This paper presents the results of an analysis of the accuracy of young deciduous tree height estimation based on a dense point cloud and high−resolution photographs acquired using an iPhone 14 Pro equipped with a viDOC RTK Rover antenna during mobile laser scanning. The photos were collected during manual vehicle driving at an altitude of 1.4 meters above ground in the Paryż Forest District belonging to the Płońsk Forest Inspectorate (Mazowieckie Voivodeship, Poland). Tree height was measured from a normalized dense point cloud (GSD 0.001 m/pix) using a moving window of approximately 0.005 m2. Ground measurements with a precision steel ruler (YT70715 YATO) served as reference data for accuracy and quality assessment. The ground control point measurements were conducted using the Pix4Dcatch Automatic Autotag Detection (AAD) with uniquely coded targets distributed across the forest plantation. A total of 900 two−years−old European beech Fagus sylvatica and 900 two−years−old pedunculate oak Quercus robur seedlings were measured: 100 trees in each of eighteen subsection for both species. The heterogeneous surface of the plantation resulted in a small camera optimization error, typically ≤1.0%. The Jensen−Shannon distance index was 0.0166 for beech and 0.0190 for pedunculate oak; the Wasserstein distance was 0.424 and 0.4725 for the same species, respectively.
Durło et al. (Thu,) studied this question.