Bayesian Approach to Tree Detection Based on Airborne Laser Scanning Data

Lähivaara, Timo; Seppänen, Aku; Kaipio, Jari P.; Vauhkonen, Jari; Korhonen, Lauri; Tokola, Timo; Maltamo, Matti

doi:10.1109/tgrs.2013.2264548

Cited by 58 publications

(49 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We found the ITD algorithms performed well in E. regnans forests as our ITD rates of 72% and 68%, using only 0.9 LiDAR pulses·m −2 , were similar to studies in Scandinavian and European forests that used LiDAR with 3-10 pulses·m −2 [15,20,46]. Our LiDAR data resolution was more comparable to [47], who used 1 pulse·m −2 in forests with Scots pine (Pinus sylvestris), Norway spruce (Picea abies) and birch (Betula sp.)…”

Section: Discussionsupporting

confidence: 84%

Using Tree Detection Algorithms to Predict Stand Sapwood Area, Basal Area and Stocking Density in Eucalyptus regnans Forest

et al. 2015

View full text Add to dashboard Cite

Managers of forested water supply catchments require efficient and accurate methods to quantify changes in forest water use due to changes in forest structure and density after disturbance. Using Light Detection and Ranging (LiDAR) data with as few as 0.9 pulses m −2 , we applied a local maximum filtering (LMF) method and normalised cut (NCut) algorithm to predict stocking density (SDen) of a 69-year-old Eucalyptus regnans forest comprising 251 plots with resolution of the order of 0.04 ha. Using the NCut method we predicted basal area (BAHa) per hectare and sapwood area (SAHa) per hectare, a well-established proxy for transpiration. Sapwood area was also indirectly estimated with allometric relationships dependent on LiDAR derived SDen and BAHa using a computationally efficient procedure. The individual tree detection (ITD) rates for the LMF and NCut methods respectively had 72% and 68% of stems correctly identified, 25% and 20% of stems missed, and 2% and 12% of stems over-segmented. The significantly higher computational requirement of the NCut algorithm makes the LMF method more suitable for predicting SDen across large forested areas. Using NCut derived ITD segments, observed versus predicted stand BAHa had R 2 ranging from 0.70 to 0.98 across six catchments, whereas a generalised parsimonious model applied to all sites used the portion

show abstract

Section: Discussionsupporting

confidence: 84%

Using Tree Detection Algorithms to Predict Stand Sapwood Area, Basal Area and Stocking Density in Eucalyptus regnans Forest

et al. 2015

View full text Add to dashboard Cite

show abstract

“…Compared with the studies focused on developing the tree detection algorithms (Wang et al 2008;Yao et al 2012;Lähivaara et al 2014), This doctoral work aimed at the unbiased estimates for the purpose of the ITD-based forest inventory. Maltamo et al (2004) manually selected the cut point, which was realized automatically in this doctoral work.…”

Section: Discussionmentioning

confidence: 99%

“…Xu Q., Hou Z., Maltamo M., Tokola T. (2014). Calibration of area based diameter distribution with individual tree based diameter estimates using airborne laser scanning.…”

Section: List Of Original Articlesmentioning

confidence: 99%

“…III. Xu Q., Hou Z., Vauhkonen J., Maltamo M., Tokola T. (2014). DBH prediction for the ALS-detected trees using the quantile-based nearest neighbor imputation.…”

Section: List Of Original Articlesmentioning

confidence: 99%

“…Hyyppä et al (2012) utilized the last pulse data for tree detection and correctly matched 6% more trees than when the first pulse was used. Lähivaara et al (2014) utilized the same data as the doctoral work, and developed a Bayesian approach to detect trees by fitting multiple 3D crown height models to point cloud data. They achieved an improvement in detection rate from 53% to 70%.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Calibration of the tree size distributions by combining the area-based approach and the individual tree detection using the airborne laser scanning

Xu¹

2014

Diss. For.

View full text Add to dashboard Cite

3 Xu, Q. 2014. Calibration of the tree size distributions by combining the area-based approach and the individual tree detection using the airborne laser scanning. Dissertationes Forestales 182. 28 p. Available at http://dx.doi.org/10.14214/df.182 ABSTRACTIndividual tree detection (ITD) -based forest inventory using the airborne laser scanning (ALS) data suffers from under-estimation problem, which arises mainly from the suppressed trees that are difficult to be detected from the air. Uncertainty of tree-level estimates, like tree height, diameter at breast height (DBH) and the modeled stem volume also contributes to the inaccuracy of the plot-level estimates. The doctoral work tackled the under-estimation problem from the perspectives of both tree and plot levels. At the plot level, suppressed trees were retrieved from the left tail of the tree size distributions derived from the area-based approach. At the tree level, DBH of single trees were predicted using the quantile-based nearest neighbor imputation.Area-based approach (ABA) -based forest inventory is able to provide accurate and unbiased plot-level estimates of forest attributes, such as total stem volume. K-MSN method is used in the ABA to simultaneously predict the forest attributes of interest. If treelevel field measurements are available in the sample plots, it's possible to apply the k-MSN method to predict tree size (DBH or height) distributions for the sample plots. The combination of the ITD-derived tree size distributions with the ABA-derived distributions makes it possible not only to improve the ABA-derived saw log estimates, but also to retrieve the suppressed trees for the ITD-derived tree size distributions. The replacement and the histogram matching were utilized to calibrate the tree size distributions. The results showed that after the calibration, the RMSE of the predicted total volume decreased by 2 %, and the bias was negligible. The quantile-based nearest neighbor imputation was able to predict the DBH as accurate as the benchmarking method, the k-MSN. It utilized the ALSmeasured tree height and crown diameter as the two predictor variables and achieved even better accuracy than the k-MSN method for larger trees with diameters 16 cm. The improved DBH estimates also benefit stem volume estimation, which led to the improved tree-and plot-level estimations of the stem volume. LIST OF ORIGINAL ARTICLESThe thesis consists of a summary followed by three articles listed below. Articles I and II are reprinted with kind permission of the publishers, while article III is author's version of the submitted manuscript.I.Xu Q., Hou Z., Maltamo M., Tokola T. (2014). Calibration of area based diameter distribution with individual tree based diameter estimates using airborne laser scanning. ISPRS Journal of Photogrammetry and Remote Sensing 93: 65-75.

show abstract