Individual tree crown detection and delineation across a woodland using leaf-on and leaf-off imagery from a UAV consumer-grade camera

Berra, Elias Fernando

doi:10.1117/1.jrs.14.034501

Cited by 13 publications

(3 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Yancho et al [48] achieved 47.6% detection accuracy with a subcrown point approach applied in a mature, coniferous dominant, temperate rainforest. In a mature, mixed-wood forest, Berra [49] yielded overall accuracies as high as 80% for the delineation of coniferous trees in plots containing approximately 40 trees using a watershed-based approach. In plantation settings, Kattenborn et al [50] and Torres-Sánchez et al [51] yielded overall accuracies of 86% and 90%, respectively for ITC delineation with object-based image analysis (OBIA) approaches.…”

Section: Itc Delineationmentioning

confidence: 99%

Automatic Delineation and Height Measurement of Regenerating Conifer Crowns under Leaf-Off Conditions Using UAV Imagery

et al. 2020

View full text Add to dashboard Cite

The increasing use of unmanned aerial vehicles (UAV) and high spatial resolution imagery from associated sensors necessitates the continued advancement of efficient means of image processing to ensure these tools are utilized effectively. This is exemplified in the field of forest management, where the extraction of individual tree crown information stands to benefit operational budgets. We explored training a region-based convolutional neural network (Mask R-CNN) to automatically delineate individual tree crown (ITC) polygons in regenerating forests (14 years after harvest) using true colour red-green-blue (RGB) imagery with an average ground sampling distance (GSD) of 3 cm. We predicted ITC polygons to extract height information using canopy height models generated from digital aerial photogrammetric (DAP) point clouds. Our approach yielded an average precision of 0.98, an average recall of 0.85, and an average F1 score of 0.91 for the delineation of ITC. Remote height measurements were strongly correlated with field height measurements (r2 = 0.93, RMSE = 0.34 m). The mean difference between DAP-derived and field-collected height measurements was −0.37 m and −0.24 m for white spruce (Picea glauca) and lodgepole pine (Pinus contorta), respectively. Our results show that accurate ITC delineation in young, regenerating stands is possible with fine-spatial resolution RGB imagery and that predicted ITC can be used in combination with DAP to estimate tree height.

show abstract

Section: Itc Delineationmentioning

confidence: 99%

Automatic Delineation and Height Measurement of Regenerating Conifer Crowns under Leaf-Off Conditions Using UAV Imagery

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Another factor that significantly contributed to the high recognition accuracy was the careful selection of the phenological period for the drone survey [47,48]. When addressing specific research objectives, it is crucial to consider the biological characteristics of the study objects and to select the appropriate seasons when the objects of interest are most distinguishable from other elements in the landscape.…”

Section: Discussionmentioning

confidence: 99%

Unveiling the Potential of Drone-Borne Optical Imagery in Forest Ecology: A Study on the Recognition and Mapping of Two Evergreen Coniferous Species

Korznikov,

Kislov,

Petrenko

et al. 2023

Remote Sensing

View full text Add to dashboard Cite

The use of drone-borne imagery for tree recognition holds high potential in forestry and ecological studies. Accurate species identification and crown delineation are essential for tasks such as species mapping and ecological assessments. In this study, we compared the results of tree crown recognition across three neural networks using high-resolution optical imagery captured by an affordable drone with an RGB camera. The tasks included the detection of two evergreen coniferous tree species using the YOLOv8 neural network, the semantic segmentation of tree crowns using the U-Net neural network, and the instance segmentation of individual tree crowns using the Mask R-CNN neural network. The evaluation highlighted the strengths and limitations of each method. YOLOv8 demonstrated effective multiple-object detection (F1-score—0.990, overall accuracy (OA)—0.981), enabling detailed analysis of species distribution. U-Net achieved less accurate pixel-level segmentation for both species (F1-score—0.981, OA—0.963). Mask R-CNN provided precise instance-level segmentation, but with lower accuracy (F1-score—0.902, OA—0.822). The choice of a tree crown recognition method should align with the specific research goals. Although YOLOv8 and U-Net are suitable for mapping and species distribution assessments, Mask R-CNN offers more detailed information regarding individual tree crowns. Researchers should carefully consider their objectives and the required level of accuracy when selecting a recognition method. Solving practical problems related to tree recognition requires a multi-step process involving collaboration among experts with diverse skills and experiences, adopting a biology- and landscape-oriented approach when applying remote sensing methods to enhance recognition results. We recommend capturing images in cloudy weather to increase species recognition accuracy. Additionally, it is advisable to consider phenological features when selecting optimal seasons, such as early spring or late autumn, for distinguishing evergreen conifers in boreal or temperate zones.

show abstract

“…The used methodology consists of local maxima ltering and watershed-based object detection using a (normalised) digital elevation model. Such work ow is typical for detecting and delineating tree crowns in various environments(Berra, 2020;Huang et al., 2018; Miraki et al, 2021). On the other hand, other methods such as Geographic Object-Based Image Analysis -GEOBIA (Díaz-Varela et al, 2015; Yurtseven et al, 2019) or neural networks approaches and machine learning methods are successfully used as well (Adhikari et al, 2020; Kestur et al, 2018).…”

mentioning

confidence: 99%

The Potential of Widespread Ready-Made UAVs in Quantifying Conifers and Delineating Their Crowns

Komárek

Klápště

Hrach

et al. 2021

Preprint

View full text Add to dashboard Cite

With the ever-improving Earth observation capabilities, variables such as tree health status, biomass storage, or stand structure are increasingly estimated through remote sensing. While many protocols of data acquisition and satellite data processing are in place, the still novel unmanned aerial vehicles (UAVs) face some challenges during data acquisition and processing. While tree height extraction seems to be a common practice, identifying individual trees and measuring their crowns is still quite tricky. We performed several flights with three different UAVs and four different popular sensors over two sites with coniferous forests of various ages at flight levels of 100–200 m above ground level (AGL) using custom settings preset by UAV solution suppliers. Considering the success rate of the individual tree identification, casual RGB cameras provided more consistent results at all flight levels (84 − 77% for Phantom 4), while the success of tree identification decreases with higher flight levels and smaller crowns in the case of multispectral cameras (77 − 54% for RedEdge-M). In general, RGB cameras yielded the best results at 150 m AGL while multispectral cameras at 100 m AGL. Regarding the accuracy of the measured crown diameters, most datasets tended to overestimate when using automatic crown delineation within the lidR package. Only RGB cameras yielded satisfactory results (Mean Absolute Error – MAE of 0.79–0.99 m and 0.88–1.16 m for Phantom 4 and Zenmuse X5S, respectively). Multispectral cameras overestimated more than RGB cameras, especially in the full-grown forest (MAE = 1.26–1.77 m); on the other hand, they offered, in addition to the structural, also spectral information. We conclude that widespread ready-made solutions mounted with low-cost RGB cameras yield very satisfactory results for describing the structural forest information at 150 m AGL. When (multi)spectral information is needed, we recommend reducing the flight level to 100 m AGL to acquire sufficient structural forest information.

show abstract

Individual tree crown detection and delineation across a woodland using leaf-on and leaf-off imagery from a UAV consumer-grade camera

Cited by 13 publications

References 40 publications

Automatic Delineation and Height Measurement of Regenerating Conifer Crowns under Leaf-Off Conditions Using UAV Imagery

Automatic Delineation and Height Measurement of Regenerating Conifer Crowns under Leaf-Off Conditions Using UAV Imagery

Unveiling the Potential of Drone-Borne Optical Imagery in Forest Ecology: A Study on the Recognition and Mapping of Two Evergreen Coniferous Species

The Potential of Widespread Ready-Made UAVs in Quantifying Conifers and Delineating Their Crowns

Contact Info

Product

Resources

About