Comparative Analysis of Novel View Synthesis and Photogrammetry for 3D Forest Stand Reconstruction and extraction of individual tree parameters

Guoji Tian,Chongcheng Chen,Hongyu Huang

2024-10-08

Abstract:Accurate and efficient 3D reconstruction of trees is crucial for forest resource assessments and management. Close-Range Photogrammetry (CRP) is commonly used for reconstructing forest scenes but faces challenges like low efficiency and poor quality. Recently, Novel View Synthesis (NVS) technologies, including Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS), have shown promise for 3D plant reconstruction with limited images. However, existing research mainly focuses on small plants in orchards or individual trees, leaving uncertainty regarding their application in larger, complex forest stands. In this study, we collected sequential images of forest plots with varying complexity and performed dense reconstruction using NeRF and 3DGS. The resulting point clouds were compared with those from photogrammetry and laser scanning. Results indicate that NVS methods significantly enhance reconstruction efficiency. Photogrammetry struggles with complex stands, leading to point clouds with excessive canopy noise and incorrectly reconstructed trees, such as duplicated trunks. NeRF, while better for canopy regions, may produce errors in ground areas with limited views. The 3DGS method generates sparser point clouds, particularly in trunk areas, affecting diameter at breast height (DBH) accuracy. All three methods can extract tree height information, with NeRF yielding the highest accuracy; however, photogrammetry remains superior for DBH accuracy. These findings suggest that NVS methods have significant potential for 3D reconstruction of forest stands, offering valuable support for complex forest resource inventory and visualization tasks.

Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to perform more accurate and efficient 3D reconstruction of trees in forest resource assessment and management, and extract the parameters of individual trees from it. Specifically, the paper explores the application effects of new view synthesis techniques (such as Neural Radiance Field (NeRF) and 3D Gaussian Point Cloud (3DGS)) and traditional photogrammetric methods in complex forest environments. The research aims to compare the advantages and disadvantages of these methods in terms of reconstruction efficiency, point cloud density, and accuracy of tree parameter extraction, in order to provide new technical support for complex forest resource surveys and visualization tasks.

### Main research questions:
1. **Compare the practical application effects of new view synthesis techniques (NVS) and traditional photogrammetric methods in complex forest environments**.
2. **Evaluate the abilities of different NVS methods (NeRF based on implicit neural networks and 3DGS based on explicit Gaussian point clouds) in tree reconstruction and generating dense point clouds**.
3. **Compare the tree parameters extracted from different 3D point cloud models, and evaluate whether NVS techniques can replace or supplement traditional photogrammetric methods and become new tools for forest scene reconstruction and forest resource surveys**.

### Research background:
- **Traditional methods**: Traditional methods for obtaining tree parameters rely on manual measurements, such as using calipers or diameter tapes to measure the diameter at breast height (DBH), and using altimeters to measure the tree height (TH). This method is time - consuming and labor - intensive, and is easily affected by human factors.
- **3D reconstruction techniques**: 3D reconstruction techniques can automatically extract the parameters of individual trees by converting forest scenes into 3D digital models, overcoming some limitations of traditional survey methods. For example, terrestrial laser scanning (TLS) and airborne laser scanning (ALS) can obtain high - density 3D point cloud models, but the equipment is expensive and the operation is complex.
- **Close - range photogrammetry (CRP)**: The CRP method reconstructs 3D point clouds through overlapping images. It is low - cost and easy to operate, but there are problems such as occlusion and illumination changes, which affect the reconstruction quality.
- **New view synthesis techniques (NVS)**: NVS techniques, especially NeRF and 3DGS, have made significant progress in the field of computer vision in recent years, and can achieve high - quality 3D reconstruction with a limited number of images. However, their applications in complex forest environments still need further verification.

### Research objectives:
1. **Compare the application effects of NVS techniques and traditional photogrammetric methods in complex forest environments**.
2. **Evaluate the abilities of NeRF and 3DGS in tree reconstruction and generating dense point clouds**.
3. **Compare the tree parameters extracted from different 3D point cloud models and evaluate the practical value of NVS techniques**.

Through these studies, the paper aims to provide new technical means for efficient and accurate surveys of complex forest resources.