Page 380 - Mechatronics for Safety, Security and Dependability in a New Era

Ch73-I044963.fm Page 364 Friday, July 28, 2006 1:57 PM
Ch73-I044963.fm
364
364 Page 364 Friday, July 28,2006 1:57 PM
level. This method can be changed by using weighting coefficients that depend of the size of the drop in
the diameter profile. Large drops are given a larger weight in order to get the fitted polynomial closer to
the measurements following large drops. Another adjustment to the weighted polynomial fitting method
is to change weighting according to the smoothness of the stem. This is motivated by the experimental
result that weighted polynomial fitting is good for stems containing many drops in the diameter profile,
whereas polynomial fitting is more suited for smooth stem profiles. The two improvements to the ordi-
nary weighted polynomial fitting can be used also together so that the overall weighting is determined by
the smoothness of the stem and local weighting by the size of the drops in the diameter profile.

The most significant difference is between the methods based on linear approximation and on polyno-
mial fitting. The former works locally and improves diameter accuracy only little or not at all. The
polynomial fitting methods that can utilize better the global knowledge of the diameter measurements
have a notable effect on the accuracy of diameter measuring. Offline smoothing could be advantageous
particularly considering volume determination. After the logs have been bucked, their volume is saved to
the harvester database. If the stem profile was smoothed offline before the volume is computed, accu-
racy could be better due to a more accurate diameter measurement. The methods that are based on
polynomial fitting or weighted polynomial fitting produce very smooth stem profiles. Smoothness could
be an advantage if the profiles are used to update a stem narrowing matrix or some other prediction tool.

CONCLUSIONS

Several methods were presented to improve the quality of stem diameter measurements. More reliable
automatic operations in forest harvesters help the operator to concentrate more on planning and other
tasks that cannot be done automatically. In addition to this, more accurate diameter measuring has sev-
eral advantages, like maximizing the value of the timber, making the work more efficient, and
eliminating the need of measuring the timber twice. Diameter measurement error along the whole stem
and in the cutting points can be reduced by using the measurement processing algorithms presented.
When the sum of square errors in all measurement points is considered, the improvement is with online
filtering up to 10.2 %, with online smoothing up to 15.3 %, and with offline smoothing up to 18.9 %.
When the sum of square errors in the cutting points is considered, the improvement is with online filter-
ing up to 9.8 %, with online smoothing up to 10.7 %, and with offline smoothing up to 14.1 %.

The results obtained in this paper should be applicable to all harvesters that have a measurement system
similar to the one used in this study. However, all manufacturers use their own algorithms and imple-
mentations in their machines, which may cause the methods presented in this paper to perform
differently. The compatibility of the algorithms must be verified case-by-case. Many of the methods
presented in this paper decrease the error in the diameter measurement. However, when considering the
value of the results, it must be taken into account that the dataset used to validate and compare the meth-
ods is relatively small, which decreases the reliability of the results. The methods contain also
parameters that have been adjusted according to the properties of this dataset. Further studies will be
needed to show if these parameters need to be changed for different felling sites.

REFERENCES

Grewal, Mohinder S. and Angus P. Andrews (1993). Kalman Filtering. Prentice-Hall, Inc., New Jersey.
Laasasenaho, Jouko (1982). Taper curve and volume functions for pine, spruce and birch. Finnish Forest
Research Institute, Helsinki.
Lappi, Juha (1986). Mixed linear models for analyzing and predicting stem form variation of scots pine.
The Finnish Forest Research Institute, Helsinki.
Metsateho (2003). Metsdteho website. URL: http://www.metsateho.fi/.