Epsilon Archive for Student Projects

Abstract

Supply line management is a big factor in the success of any forest company’s ability to produce competitive forest products from raw materials. To achieve an efficient and profitable procurement of these raw materials it is important to have satisfactory inventory of the resources throughout the supply line. The present method for inventory of stockpiles in lumberyards and terminals is a manual mensuration process with subjective assessments of the stockpiles. The precision requirement of today is that the result needs to be within 10 % of the true value. The aim of this study is to test if the volume estimations can be made using Structure from Motion (SfM) for mantle volume estimations and image segmentation for wood proportion assessment. The traditional wood proportion estimation was also used for the final volume estimations thus two results are compared to each stockpile. The study was conducted on four stockpiles of two different assortments of lumber at Norra Skogsägarnas sawmill “Sävarsågen” with assistance from VMF Nord for reference data to compare the final volume estimations with. A terrestrial laser scanner was used on one of the stockpiles for validation of the mantel model produced by the SfM process. The final wood volume from the SfM process in combination with the traditional wood proportion assessment made overestimations between 5.78 % and 25.56 % of the true value. Two out of the four piles made volume estimations in accordance to present standards. For the volume estimations using image segmentation for assessing the solid wood proportion the accuracy ranged from -0.04 % to 3.58 % of the real value. All four estimations was within present requirements using image analysis. This study concludes that SfM is a viable option for stockpile inventory that replicates present day methods but with the possibility of superior accuracy in measurement of the mantle volume. Also the method has the possibility of being conducted objectively.