 98.27%,and the Kappa coefficient was 0.98.The feature fusion image was combined with the object-oriented classification method to realize the classification of tree species in the visible light image,which provided data reference for tree species counting,statistics and classification.