Quality assessment of standing trees using 3D laserscanning

Similar documents
DESCRIBING FOREST STANDS USING TERRESTRIAL LASER-SCANNING

AUTOMATIC DETERMINATION OF FOREST INVENTORY PARAMETERS USING TERRESTRIAL LASER SCANNING

Individual Tree Parameters Estimation from Terrestrial Laser Scanner Data

1 INTRODUCTION. on the branch surface is described. Based on this, the computation of a smooth branch axis and surface is presented.

KEY WORDS: Laser scanning, Close Range, Modelling, Forestry, Automation, Measurement, Algorithms

APPLICATION OF TERRESTRIAL LASER SCANNERS FOR THE DETERMINATION OF FOREST INVENTORY PARAMETRS

GEODETIC MEASURING METHODS AND SHAPE ESTIMATION OF CONCRETE THIN SHELL SURFACE

IMAGER The new way of scanning highly accurate, fast, reliable and flexible.

5 Applications of Definite Integrals

AUTOMATIC ORIENTATION AND MERGING OF LASER SCANNER ACQUISITIONS THROUGH VOLUMETRIC TARGETS: PROCEDURE DESCRIPTION AND TEST RESULTS

SKELETONIZATION AND SEGMENTATION OF POINT CLOUDS USING OCTREES AND GRAPH THEORY

3D BUILDINGS MODELLING BASED ON A COMBINATION OF TECHNIQUES AND METHODOLOGIES

The potential of the terrestrial laser scanning for geometrical building facades inspection

High Resolution Tree Models: Modeling of a Forest Stand Based on Terrestrial Laser Scanning and Triangulating Scanner Data

How we build reality. Company Overview. Case Study Stonehenge in High Defi nition

Automatic forest inventory parameter determination from terrestrial laser scanner data

VEGETATION MODELLING BASED ON TLS DATA FOR ROUGHNESS COEFFICIENT ESTIMATION IN RIVER VALLEY

Virtual Assembly Technology on Steel Bridge Members of Bolt Connection Zhu Hao

REPRESENTATION REQUIREMENTS OF AS-IS BUILDING INFORMATION MODELS GENERATED FROM LASER SCANNED POINT CLOUD DATA

METHODS FOR THE AUTOMATIC GEOMETRIC REGISTRATION OF TERRESTRIAL LASER SCANNER POINT CLOUDS IN FOREST STANDS

SIMULATED LIDAR WAVEFORMS FOR THE ANALYSIS OF LIGHT PROPAGATION THROUGH A TREE CANOPY

Measurements using three-dimensional product imaging

A VOXEL-BASED TECHNIQUE TO ESTIMATE THE VOLUME OF TREES FROM TERRESTRIAL LASER SCANNER DATA

THREE-DIMENSIONAL FOREST CANOPY STRUCTURE FROM TERRESTRIAL LASER SCANNING

The suitability of airborne laser scanner data for automatic 3D object reconstruction

S8.6 Volume. Section 1. Surface area of cuboids: Q1. Work out the surface area of each cuboid shown below:

USING UNUSUAL TECHNOLOGIES COMBINATION FOR MADONNA STATUE REPLICATION

Lidar Sensors, Today & Tomorrow. Christian Sevcik RIEGL Laser Measurement Systems

Quantitative structure tree models from terrestrial laser scanner data Pasi Raumonen! Tampere University of Technology!

Stress Wave Propagation on Standing Trees -Part 2. Formation of 3D Stress Wave Contour Maps

Small-footprint full-waveform airborne LiDAR for habitat assessment in the ChangeHabitats2 project

EDEXCEL NATIONAL CERTIFICATE UNIT 4 MATHEMATICS FOR TECHNICIANS OUTCOME 1

HIGH RESOLUTION COMPUTED TOMOGRAPHY FOR METROLOGY

REDUCING THE ERROR IN TERRESTRIAL LASER SCANNING BY OPTIMIZING THE MEASUREMENT SET-UP

NATIONWIDE POINT CLOUDS AND 3D GEO- INFORMATION: CREATION AND MAINTENANCE GEORGE VOSSELMAN

TREE HEIGHT ESTIMATION METHODS FOR TERRESTRIAL LASER SCANNING IN A FOREST RESERVE

Error budget of terrestrial laser scanning: influence of the incidence angle on the scan quality

Point clouds to BIM. Methods for building parts fitting in laser scan data. Christian Tonn and Oliver Bringmann

TERRESTRIAL LASER SYSTEM TESTING USING REFERENCE BODIES

Calypso Construction Features. Construction Features 1

Single Tree Stem Profile Detection Using Terrestrial Laser Scanner Data, Flatness Saliency Features and Curvature Properties

Flank Millable Surface Design with Conical and Barrel Tools

GENERATING BUILDING OUTLINES FROM TERRESTRIAL LASER SCANNING

POINT CLOUD REGISTRATION: CURRENT STATE OF THE SCIENCE. Matthew P. Tait

Chapter - 13 (Surface areas and Volumes)

Impact of 3D Laser Data Resolution and Accuracy on Pipeline Dents Strain Analysis

Shadow Casting in World Builder. A step to step tutorial on how to reach decent results on the creation of shadows

QUALITY CONTROL OF CONSTRUCTED MODELS USING 3D POINT CLOUD

Snow cover change detection with laser scanning range and brightness measurements

KRISTAPS KLAVA - HEAD OF LASER SCANNING DEPARTMENT AT MERKO 3D LASER SCANNING POSSIBILITIES IN REAL ESTATE

Automatic analysis and weld indications classification on TFM acquisitions

Linear and Rotary Infrared Scan System for Measuring Circumference

USE OF A POINT CLOUD CO-REGISTRATION ALGORITHM FOR DEFORMATION MEASURING

LIDAR: MAPPING WITH LASERS

AUTOMATIC EXTRACTION OF BUILDING FEATURES FROM TERRESTRIAL LASER SCANNING

A New Method in Shape Classification Using Stationary Transformed Wavelet Features and Invariant Moments

Accurate 3D Face and Body Modeling from a Single Fixed Kinect

Landmark Detection on 3D Face Scans by Facial Model Registration

IDENTIFYING STRUCTURAL CHARACTERISTICS OF TREE SPECIES FROM LIDAR DATA

11. Mensuration. Q 2 Find the altitude of a trapezium, the sum of the lengths of whose bases is 6.5 cm and whose area is 26 cm 2.

Burkhard Plinke; Xavier Le Fur; Peter Meinlschmidt; Friedrich Schlüter

Find the maximum value and minimum values of f(x) for x in [0, 4]. Graph f(x) to check your answers. ( answer)

No Brain Too Small PHYSICS

Further Volume and Surface Area

Using Perspective Rays and Symmetry to Model Duality

3D-Analysis of Microstructures with Confocal Laser Scanning Microscopy

Experimental accuracy assessment of different measuring sensors on workpieces with varying properties

Recognition and Measurement of Small Defects in ICT Testing

RECOGNISING STRUCTURE IN LASER SCANNER POINT CLOUDS 1

3D Digitization of Human Foot Based on Computer Stereo Vision Combined with KINECT Sensor Hai-Qing YANG a,*, Li HE b, Geng-Xin GUO c and Yong-Jun XU d

TREE DETECTION AND DIAMETER ESTIMATIONS BY ANALYSIS OF FOREST TERRESTRIAL LASERSCANNER POINT CLOUDS

Structured light 3D reconstruction

TERRESTRIAL LASER SCANNING FOR AREA BASED DEFORMATION ANALYSIS OF TOWERS AND WATER DAMNS

Comparison of Probing Error in Dimensional Measurement by Means of 3D Computed Tomography with Circular and Helical Sampling

February 07, Dimensional Geometry Notebook.notebook. Glossary & Standards. Prisms and Cylinders. Return to Table of Contents

13. Surface Areas and Volumes. Q 2 The diameter of a garden roller is 1.4 m and it is 2 m long. How much area will it cover in 5 revolutions?

CRONOS 3D DIMENSIONAL CERTIFICATION

Registration of Moving Surfaces by Means of One-Shot Laser Projection

3D Terrestrial Laser Scanner Innovative Applications for 3D Documentation

Photogrammetry and 3D Car Navigation

Available online at ScienceDirect. Energy Procedia 69 (2015 )

Grade 6 Math Circles October 16 & Non-Euclidean Geometry and the Globe

Advanced Image Processing, TNM034 Optical Music Recognition

Fiber Composite Material Analysis in Aerospace Using CT Data

SP about Rectangular Blocks

FILTERING OF DIGITAL ELEVATION MODELS

Principal Roll Structure Design Using Non-Linear Implicit Optimisation in Radioss

Electromagnetic Field Simulation in Technical Diagnostics

PLC Papers. Created For:

Using SLAM-based Handheld Laser Scanning to Gain Information on Difficult-to-access Areas for Use in Maintenance Model

DETECTION AND ROBUST ESTIMATION OF CYLINDER FEATURES IN POINT CLOUDS INTRODUCTION

Nearest Neighbor Methods for Imputing Missing Data Within and Across Scales

Three-Dimensional Laser Scanner. Field Evaluation Specifications

Interactive Collision Detection for Engineering Plants based on Large-Scale Point-Clouds

青藜苑教育 Volume of cylinder = r h 965 = r = 6 r 965 = r 9.98 = r = r So the radius of the cylinde

Aim: How do we find the volume of a figure with a given base? Get Ready: The region R is bounded by the curves. y = x 2 + 1

Z-LASER Improved Beam Modeling With Optical Fibers. Vision Technology Forum April 15th, 2015

Section 7.2 Volume: The Disk Method

Strategy. Using Strategy 1

Grade 9 Surface Area and Volume

Transcription:

Quality assessment of standing trees using 3D laserscanning G.R.M. van Goethem 1, J.W.G. van de Kuilen 2, W.F. Gard 3, W.N.J.Ursem 4 1,3 Delft University of Technology, Faculty of Civil Engineering and Geosciences, The Netherlands 2 Delft University of Technology, NL, CNR-Ivalsa, Italy 4 Delft University of Technology, Botanical Garden, The Netherlands 1 gvangoethem@gmail.com, 2 J.W.G.vandeKuilen@tudelft.nl, 3 W.F.Gard@tudelft.nl, 4 W.N.J.Ursem@tudelft.nl ABSTRACT 3D laserscanning of trees in the forest is becoming more and more available to everybody with the upcoming of better and faster scanners. Where in the past one single scan could only cover 40 degrees, nowadays 360 degree scanners are commercially available and delivering very good data with a range up to 80 meters within 3 minutes per scan. The use of these scanners is one thing; good commercial use of the obtained data is another. To be able to put the scan data to good use, it is necessary to know how the data can be used as optimal as possible. Earlier research has proven that defining trees and making a high quality assessment of the tree is possible. Also some papers discuss the link between outside stem parameters and inside knot properties. In this paper the results of research on scanning beams and reconstructing the single tree will be discussed. First, the tree is scanned in the forest and then boards with a length of 3 meters are sawn. Next, knots are defined based upon the scan. The knots are then connected to a model of the interior of the tree. The position of these knots defines the quality class of the single beams extracted from the tree. With the knowledge of the position of the bigger knots it is possible to turn the tree in the best possible direction before sawing, to have the best quality end product, with the highest possible strength class assigned. Finally, the reconstruction of the tree is discussed and an automatic reconstruction program is proposed. INTRODUCTION Quality assessment of standing trees helps optimizing forestry management. When an up to date and in depth database of all trees in the forest is available, the use of high quality timber can be optimized and loss of timber because of poor tree selection can be minimized,. This will bring more revenue for forest owners and better use of resources, both from an economic and environmental point of view. To make sure that the data collected is proficient enough to reach these goals, a high quality measuring technique is needed. 3D laserscanning (or similar) may provide this. The use of a 3D laser scanner for checking outside quality of trees has already been proven. In this research the laser scanner will also be used to check and validate the results from the tree surface measurement and compare it with the findings from the per beam measured internal quality. This research will try to give answers to the following questions: 1. With 3D data available can 3D laser scanning deliver enough information to determine single tree characteristics and can this data be used to increase the quality of sawn wood? 2. Can 3D laserscanning techniques be used as a basis for forest management? 145

SCANNING OF TREES In the Botanical Garden of Delft University of Technology, the opportunity arose to scan and use trees for this research. A Honey tree (Styphonolobium japonicum (L.)) and Chestnut tree (Castanea sativa Mill.) was scanned in high to super high resolution, see Fig. 1. With this resolution, details on the bark of the tree are easily visible as Figure 1 Honey tree and Chestnut tree can be seen from the picture below. Subsequent branches and other imperfections cannot be missed, Fig. 2. Because the scanning took place at the end of the winter, no leaves were blocking the scan and the whole stem was visible. Figure 2 Detail of Chestnut 146

RECONSTRUCTION OF THE CHESTNUT TREE The tree was cut into single trunks with length of 3 meters, which were sawn to non squared edged boards. Because of the sawing process some parts of the tree were not delivered. As a consequence, the whole diameter of the tree could not be reconstructed using the beams. The beams that did connect were, because of the high quality of the surface scan, coupled using a patch on both connecting sides along with 3 or more vertices (Fig. 3). This process was used because the ICP algorithm (Iterative Closest Point), normally used to couple different scans could not be used. This being done, all 5 boards that originated from the top trunk of the tree was coupled into one trunk. The same workflow was used for the middle section of the tree. The results are shown in the Fig. 4 and 5. In Fig. 4 the lines between the boards are clearly visible. Quality of the beam registration can be derived from the nice connection of the outer flaws. Because of the absence of all beams after the sawing process only 5 beams were available. This should be, although not complete, enough for the research of the outer flaws, since the whole tree will be used later on to combine with the not measured beams. Of the middle part of the tree more beams were available for scanning, so a more complete image is shown, Fig. 5. While this part has more data available, this part will be used as basis for the determination of the knot propagation inside the tree. Determination of knots The single knots which are visible on the beams are defined as described earlier in this paper. When the found knots are connected we find the following pattern on the middle part of the tree. Bare in mind, this pattern only gives an accurate result perpendicular to the sawing direction. Knots in the other direction are not visible on the beams as ovals but more as lines, as can be seen in Fig. 6 and 7. On the middle part of the tree these lines were not clearly visible in the scans or pictures and thus this direction is not taken into account. Figure 3: Vertices used for scan coupling Figure 4: Beam composed of single boards, top part of tree Figure 5: Beam composed of single boards, centre part of tree 147

Figure 6: Scanned knots direction SCANNING OF SINGLE BOARDS Larch boards (Larix kaempferi (LAMB) CARR) The beams are scanned from 2 to 7 different positions, see Fig. 8. This is done to make sure all knots are measured in the highest resolution possible. Note that these beams are not specific from the chestnut tree, but from a larch tree, which were scanned and used in the research done for the SBB (National Forest Association of the Netherlands). Apart from a time constraint, another reason to use these larch boards was that more small knots are available that need to be assessed. From this position, using ZF- Lasercontrol software, the visible knots are manually defined and numbered, see Fig. 9. For every beam this process is repeated on both sides. The numbered vertices can then be connected at each side of the beam. With this approach the direction and position of every knot can be determined. With these positions defined, the individual beam strength grade can be determined using grading standards such as DIN 4074. The larch boards were dried and scanned some months after the actual sawing process. This gave very good results regarding to the knot detection, but relatively poor results in reconstructing the tree. A number of reasons could be identified. The first reason was that the boards had been placed during prolonged periods in a dry indoor climate, resulting in large Figure 7: Knot section of Chestnut tree Figure 8: Single scan position distortions, Fig. 10. This made it extra difficult to reconstruct the tree. A second reason was that the trees were cut into boards in a planer mill, severely decreasing the cross section, making it difficult to reproduce the stem s surface. 148

Figure 9: Beam with position of knots Figure 10: Twisted board Boards of Chestnut The beams from the chestnut tree were still wet, but had suffered from some fungus build-up. This gave more noise in the scanning data and thus did not give the same clean results as with the Larch. Although this was a small problem, the result of the manual positioning of the knots in the chestnut tree was satisfactory. For the correct definition of the position of the knots the following could be concluded: - the board surface has to be clean and free from fungi. - boards have to be dried properly in order to clearly see the differences between? normal wood and knots. Alternatively, when the beams are processed directly after the sawing process, no fungi will form, and scans will be crisp and accurate. Figure 11: Discoloration caused by fungi 149

ASSIGNED VISUAL GRADE BASED ON KNOT SURFACE A first estimation of the visual grade of the beam will be carried out based on the minimum knot size compared to the width of the beam in accordance to DIN 4074. For this provision only knots with a minimum size of 5 mm or more are considered. For this provision the smallest diameter of the biggest knot is governing. To give an example of the possibilities that scanning gives for recognition and computing these values an example of a single beam is given: Figure 12: Knot width 1. First, a check is done which beam can be used; 2. Normative knots are processed in 3D; 3. The measurements and assignment are made according to DIN 4074, see Fig. 12 Table 1: Grading characteristics Sorting method : Knots Strength class (DIN 4074) LS 7, LS 7K LS 10, LS20K LS 13, LS13K General A 3/ 5 A 2 / 5 A 1/ 5 Oak A 3/ 5 A 2 / 5 A 1 / 6 Grade characteristic Sorting class LS 7 LS 10 LS 13 Single Knot < ½ < 1 / 3 < 1 / 5 Knot collection < 2 / 3 < ½ < 1 / 3 Knot at slim side < 2 / 3 < 1 / 3 In the next picture the knots that will be measured are specified (Fig. 13): Knot 2 Knot 1 Figure 13: Measured knots 150

Conference COST E53, 29-30 October 2008, Delft, The Netherlands Knot 1 This knot is on both sides of the beam clearly visible. First, one side is examined. Afterwards in the same pointcloud the other side is examined whereupon a calculation can be made of the knot size according to the pictures shown below. Measurements in this example are taken directly from the software. This means sub-millimeter accuracy can be achieved. To the example above the following applies, from Fig. 14 and 15: 13.16 (1) 0.0756 174 From the other side the following measurement to the knot is made: A 18.38 0.1056 (2) 174 Finally, using the first formula the following conclusion can be made: A A 13.26 18.38 2 *174 0.0909 (3) Strength class according to table 1: LS13 Knot 2 The same process is used to measure the second knot. (Fig. 16 and 17). The first normative width of this knot is shown is the next picture: 11.26 174 0.065 Figure 14: Knot size 1 (4) From the other side the following picture can be seen: 13.76 0.079 (5) 174 Combined: 11.26 13.76 A 0.072 (6) 2 174 A Figure 15: Knot size 1 Strength class according to table 1: LS13 151

CONCLUSIONS ON THE KNOT DETECTION CAPABILITY It can be concluded that by using laser scans and the natural colour deviation of knots it is very well possible to define the strength grade. Thanks to the colour deviation it is possible to automatically isolate the knots and to make them visible. An example of this method is given in the next sequence of pictures. When a scanner is used, where Figure 16: Knot size 2 RGB values are also measured (e.g. the Imager 5006 of Zoller & Fröhlich where a high resolution photograph is coupled to the scandata) it will be even easier to define the different knots. It will also be possible when a high resolution picture is made to define the exact centre of the knot. This will result in a better indication of the direction and size and shape of the knot, see Fig. 18. With the knowledge that it is possible to calculate and define the position of the knots Figure 17: Knot size 2 internally inside beams it can now be analysed how well it will be possible to directly derive the quality of a standing tree, based on the outside parameters. Figure 18: Isolate knot intensity of different knots in 2D or 3D 152

Therefore, ways to isolate the outside flaws need to be defined. Since it is possible to automatically define the inside wood flaws, the outside flaws can be defined using the same approach. AUTOMATIC EXTERNAL WOOD QUALITY DETECTION According to the research done in Schütt et al. (2004) and Thomas et al. (2006), it is very well possible to define the exact location of the wood flaws by determining the 3D position on a reference cylinder and changing these coordinates to a 2D picture. This same process will be used in the research on the chestnut tree. The semi-automatic research, with the outcome of the position of the knots, based on scans of the beams, will be checked. A procedure was written that will determine the centerline of the tree and automatically map the external flaws onto a 2D picture that can be analyzed by neural analyses software on different wood flaws (Schütt et al.) The process used for this is partly described in Pfeifer et al. (2004). The principle idea exploited in the reconstruction is that tree and branches are described as right circular cylinders. This model deviates from reality in two aspects. First, the cross-section is in most cases not circular, but of a more general form. Taking this into account, the exact centerline of the tree is hard to find, but can be done according to Pfeifer and Winterhalder (2004). This article first describes a way to determine the tree axis, using approximation of the stem with a cylinder. After this has been done, using this axis direction, the more accurate free form of the tree is automatically modelled. In Fig. 19 and 20, the approximation of the stem-axis using cylinders is shown. The real stem form is visible using a top view of a slice on a height z from the bottom. Figure 19: Modeled stem, approximation Figure 20: Modeled stem, real 153

Reconstruction of Chestnut tree In this chapter the reconstruction of the chestnut tree will be discussed. This will first be done using meshing software. After cleaning the point cloud a mesh is created and each 10 cm a curved freeform is generated from this mesh. The result is shown in Fig. 22. As can be seen the result is comparable to the results shown in Fig. 19. The difference from the approach described in Pfeifer et al. (2004), is that the stem axis is not defined first, but is defined by determining the centre of each free-form line. When connecting these centres the stem form should appear, with high accuracy. Downside of this workflow is that a mesh is needed. In the software these freeform lines are directly fit into the pointcloud, thus removing another calculation step. To make the result more accurate, it is possible to make cutplanes not only perpendicular to the z- direction, but also in different angles. The result, with two 45 degree cutplanes is shown in Fig. 22. Figure 21: Free form of the stem Figure 22: Free form of the stem, final and detailed Now that the centerline of the tree is defined by free forms, the coordinates of the 3D points can be translated from 3D to 2D. When this is done, a picture of the outside bark of the tree is created, and this can be, like the single beams, used for automatic wood quality detection. With all the parameters present of the location of these flaws, it s an easy step to go back and forth from 2D to 3D. 154

CONCLUSIONS Based on 3D laser scanning, it is possible to define the outside parameters of a standing tree. With existing and techniques to be developed, from one single tree one can define: 1. Stem volume; 2. Stem direction; 3. Branch position; 4. Branch direction; 5. Internal position of knots; 6. External position of knots; 7. 3D model of tree; 8. 3D database of tree models, with 3D coordinates (forest); 9. Exact position of the tree within a forest. From the list above it becomes clear that laserscanning trees in a forest with the goal to use the data for forest management is possible, and cost effective. With a 3D database available, with all important features of all trees available, management is simpler and less wood will be wasted. The exact workflow and results need to be tested thoroughly, in order to create a robust program that can be used commercially. Research is furthermore proposed on the size of the knot. In above research the knot is only defined by a line. As was pointed out in the chapter calculating the strength of the beam, the width and breadth of the knot define the strength class. Instead of using a line, the knot propagation more resembles a cone. When this cone is defined an even more accurate picture of the inside of the tree will appear, and at some point the link between breadth of the knot and breadth of the beam do not influence the quality anymore. This means fewer calculations need to be done. REFERENCES DIN 4074-1 Sortierung von Holz nach Tragfähigkeit - Sortierkriterien für Nadelholz Ikonen, V., Kellomäki, S., Peltola, H., Linking tree stem properties of Scots pine (Pinus sylvestris L.) to sawn timber properties through simulated sawing, Faculty of Forestry, University of Joensuu, Finland, 2002. Pfeifer, N., Gorte, B., Winterhalder, D., Automatic reconstruction of single trees from terrestrial laserscanner, Delft, The Netherlands, 2004 Pfeifer, N., Winterhalder, D. Modelling of tree cross sections from terrestrial laser scanning data with free form curves, Delft, The Netherlands, 2004 Schütt, C., Spiecker, H., Thies, M. Qualitätsbestimmung von Wertholzstammen, Einsatz moderner Laser zur Vermessung äußerlich sichtbarer Holzfehler stehender Bäume und deren Ausprägung im Inneren, Freiburg, Germany, 2004 155

Schütt, C., Aschoff, T., Winterhalder, D., Thies, M., Kretschmer, U., Spiecker, H., Approaches for recognition of wood quality of standing trees based on terrestrial laserscanner, Freiburg, Germany, 2004 Thomas, E., Thomas, L., Shaffer, C., Defect detection of severe surface defect on barked hardwood logs, Virginia, USA, 2006 Http://en.wikipedia.org/wiki/Iterative_Closest_Point ACKNOWLEDGMENT Special thanks go out to P. Holverda, who made it possible to code the software necessary for this research. 156