树种分类是林业遥感中重要的应用领域,在林业可持续管理、生物多样性研究和生态环境保护领域都有广发的应用场景。结合2000年后该领域的研究成果以及近些年林业生产过程中的应用实践,对多源数据在树种分类的应用加以汇总。从工作流程及相应算法两个角度,基于图像分类与数理统计层面对该问题进行了解析与比较。面对林业遥感在树种分类应用过程中遇到的问题与挑战,提出了基于语义分割及实例分割的不同工作思路,对未来多源遥感数据融合获取及硬件处理设备的发展前景进行了展望。

Tree species classification is an important application area in forestry remote sensing.It has widely used application scenarios in the areas of sustainable forestry management,biodiversity research,and ecological environmental protection.Based on the research results in this field after 2000 and the application practice in the forestry production process in recent years,the application of multi-source data in tree classification is summarized.From the perspectives of workflow and corresponding algorithms,the issue is analyzed and compared based on image classification and mathematical statistics.Facing the problems and challenges encountered in the application of forestry remote sensing in the classification of tree species,different working ideas based on semantic segmentation and instance segmentation were proposed,and the future development of multi-source remote sensing data fusion acquisition and hardware processing equipment was prospected.