Tree stem diameter estimation from terrestrial point clouds

Mona Forsman

Tree stem diameter estimation from terrestrial point clouds

Mona Forsman

Institutionen för skoglig resurshushållning

Publikation: Avhandling › Doktorsavhandling

Sammanfattning

Skogsägare, myndigheter och miljöorganisationer behöver skoglig information för planering av skogsskötsel, värdering och miljöövervakning. Informationen samlas in med en kombination av flyg- och satellitfjärranalys och stickprovsinventeringar av provytor i fält. Stamdiameter mäts med klavar, vilket är arbetsintensivt. Markbaserade sensorer skulle kunna göra inventeringen snabbare, och fler stickprov skulle kunna inventeras. Sensorer på skördare skulle kunna leverera trädkartor över vad som är kvarlämnat efter en avverkning, eller de skulle kunna samla in data till ett förarstödsystem. Den första artikeln beskriver en fotogrammetrisk metod för skattning av stamdiameter och position på provytor. Problematiska ljusförhållanden minskade mängden användbara provytor. På de användbara provytorna positionerades 76 % av träden och för 40 % kunde diametrarna uppskattas. I den andra artikeln förbättrades resultaten från ett mobilt laserskanningsprojekt genom att behandla punktmolnet linjevis, och genom att använda punkternas intensitet som ett kvalitetsvärde. Stamdiameterskattningens RMSE reducerades från 24% till 14%, men biaset ökade något från -1.9% till 2.3%. Kantpunkterna på stammarna identifierades som en felkälla eftersom de inte hittades längs den förväntade cirkeln. Den tredje artikeln undersöker detta kantpunktsproblem genom simulering av laserskanner / trädkombinationer. En relation mellan diameterfelet och laserns träffyta i förhållande till stamdiametern hittades. Vanligt använda mobila laserscannrar visade sig ge en relativ bias på 10 % eller mer i skogliga tillämpningar. I den fjärde artikeln användes en panoramabild av intensiteten hos ett laserpunktmoln för att upptäcka träd med lyckat resultat. Slutsatserna är att punktmoln från de olika sensorerna är användbara för uppskattning av träddiametrar och positioner, men de har sensorberoende egenskaper som kan införa fel. Dessa egenskaper och precisionskravet bör beaktas när datainsamlingen planeras och sensorn väljs.

Originalspråk	Engelska
Kvalifikation	Doktor i filosofi
Förlag	Swedish University of Agricultural Sciences
Tryckta ISBN	978-91-7760-248-4
Elektroniska ISBN	978-91-7760-249-1
Status	Publicerad - 2018

Nyckelord

Error analysis
Forest inventory
Mobile laser scanning
Mobile mapping
Point cloud processing
Precision forestry
Simulation
Terrestrial laser scanning
Terrestrial photogrammetry
Tree stem diameter

SLU-serie

Acta Universitatis Agriculturae Sueciae

Åtkomst av dokument

FulltextSlutlig publicerad version, 847 KBLicens: Ej specificerad

Andra filer och länkar

Permanent länk till denna sida

Citera det här

@phdthesis{c5196e5e06a54ebab473fe5802ab129d,

title = "Tree stem diameter estimation from terrestrial point clouds",

abstract = "Forest owners, governments, and environmental organizations demand forest information for planning of forest operations, estimation of value, and for environmental monitoring. This information is collected using airborne and satellite remote sensing combined with field inventory of sample plots. Stem diameter is measured with calipers, which is labor-intensive. Terrestrial sensors could make the inventory faster, and more samples could be taken. Sensors mounted on forest harvesters could produce maps of the trees left after forest operations, or collect data for an operator support system. The first article describes a photogrammetric method using a multi-camera rig for estimation of stem diameter and position on field plots. Problematic light conditions reduced the usable amount of field plots. On adequate field plots, 76\% of the trees were detected and positioned, and on 40\% of the trees the diameters could be estimated. In the second article, the results from a mobile laser scanning project was improved by treating the data line-wise, and by using the intensity of the laser points as a quality value. The RMSE of the stem diameters was reduced from 24\% to 14\%, but the bias increased slightly from -1.9\% to 2.3\%. The edge points on the stems were identified as an error source since they were not found along the expected circle. The third article investigates this edge point problem by simulation of laser scanner/tree combinations. A relationship between the diameter error and the footprint size relative to the stem diameter was found. Commonly used mobile laser scanners were concluded to give a relative bias of 10\% or more when estimating diameters using circle fit methods. In the fourth article, a panorama image of the intensities of a laser scanner point cloud was used to detect trees, with adequate results. The overall conclusions are that point clouds from the various sensors are useful for estimation of tree diameter and positions, but they have sensor-dependent properties that can introduce errors. These properties, and the precision requirement should be considered when the data acquisition is planned and the sensor is selected.",

keywords = "Error analysis, Forest inventory, Mobile laser scanning, Mobile mapping, Point cloud processing, Precision forestry, Simulation, Terrestrial laser scanning, Terrestrial photogrammetry, Tree stem diameter, Error analysis, Forest inventory, Mobile laser scanning, Mobile mapping, Point cloud processing, Precision forestry, Simulation, Terrestrial laser scanning, Terrestrial photogrammetry, Tree stem diameter",

author = "Mona Forsman",

year = "2018",

language = "English",

isbn = "978-91-7760-248-4",

series = "Acta Universitatis Agriculturae Sueciae",

publisher = "Swedish University of Agricultural Sciences",

number = "2018:54",

}

TY - BOOK

T1 - Tree stem diameter estimation from terrestrial point clouds

AU - Forsman, Mona

PY - 2018

Y1 - 2018

N2 - Forest owners, governments, and environmental organizations demand forest information for planning of forest operations, estimation of value, and for environmental monitoring. This information is collected using airborne and satellite remote sensing combined with field inventory of sample plots. Stem diameter is measured with calipers, which is labor-intensive. Terrestrial sensors could make the inventory faster, and more samples could be taken. Sensors mounted on forest harvesters could produce maps of the trees left after forest operations, or collect data for an operator support system. The first article describes a photogrammetric method using a multi-camera rig for estimation of stem diameter and position on field plots. Problematic light conditions reduced the usable amount of field plots. On adequate field plots, 76% of the trees were detected and positioned, and on 40% of the trees the diameters could be estimated. In the second article, the results from a mobile laser scanning project was improved by treating the data line-wise, and by using the intensity of the laser points as a quality value. The RMSE of the stem diameters was reduced from 24% to 14%, but the bias increased slightly from -1.9% to 2.3%. The edge points on the stems were identified as an error source since they were not found along the expected circle. The third article investigates this edge point problem by simulation of laser scanner/tree combinations. A relationship between the diameter error and the footprint size relative to the stem diameter was found. Commonly used mobile laser scanners were concluded to give a relative bias of 10% or more when estimating diameters using circle fit methods. In the fourth article, a panorama image of the intensities of a laser scanner point cloud was used to detect trees, with adequate results. The overall conclusions are that point clouds from the various sensors are useful for estimation of tree diameter and positions, but they have sensor-dependent properties that can introduce errors. These properties, and the precision requirement should be considered when the data acquisition is planned and the sensor is selected.

AB - Forest owners, governments, and environmental organizations demand forest information for planning of forest operations, estimation of value, and for environmental monitoring. This information is collected using airborne and satellite remote sensing combined with field inventory of sample plots. Stem diameter is measured with calipers, which is labor-intensive. Terrestrial sensors could make the inventory faster, and more samples could be taken. Sensors mounted on forest harvesters could produce maps of the trees left after forest operations, or collect data for an operator support system. The first article describes a photogrammetric method using a multi-camera rig for estimation of stem diameter and position on field plots. Problematic light conditions reduced the usable amount of field plots. On adequate field plots, 76% of the trees were detected and positioned, and on 40% of the trees the diameters could be estimated. In the second article, the results from a mobile laser scanning project was improved by treating the data line-wise, and by using the intensity of the laser points as a quality value. The RMSE of the stem diameters was reduced from 24% to 14%, but the bias increased slightly from -1.9% to 2.3%. The edge points on the stems were identified as an error source since they were not found along the expected circle. The third article investigates this edge point problem by simulation of laser scanner/tree combinations. A relationship between the diameter error and the footprint size relative to the stem diameter was found. Commonly used mobile laser scanners were concluded to give a relative bias of 10% or more when estimating diameters using circle fit methods. In the fourth article, a panorama image of the intensities of a laser scanner point cloud was used to detect trees, with adequate results. The overall conclusions are that point clouds from the various sensors are useful for estimation of tree diameter and positions, but they have sensor-dependent properties that can introduce errors. These properties, and the precision requirement should be considered when the data acquisition is planned and the sensor is selected.

KW - Error analysis

KW - Forest inventory

KW - Mobile laser scanning

KW - Mobile mapping

KW - Point cloud processing

KW - Precision forestry

KW - Simulation

KW - Terrestrial laser scanning

KW - Terrestrial photogrammetry

KW - Tree stem diameter

KW - Error analysis

KW - Forest inventory

KW - Mobile laser scanning

KW - Mobile mapping

KW - Point cloud processing

KW - Precision forestry

KW - Simulation

KW - Terrestrial laser scanning

KW - Terrestrial photogrammetry

KW - Tree stem diameter

UR - https://res.slu.se/id/publ/104216

M3 - Doctoral thesis

SN - 978-91-7760-248-4

T3 - Acta Universitatis Agriculturae Sueciae

PB - Swedish University of Agricultural Sciences

ER -