Roundwood measurement of truck loads by laser scanning A field study at Arauco pulp mill Nueva Aldea
Mats Nylinder, Tonny Kubénka, Mikael Hultnäs Introduction
In Chile, Brazil and other countries in South America systems for automatic scanning of trucks to determine the volume of wood have been used for several years. This system is based on laser scanning and advanced mathematical algorithms to calculate the volume. One manufacturer of such systems is Woodtech of Chile, with its’ Logmeter 4000 system. The system is also used to determine the volume for chips and coal (Anon 2008/1). Laser scanning of individual logs is used in many sawmills and is a well known technique (Björklund, L., 2003, Edlund, J., 2004). Scanning of whole trucks to determine the volume of wood is done in some places in Finland and Sweden but it is not as established as in South America (Anon 2006/2, Huttunen, T., 2006, Marjomaa, J. & Sairanen, P. 1996, Moilanen, P., 2003 Nylinder, M., 1992). The purpose of this study is to compare the manual measurement of pulpwood used in Sweden with automatic laser scanning of truckloads to try to develop a more efficient way of measuring wood for Scandinavian conditions.
In the pulp industry in Sweden the wood raw material represents about 50% of the total cost of production and in the sawmill industry about 70%. Considering that this is a large proportion of the total cost, relatively few resources are put into the development of new systems for measurement the quantity and quality of incoming wood. Today in Sweden most of the pulpwood is measured when it arrives at the pulp mills. The trucks arrive at the measuring station where each pile of wood is measured manually with a stick by scaling its height, length and width. The solid volume content under bark is then visually estimated. The wood is also graded with respect to species, dimension, rot, reject etc. To continuously check and control the measuring, samples are taken for manual log by log measuring and determination of solid volume under bark. This sample measuring is, for some methods, also used to adjust the original measurement. For the measuring of samples a new automatic technique is under development (Anon 2008/2).
Material and methods
Pulpwood is graded in many countries by its weight, green or dry. The problem with this method is the variation in its moisture content and the technique to determine moisture content is still very costly and time consuming. Over time several investigations have been performed regarding possibilities of changing today’s pulpwood volume measuring system, with prospects of replacing it with weighing the pulpwood and determining its moisture content, e.g. (Björklund, L., 1988), Thygesen, L., 1996 and Hultnäs, M., 2008).
The study was carried out in November 2008 at the Arauco Nueva Aldea pulp mill in the region of Bio Bio in Chile. Logs of eucalyptus (Eucalyptus globulus) pulpwood were sorted in three dimension classes: large, medium and small. The goal was to get at least a truck pile of logs of each dimension class. The logs were manually debarked. Each log in each dimension class was manually measured with a calliper and a ruler for its diameter and length. The diameter measurements were taken 10 cm from the top end, in the middle of the log and 10 cm from the butt end. The length was recorded. For logs which were noted as butt logs, the diameter was measured 50 cm from the butt end. All logs were numbered to facilitate the identification if some measures were falsely recorded. After this measurement, 26 randomly chosen logs of each dimension class were measured once more to determine the accuracy and repeatability of the manual measurement.
In Scandinavia forest growth, standing volume and cost for harvesting is related to volume, not weight, and although a system for weight is used there is a need for quantities in volume. This means that it is of great interest to develop systems to have a lower cost and more efficient system to determine the volume. Volume in combination with weight is probably the overall best system as this will also give information of density and the freshness of wood.
The diameters were recorded in cm classes, values within a cm range were recorded as class bottom, 1
e.g. a log that had the actual diameter 10.6 cm was recorded as 10 cm in the protocol. For the volume calculation half a cm was added to the recorded value. The length measurement was recorded in dm at class bottom. When calculating the volume, five centimetres were added to the recorded volume. The volume of the logs was calculated in two ways: A: Mid-diameter. (Huber’s formula) Based on the mid-diameter and length. The following formula was applied: Va =
Dm 2 1 ×π × × L 100000 4
(1)
Va = Volume in m3 Dm = Mid-diameter in cm L = Log length in dm B: Top-butt-diameter. Based on top and butt diameters. The volume calculation is based on a model developed for Swedish roundwood, mainly Scots pine (Pinus silvestris) and Norway spruce (Picea abies) but also for birch and other hardwoods. (Anon 2000)
Figure 1. Woodtech’s Logmeter 4000. Laser-scanners are circled in yellow.
laser-scanning system (Anon 2008/1). The piles were then rearranged or switched with another pile and measured again in the same way.
The following formula was applied:
The manual measurement of the piles was carried out according to a method developed for Swedish pulpwood and Swedish conditions. In this method frame volume of the pile is measured by a scaler while standing on a scaling bridge at the same level as the pile. In the study, this was done at ground level. To estimate the solid volume content, a standard value for the specific wood species is used, and then adjusted by the estimation of diameter, stacking quality, straightness, delimbing quality, bark content, taper etc (Kubénka, T., 2008).
Va = 100000 × π × L × (α × (Db )2 + (1 − α )× (Dt )2 ) (2) 4
Va = Volume in m3 Dt = Diameter in top end in cm Db = Diameter in butt end in cm L = Log length in dm α = Constant according to Table 1 Table 1. The constant α Diameter top (cm) -349
Length (cm) 350-449
450+
-14 0.485
0.485
0.485
15-24
0.460
0.455
0.465
The roundwood truck measurement was done by laser scanning the periphery of the load from both sides and from above, with the Logmeter 4000. The truck is driven through the system at a fairly constant and low speed to get as good raw data as possible. Using the information generated, Woodtech has developed algorithms to estimate length, diameter, stack volume and solid volume. As pulpwood trucks in South America seldom have truck-mounted cranes, algorithms for the cranes are not yet fully developed. Due to this, the influence
The logs were loaded pile-wise on a truck for pulpwood. After a pile was loaded on the truck and on the trailer, stack volume and solid volume content of the pile were manually estimated. After the manual measurement the truck drove to the Logmeter, Figure 1, where it was measured by the 2
Diameters Next, each cross-section of the load is analyzed, determining the best fit for the diameter of each periphery log. This is replicated throughout the load, leading to the identification of all exterior logs.
of the crane was manually adjusted before all the laser data was presented. Woodtech, who developed the scanning system, describes the equipment as follows: ”Logmeter 4000 is the most advanced laser-scanner system available for the scaling of wood loads. The measurement process is simple and automatic, with human intervention in less than 5% of cases. The scanning process takes less than a minute, which allows more than 600 trucks to be scanned per day. The system can measure different configurations, including fixed and variable length, tree-length, chip and forest residue loads. Scanning As the truck enters the measurement area, high precision laser-scanners sweep the truck, generating hundreds of cross-sections of the load and creating a 3D representation with more than 1 million individual measurements.
Figure 4. Principle for automatic identification of periphery logs.
Biometrics With each log of the periphery modelled in space, the system then calculates the biometric information of the log, including diameter, length, taper, and crook.
Figure 2. Principle for 3-D Interface for log identification. Figure 5. Principle for individualization of periphery logs and identification of biometric characteristics.
Segmentation Using image processing algorithms, the limits of the load are identified and the unwanted elements, such as truck wheels, drawbars, and platform, are eliminated.
Solid volume In the final step of the process, the solid volume is calculated. This is done using mathematical models which are calibrated for each specific site where a Logmeter is installed. The model determines the relationship between the parameters calculated for the periphery of the load and the total volume. The model can be easily updated over time, thus ensuring precision. Generated results include volume, diameter, and length classes for each bundle of logs. All results and images are stored in the mill’s administrative system. The system allows for trucks to be first measured and then calculated at different moments
Figure 3. Principle for automatic segmentation of the load.
3
Table 2. The amount of measurements Pile Number of logs Number of pile measurements on trailer
Number of pile measurements on truck
Blue, small dimension
132
1
3
Green, medium dimension
118
3
3
Red, large dimension
102
3
1
Figure 6. The blue pile (left), the green pile (centre) and the red pile (right).
in time. This is useful for reducing truck lines and even allows for unmanned operation. The Logmeter includes a powerful auditing system which allows all data to be reviewed at a later date, even permitting electronic re-measurements if necessary. This maximizes the transparency and traceability of the measurement process. Results can also be shared by Internet with suppliers, management, and government” (Anon, 2008/1).
Table 3. Description of the logs and piles Variable Number of logs
The last part of the study was a study of repeatability with the laser-scanner. The truck and trailer were driven ten times through the Logmeter without rearranging the load.
Pile Green
Red
132
118
102
Amount of butt logs (%)
17
29
44
Middle diameter (cm) Average Std
9.2 2.4
13.2 2.1
16.5 2.4
Length (dm) Average Std
56.1 8.5
57.3 8.1
57.0 7.9
Taper, (mm/m) Average Std
5.5 1.99
5.2 1.95
4.8 2.07
0.045 0.024 5.877
0.080 0.029 9.484
0.133 0.041 13.587
Volume, ”top/but diam.” (m3) Average 0.046 Std 0.024
0.083 0.032
0.138 0.043
Volume, ”mid diam.” (m3) Average Std Total
For the analysis of the results, standard formulas for mean and standard deviation were used. The correlation between two variables reflects the degree to which the variables are related. In this paper we have used the Pearson Product Moment Correlation (Yamane, T., 1969).
Blue
Table 4. Result of manual control measurement of 26 logs per pile
Result The characteristics of the logs according to the manual log by log measurement are described in Table 3. After the original measurement, 26 logs from each pile were randomly sampled and measured once more. The result of this measurement was compared with the first original measurement. Table 4 shows the control measurement minus the original measurement and the standard deviation of the difference.
Method for volume measurement
Of the four blue piles, three piles were on the truck 4
Blue
Pile/Stack Green
Red
Volume mid, (m3) Mean of diff. (m3) Mean diff. (%) Std for diff. (m3)
1.2832 -0.0027 0.21 0.0055
2.1714 3.4682 0.0011 -0.0007 0.05 0.02 0.0047 0.0057
Volume top/but, (m3) Mean of diff. (m3) Mean diff. (%)
1.3118 -0.0005 0.04
2.2572 3.6583 -0.0009 -0.0014 0.05 0.04
Diameter laser/diameter log-by-log r2 = 96.9%, P = 0.000, F = 374
and one pile was on the trailer. Three of the green piles were on the truck and three were on the trailer. One of the red piles was on the truck and three were the trailer. The estimation of diameter and length is given in Table 5 and the estimation of solid volume in Table 6.The correlation between the two methods is:
Length laser/ length log-by-log r2 = 35.2%, P = 0.025, F = 6
Table 5. Recorded diameter and length by the laser and log-by-log measurement Stack Trailer/ Load Truck No
Diam. Laser (cm)
Blue Trailer 1 Truck 1 Truck 2 Truck 3 Mean Std
Diam. Log-by-log (cm)
Length, Laser (dm)
Length, Log-by-log (dm)
11.3 9.2 10.9 9.2 10.4 9.2 10.8 9.2 10.9 9.2 0.32
54.8 56.1 54.1 51.1 54.0 2.12
56.1 56.1 56.1 56.1 56.1
Green Trailer 1 Trailer 2 Trailer 3 Truck 1 Truck 2 Truck 3 Mean Std
13.4 13.2 13.5 13.2 13.1 13.2 14.2 13.2 13.5 13.2 13.5 13.2 13,5 13.2 0.37
58.4 57.9 57.4 56.6 53.4 57.3 56.8 1.79
57.3 57.3 57.3 57.3 57.3 57.3 57.3
Red Trailer 1 Trailer 2 Trailer 3 Truck 1 Mean Std
16.5 16.5 17.0 16.5 16.6 16.5 16.8 16.5 16.7 16.5 0.22
58.4 58.7 58.6 55.7 57.9 1.44
57.0 57.0 57.0 57.0 57.0
Table 6. Manual and laser determination of solid volume for the piles Stack Trailer/ Load Laser Log-by-log (m3) Truck No (m3)
Manual, Stack estim. (m3)
Diff. Laser, Log-by-log (%)
Diff. Laser Manual (%)
Blue Trailer 1 Truck 1 Truck 2 Truck 3 Mean Std
6.05 6.09 6.10 6.09 6.20 6.09 6.02 6.09 6.09 0.08
5.66 6.33 6.63 6.17 6.18 0.41
-0.7 0.2 1.8 -1.0 0.1 1.3
6.4 -3.8 -6.9 -2.5 -1.7 5.7
Green Trailer 1 Trailer 2 Trailer 3 Truck 1 Truck 2 Truck 3 Mean Std
9.75 9.85 9.43 9.85 9.50 9.85 9.99 9.85 9.75 9.85 9.95 9.85 9.73 0.22
9.33 9.66 9.86 9.98 10.80 10.42 10.01 0.53
-1.0 -4.4 -3.7 1.4 -1.0 1.0 -1.3 2.4
4.3 -2.4 -3.7 0.1 -10.8 -4.7 -2.9 5.0
Red Trailer 1 Trailer 2 Trailer 3 Truck 1 Mean Std
13.08 14.05 14.03 14.05 13.30 14.05 13.59 14.05 13.50 0.40
13.57 14.62 14.04 14.31 14.14 0.45
-7.4 -0.1 -5.6 -3.4 -4.2 3.1
-3.7 -4.2 -5.6 -5.3 -4.7 1.0
5
Figure 7. Relationship between manual and laser-scanner estimation (left) and between log by log and laser-scanner estimation (right) of solid volume. Table 7. Manual and laser determination of stacked volume Stack
Trailer/ Truck
Load No
Laser (m3)
Manual (m3)
Diff. Laser-Manual (%)
Blue Trailer 1 Truck 1 Truck 2 Truck 3 Mean Std
11.12 11.78 11.48 12.56 11.74 0.61
10.89 11.94 13.82 12.35 12.25 1.05
2.0 -1.4 -20.4 1.7 -4.5 10.6
Green Trailer 1 Trailer 2 Trailer 3 Truck 1 Truck 2 Truck 3 Mean Std
17.04 16.62 17.83 16.92 18.78 18.52 17.62 0.90
15.55 16.95 16.99 16.63 19.64 18.28 17.34 1.43
8.7 -2.0 4.7 1.7 -4.6 1.3 1.7 4.7
Red Trailer 1 Trailer 2 Trailer 3 Truck 1 Mean Std
22.13 24.87 22.6 21.42 22.76 1.49
20.87 23.97 21.94 22.35 22.28 1.29
5.7 3.6 2.9 -4.3 2.0 4.3
The correlation between laser and manual estimation were for: Solid volume , laser/manual r2 = 98.7%, P = 0.000, F = 926 Solid volume, laser/ log by log = 0.000, F = 1553
r2 = 99.2%, P
Stacked volume, laser/manual = 0.000, F = 232
r2 = 95.1%, P
A measure of the stacking in the piles is the solid volume percent, the solid volume in relation to the stack volume, Table 8.
Figure 8. Relationship between manual and laser-scanner measurement of stacked volume.
6
Table 8. Estimation of solid volume percentage
Blue
Green
Red
Number of obs. Mean laser (%) Std (m3) Mean, manual (%) Std
4 52 3.0 51 2.2
6 55 2.8 58 1.9
4 59 2.9 64 1.7
At the end of the study the trailer and truck were scanned 10 times with the laser without rearranging the piles between each measurement. Figure 9. Example of end cracks.
Table 9. Result from ten independent measurements of the blue and red piles without rearrangement of the load Statistics
Diam. (cm)
Length Stack Solid (dm) vol. (m3) vol. (m3)
Blue pile, trailer Mean Std
11.6 0.31
55.6 1.38
11.24 0.30
6.1 0.14
Red pile, truck Mean
16.6
56.8
21.7
13.7
Std
0.29
0.56
0.21
0.21
Discussion The log by log measurement is supposed to yield the ”true” volume not including whorls or sections were the diameter is increasing from butt to top end. Some butt logs were also shortening due to quite large fell combs. Quite a few logs were split and had cracks and the diameters were taken in a way trying to disregard these defects.
Figure 10. Crack along the log.
The way of calculating the volume by the top/butt method is also based on formulas developed for Swedish conditions and not for eucalyptus. The debarking of the logs was not 100% perfect and compared to bark on pine and spruce in Sweden the eucalyptus bark has more strips in the longitudinal direction which probably makes it more difficult to measure precisely disregarding the bark which was left. The cracks and splits, bark and fell combs will of course for both the manual and the laser scanning generate errors.
Figure 11. Example of fell comb.
The arithmetic mid diameter for the small dimension wood was recorded in the manual measurement as 9.6 cm, which is less than the average diameter for spruce pulpwood in Sweden, which is about 11-12 cm. The average diameter for softwood and birch pulpwood in Sweden is about 13 cm (Statistics, VMF Qbera). This means that the medium class in this study, green, has about the same average
Figure 12. Logs before loading on truck and trailer.
7
diameter as Swedish pulpwood. The manually measured length of the wood was recorded at about 57 dm which compared to random length pulpwood in Sweden is at least 10 dm longer. The taper was measured to be about 5 to 6 mm/m. The taper for pine and spruce pulpwood in Sweden is normally a bit more.
which is the correct diameter, length and volume. In Table 7 the solid volume estimated for piles on the trailer and truck is given. The laser scanning gives between 0 to 4% less volume than the log by log measurement. Compared with the manual estimation the laser gives 1.5% less volume for the small dimension and 3% resp. 5% less volume for the green and blue pile. The correlation, r2 = 99%, between the two methods is very good and significant which indicates a potential that the methods can be calibrated to each other. The correlation between the laser-scanning and log by log measurement is also significant and strong, with r2 = 99%.
The two manual methods to calculate volume of the logs yield about the same volume and the control measurement indicates that the manual log by log measurement has a good repeatability. One must however mention that the repeatability measurement was done by the same persons and other persons measuring the same logs will probably give other results. Another factor that can influence the log by log measuring is weather conditions. The logs themselves are also a factor since it is probably easier to get a stable and correct manual log measurement on large dimension logs which are properly debarked and of same length compared to small crooked and cracked logs.
The laser-scanning for the small dimension calculates a smaller stacked volume compared to the manual estimation, by about 5%. For the green and red piles the laser scanning gives a 2% larger volume. One observation has a difference of 20% which is difficult to explain but could be a registration error. The correlation between the two methods is very strong r2 = 95%, but not as significant as for the solid volume.
The estimated log diameter by the laser-scanner measurement was for the blue logs 10.9 cm compared to the manual measurement which gave 9.2 cm. For the green pile, the corresponding values were 13.5 for the laser-scanner and 13.2 for the manual measurement and for the red pile, 16.7 cm and 16.5 cm respectively. For the medium and large dimension logs, the diameter is almost the same for the two methods. For the small dimension logs the difference is about 10%. The correlation between the two methods, r2 = 97% is however strong and the relation is significant which indicates that it has a good potential for calibrating. An exact figure on diameter is often of less importance – it is normally a rough value which can be of interest for pricing and sorting of wood in Scandinavia
When looking at precision and exactness it should be seen in relation to the stack volume of the piles. A large volume, such as for the red piles (22 m3), means that a smaller proportion of the logs of the pile are directly scanned by the laser in comparison to the case of the small blue pile (12 m3). The ten repeatability measurements of the small dimension logs indicate a very good repeatability of the laser-scanning method. This is as always a very strong side of a mechanical automatic method compared to a manual measurement. As the laser-scanner has a high repeatability, it has a high potential to be adjusted to different methods used by different mills and standards existing in different countries. The scanning is very fast (
