Individual tree segmentation in deciduous forests using geodesic voting

Parkan, Matthew; Tuia, Devis

doi:10.1109/igarss.2015.7325844

Cited by 11 publications

(7 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Individual tree crown boundaries were detected from combining a UAV‐derived ortho‐mosaic (flights were performed during the fall of 2013) and a ALS‐derived canopy height model (Torabzadeh et al., 2014 , 2019 ). Individual tree segments from ALS were derived with the approach of Parkan and Tuia ( 2015 ) using geodesic voting. Ground‐truthed tree positions were linked to ITCs and geodesing voting‐derived point clouds using coordinates, crown size, and vertical layering as main constraints.…”

Section: Methodsmentioning

confidence: 99%

Remotely sensed between‐individual functional trait variation in a temperate forest

Guillén-Escribà

Schneider

Schmid

et al. 2021

Ecology and Evolution

View full text Add to dashboard Cite

show abstract

Section: Methodsmentioning

confidence: 99%

Remotely sensed between‐individual functional trait variation in a temperate forest

Guillén-Escribà

Schneider

Schmid

et al. 2021

Ecology and Evolution

View full text Add to dashboard Cite

show abstract

“…For the DBH estimation, the trunk point cloud quality may vary in different environments, and the DBH estimation methods, which use a certain height point cloud of the trunk to fit a circle or cylinder [11,17,18,20], may produce a low accuracy when the trunk has large noise. Thus, we propose a DBH estimation method based on multi-height diameters.…”

Section: Methodsmentioning

confidence: 99%

“…The correct single tree segmentation directly influences the DBH estimation and forest inventory, Scholars previously used the Euclidean cluster extraction method [13,15], the hierarchical clustering method [14], and arc detection algorithms [16] etc. to segment the trees from the MLS point cloud, and some scholars developed a graph based segmentation for the Airborne Laser Scanning data [17], they all achieved good results. However in some forests, the forest ground was characterized by rugged terrain and obstacles, such as rocks, shrubs, and deadwood [18], and there are lots of shrubs and leaves that may influence the single tree segmentation.…”

Section: Introductionmentioning

confidence: 99%

Single Tree Segmentation and Diameter at Breast Height Estimation With Mobile LiDAR

et al. 2021

View full text Add to dashboard Cite

The tree diameter at breast height (DBH) is one of the most important variables for monitoring the forest ecology. Mobile laser scanning (MLS), which has been widely applied in the forestry field, makes DBH measurement fast and convenient. However, there are many shrubs and deadwood in the neutral forest environment and the point clouds quality from MLS are easily affected by the environment which results in low single tree segmentation and DBH estimation accuracy. To improve the accuracy in a complex forest environment and low point cloud quality, we propose a relative density segmentation method for the single tree segmentation and DBH estimation method based on multi-height diameters for the DBH estimation. The relative density segmentation method calculates the relative density according to the ratio of density in two different scales, and segments the tree trunks by the higher relative density of trunk point clouds compared with their surroundings points. In the natural forest plot, the precision and recall of the proposed segmentation method reached 0.966 and 0.946, respectively; In the urban forest plot, the precision and recall reached 1 and 0.966, respectively. The proposed DBH estimation method was used to estimate the DBH of trees using multi-height diameters. The multi-height diameters combined with the outlier detection algorithm were able to improve the accuracy and robustness when the trunk point clouds have large noise. For the DBH estimation results, the mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) were 2.5 cm, 11.54%, and 3.17 cm, respectively, in the natural forest plot and 1.65 cm, 6.31%, and 1.97 cm, respectively, in the urban forest plot. The good experiment results indicate that the proposed method can achieve accurate and robust DBH extraction and provide fundamental data for supervision and sustainable development of forest resources.

show abstract

“…Moreover, high numbers of commission errors are usually obtained due to the presence of several local peaks within the same canopy [7]. Few works focus the attention on the detection of deciduous trees, mainly assuming leaf-off acquisition condition [8,9]. Here, the absence of leaves in the canopy strongly facilitates the penetration of the laser beam, thus making the delineation of the canopy structure easier.…”

Section: Introductionmentioning

confidence: 99%

An Automatic Technique for Deciduous Trees Detection in High Density Lidar Data Based on Delaunay Triangulation

Marinelli

Paris

Bruzzone

2019

IGARSS 2019 - 2019 IEEE International Geoscience and Remote Sensing Symposium

View full text Add to dashboard Cite

Individual tree detection in Light Detection and Ranging (LiDAR) data has been widely investigated in the literature. However, most of the methods work well on conifers but lead to poor accuracy in broad-leaved forest. The detection of deciduous trees is a complex task due to: (i) multiple local maxima present in the same canopy, and (ii) the tree-top (TP) can be in a different location from the canopy center. This paper presents an automatic technique which exploits high density LiDAR data to refine the detection of deciduous trees. First, the candidate tree-tops (CTPs) are detected using the standard level set method (LSM). Then, the Delaunay triangulation is used to generate a network topology which connects neighboring CTPs. For each pair of connected CTPs, geometrical features are extracted to automatically determine if the CTPs pair belongs to the same tree or to different canopies. The groups of CTPs identified as belonging to the same tree crown are merged into one TP. Preliminary numerical results show that the proposed method halves the commission errors of the initial TP detection by increasing the overall detection accuracy of 8.2%.

show abstract

Individual tree segmentation in deciduous forests using geodesic voting

Cited by 11 publications

References 7 publications

Remotely sensed between‐individual functional trait variation in a temperate forest

Remotely sensed between‐individual functional trait variation in a temperate forest

Single Tree Segmentation and Diameter at Breast Height Estimation With Mobile LiDAR

An Automatic Technique for Deciduous Trees Detection in High Density Lidar Data Based on Delaunay Triangulation

Contact Info

Product

Resources

About