Page 375 - Mechatronics for Safety, Security and Dependability in a New Era

Ch73-I044963.fm Page 359 Friday, July 28, 2006 1:57 PM
Friday, July 28,2006
1:57 PM
Ch73-I044963.fm
Page 359
359
359
SOFT-SENSOR BASED TREE DIAMETER
MEASURING

Vesa Holtta

Control Engineering Laboratory, Helsinki University of Technology
P. O. Box 5500, FI-02015 TKK, Finland

ABSTRACT
The forest harvester used for felling timber is a complex machine with a high degree of automation. To
work properly, automatic functions need accurate measurements. Tn this paper tree diameter measure-
ment is improved using different filtering and smoothing algorithms. The cases where smoothing is done
during stem processing and after the stem has been processed are treated separately. Validation using
manually measured data indicates that the methods that are presented improve the performance of the
diameter measurement considerably.

KEYWORDS
Measurement, filtering, smoothing, Kalman filtering, Kalman smoothing

INTRODUCTION

Currently the vast majority of wood felled in Finland is felled with a forest harvester. In spite of the con-
siderable amount of automation that helps the harvester operator in his work, felling timber can still be
seen as handicraft. The operator must be a trained professional who is able to navigate the harvester in
the woods without damaging the environment, and to choose the trees to cut so that future growth of the
forest is guaranteed. Achieving these goals is compromised if the operator must concentrate on too many
secondary tasks. Consequently, the need of operator interventions in less important tasks should be
minimized. This is not possible unless the operator can rely on the automatic functions.

One field where a computer can outperform a human operator is bucking, i.e. selecting the points where
the stem should be cut to logs. The price of a log is determined by its volume, but also by its grade
(stock, paper-wood, etc.). The grade can often be changed by choosing the cutting points differently.
Optimizing the bucking such that the value of the stem is maximized is a well-suited task for a com-
puter. In order for the optimization to be successful, the length and diameter measurements that are fed
to the optimization algorithm must be reliable. However, several factors can degrade the quality of the