Epsilon Archive for Student Projects

Abstract

In Sweden the cap portion of the harvested stem volume derives from Non-Industrial Private Forest (NIPF) owners. In the current study a Decision Support Tool (DST) for wood procurement planning based on stereo-matching of aerial images is presented. Two stages are described, namely (1) automatic segmentation using a Mean Shift algorithm; (2) wall-to-wall mapping of the stands using Semi-Global Matching (SGM) in combination with a high-resolution Digital Elevation Model (DEM). The study was conducted in a coniferous boreal forest area in northern Sweden. 365 sample plots (8 m radius) were measured in the field where HGV (dm) ranged between 49.0 - 246.0 dm (mean 139.3 dm), DGV 67.0 - 400.0 mm (mean 196.8 mm), VOL 7.0 - 665.0 m3/ha (mean 151.1 m3/ha) and BA 20.0 - 635.0 dm2/ha (mean 204.9 dm2/ha). Point clouds were extracted from the aerial images with 60% forward overlap. A canopy cover metric was used to improve the VOL and BA estimations. Plot level accuracies were calculated using leave-one-stand-out-cross-validation resulting in a Root Mean Square Error (in percent of surveyed mean) for: HGV 11.2%, DGV 15.2%, VOL (m3/ha) 33.5% and BA 30.3%. Each stand was given an average of the estimated forest variables enabling ranking between the stands based on their estimated values. The results indicated that the proposed DST can be used as a support in wood procurement planning. Aerial images are an appropriate data source in the proposed DST, mainly because of the readily availability and low cost.