Individual tree crown detection and delineation (ITCD) mainly depend on high-resolution aerial photos and satellite images or LiDAR, and these data can be costly to obtain. The advent of unmanned aerial vehicle (UAV) remote sensing technology provides an economic and effective method for data acquisition. Therefore, the research of ITCD based on UAV high-resolution images is of significance to improve the efficiency and accuracy of forest resource inventory and remote sensing validation. However, in very high-resolution (defined here with the pixel size smaller than 10 cm) images, the inhomogeneities in the canopy texture can be detrimental to the correct detection of individual trees by computer processing. We applied the bias field estimation, which is used in medical image segmentation, to reduce the within-canopy spectral heterogeneity in the very high-resolution UAV-derived orthophoto. By selecting young Osmanthus and Podocarpus trees that grow in a nursery as the study objects, we tested our method in an orthophoto (with a ground resolution of 2.5 cm) generated from overlapping UAV images. A local intensity clustering was first applied to produce a smoothed bias field image; then, by using morphological operation of opening and closing, the fine texture of the canopy was further smoothed. Finally, individual tree crowns were extracted by applying the marker-controlled watershed segmentation algorithm. The segmentation results were validated by comparing to the manually drawn individual tree crown polygons. The $F$-scores of the detection rate for Osmanthus and Podocarpus were 98.2% and 93.1%, respectively. The result is considerably better than those achieved with similar processing steps but without the bias field treatment. Our study proves that it is feasible and effective to detect and delineate individual tree crowns based on the bias field and marker-controlled watershed segmentation in very-high-resolution images obtained from UAV.

With the progression of LiDAR (Light Detection and Ranging) towards a mainstream resource management tool, it has become necessary to understand how best to process and analyze the data. While most ground surface identification algorithms remain proprietary and have high purchase costs; a few are openly available, free to use, and are supported by published results. Two of the latter are the multiscale curvature classification and the Boise Center Aerospace Laboratory LiDAR (BCAL) algorithms. This study investigated the accuracy of these two algorithms (and a combination of the two) to create a digital terrain model from a raw LiDAR point cloud in a semi-arid landscape. Accuracy of each algorithm was assessed via comparison with >7,000 high precision survey points stratified across six different cover types. The overall performance of both algorithms differed by only 2%; however, within specific cover types significant differences were observed in accuracy. The results highlight the accuracy of both algorithms across a variety of vegetation types, and ultimately suggest specific scenarios where one approach may outperform the other. Each algorithm produced similar results except in the ceanothus and conifer cover types where BCAL produced lower errors.

Tree height is an important vegetative structural parameter, and its accurate estimation is of significant ecological and commercial value. We collected UAV images of six tree species distributed throughout a subtropical campus during three periods from March to late May, during which some deciduous trees shed all of their leaves and then regrew, while other evergreen trees kept some of their leaves. The UAV imagery was processed by computer vision and photogrammetric software to generate a three-dimensional dense point cloud. Individual tree height information extracted from the dense photogrammetric point cloud was validated against the manually measured reference data. We found that the number of leaves in the canopy affected tree height estimation, especially for deciduous trees. During leaf-off conditions or the early season, when leaves were absent or sparse, it was difficult to reconstruct the 3D canopy structure fully from the UAV images, thus resulting in the underestimation of tree height; the accuracy improved considerably when there were more leaves. For Terminalia mantaly and Ficus virens, the root mean square errors (RMSEs) of tree height estimation reduced from 2.894 and 1.433 m (leaf-off) to 0.729 and 0.597 m (leaf-on), respectively. We provide direct evidence that leaf-on conditions have a positive effect on tree height measurements derived from UAV photogrammetric point clouds. This finding has important implications for forest monitoring, management, and change detection analysis.

Three-dimensional (3D) reconstruction of trees has always been a key task in precision forestry management and research. Due to the complex branch morphological structure of trees themselves and the occlusions from tree stems, branches and foliage, it is difficult to recreate a complete three-dimensional tree model from a two-dimensional image by conventional photogrammetric methods. In this study, based on tree images collected by various cameras in different ways, the Neural Radiance Fields (NeRF) method was used for individual tree dense reconstruction and the exported point cloud models are compared with point clouds derived from photogrammetric reconstruction and laser scanning methods. The results show that the NeRF method performs well in individual tree 3D reconstruction, as it has a higher successful reconstruction rate, better reconstruction in the canopy area and requires less images as input. Compared with the photogrammetric dense reconstruction method, NeRF has significant advantages in reconstruction efficiency and is adaptable to complex scenes, but the generated point cloud tend to be noisy and of low resolution. The accuracy of tree structural parameters (tree height and diameter at breast height) extracted from the photogrammetric point cloud is still higher than those derived from the NeRF point cloud. The results of this study illustrate the great potential of the NeRF method for individual tree reconstruction, and it provides new ideas and research directions for 3D reconstruction and visualization of complex forest scenes.

Plant structural parameters are important for ecological studies and for monitoring the environment. Terrestrial laser scanning has become a widely accepted technique for acquiring accurate high‐density three‐dimensional information about plant surfaces; however, this instrument is expensive, technically challenging to operate, heavy, and difficult to transport to hard‐to‐reach areas such as dense forests and undeveloped areas without easy vehicle access. Using Haloxylon ammodendron, a plant widely distributed in arid lands, as an example, we used a consumer‐grade handheld camera to take a series of overlapping images of this plant. Computer vision and photogrammetric software were used to reconstruct highly detailed three‐dimensional data of the plant surface. This surface data was compared to the point cloud of the plant acquired from concomitant terrestrial laser scanning. We demonstrated that the accuracy and degree of completeness of the image‐derived point clouds are comparable to that of laser scanning. Plant structural parameters (such as tree height and crown width) and three‐dimensional models extracted from the point clouds also agree well with a relative difference of less than 5%. Our case study shows that a common camera and image processing software can be an affordable, highly portable, and viable option for acquiring accurate and detailed high‐density and high‐resolution three‐dimensional information about plant structure in the environment. This digitization technique can record the plant and its surrounding environment effectively and efficiently, and it can be applied to many ecological fields and studies.

With the rapid development of Unmanned Aerial Vehicle (UAV) technology, more and more UAVs have been used in forest survey. UAV (RGB) images are the most widely used UAV data source in forest resource management. However, there is some uncertainty as to the reliability of these data when monitoring height and growth changes of low-growing saplings in an afforestation plot via UAV RGB images. This study focuses on an artificial Chinese fir (Cunninghamia lancelota, named as Chinese Fir) young forest plot in Fujian, China. Divide-and-conquer (DAC) and the local maximum (LM) method for extracting seedling height are described in the paper, and the possibility of monitoring young forest growth based on low-cost UAV remote sensing images was explored. Two key algorithms were adopted and compared to extract the tree height and how it affects the young forest at single-tree level from multi-temporal UAV RGB images from 2019 to 2021. Compared to field survey data, the R² of single saplings’ height extracted from digital orthophoto map (DOM) images of tree pits and original DSM information using a divide-and-conquer method reached 0.8577 in 2020 and 0.9968 in 2021, respectively. The RMSE reached 0.2141 in 2020 and 0.1609 in 2021. The R² of tree height extracted from the canopy height model (CHM) via the LM method was 0.9462. The RMSE was 0.3354 in 2021. The results demonstrated that the survival rates of the young forest in the second year and the third year were 99.9% and 85.6%, respectively. This study shows that UAV RGB images can obtain the height of low sapling trees through a computer algorithm based on using 3D point cloud data derived from high-precision UAV images and can monitor the growth of individual trees combined with multi-stage UAV RGB images after afforestation. This research provides a fully automated method for evaluating the afforestation results provided by UAV RGB images. In the future, the universality of the method should be evaluated in more afforestation plots featuring different tree species and terrain.