朱雨涵

针对森林资源调查中碳汇计量与生态研究对树种精准分类的迫切需求，本研究提出一种基于智能手机影像的低成本、高效率林木三维重建与单木分割方法，以实现精准树种分类。传统树种分类多依赖激光雷达等专业设备或人工目视识别，存在设备成本高、操作复杂、效率低及主观性强的局限。本研究通过规划林区采集路径，由智能手机平稳采集视频序列，经抽帧与运动重建生成林木初始点云；针对点云噪声冗余、密度不均及地面点干扰问题，设计系统化预处理流程，集成改进的降采样、噪声去除、点云补全与地面点滤除算法，结合点云聚类优化，最终实现复杂林分环境下单木点云精准自动分割，进而完成高效精准的树种分类。

手机影像；林木三维重建；点云聚类；树种分类

[1] XU Y,QIU Q.Challenges and Prospects of Forest Biological Resource Transformation under the Dual-Carbon Policy Framework[J].Journal of Bioresources and Bioproducts,2026:100231. [2] LIAN Y,FENG Z,HUAI Y,et al.Terrestrial Videogrammetry for Deriving Key Forest Inventory Data:A Case Study in Plantation[J].Remote Sensing,2021,13(16):3138. [3] LI J,WANG J,CHEN J,et al.Towards robust point cloud recognition with sample-adaptive auto-augmentation[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2025. [4] LIU C,GU M,HUANG H,et al.Hyperspectral images 3D reconstruction based on structure-from-motion and multi-view stereo[J].Optics and Lasers in Engineering,2026,199:109578. [5] ENGEL N,BELAGIANNIS V,DIETMAYER K.Point transformer[J].IEEE access,2021,9:134826-40.

点击下载PDF