Automated assessment of hardwood and shrub competition in regenerating forests using leaf-off airborne imagery

Abstract

Forest regeneration assessment considers the abundance and condition of crop trees as well as the level of competing vegetation. Most remote sensing research has been conducted on conifer crop tree detection and assessment using imagery acquired in deciduous leaf-off conditions. Some research on competition assessment has been conducted using leaf-on imagery, but it is too costly and time consuming to require both leaf-off and leaf-on image acquisition and analysis for complete regeneration assessment. This paper evaluates the potential of automated methods for assessment of woody stem competition using very high-resolution (2 cm) leaf-off imagery. The intent is to couple the competition-evaluation methods with previously developed leaf-off conifer assessment methods. The automated method combined texture analysis, classification, and line detection. Results show that competition measures extracted from the processed imagery agreed well with estimates derived from both manual competition interpretation and field measurements. Manual estimates slightly outperformed automated extraction when compared against field measurements, but the best approach may be to combine them to optimize processing time and achieve the highest possible precision, particularly in areas where competition abundance is estimated to be close to a given silvicultural decision threshold.