Naslov
The influence of different orientation methods of UAV images on the quality of point cloud generation and plot-level tree height estimates

Autori
Jurjević, Luka ; Balenović, Ivan

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Book of Abstracts of "The 15th International Phytotechnology Conference" / Orlović, Saša - Novi Sad : Institute of Lowland Forestry and Environment (ILFE), University of Novi Sad, 2018, 225-225

ISBN
978-86-912323-9-9

Skup
The 15th International Phytotechnology Conference (15IPC-2018)

Mjesto i datum
Novi Sad, Srbija, 01.10.2018. - 05.10.2018

Vrsta sudjelovanja
Predavanje

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
digital photogrammetry ; forest inventory ; unmanned aerial vehicle

Sažetak
Digital photogrammetry ; forest inventory ; unmanned aerial vehicle

Izvorni jezik
Engleski

Znanstvena područja
Geodezija, Šumarstvo

Napomena
Unmanned Aerial Vehicles (UAVs) have the potential
to provide data of very high spatial and temporal
resolution, and therefore they recently attracted
great attention in many environmental sciences,
including forest inventory. By photogrammetric
processing of UAV-images, a dense point cloud of
forest canopy can be derived. To obtain normalized
point cloud (height above ground) useful for
forest inventory, a highly accurate Digital
Terrain Model (DTM) is needed. In combination with
field reference data, the normalized point clouds
could serve to estimate various forest inventory
attributes which has been confirmed by a number of
recent studies. However, due to a high variation
in the technical characteristics of UAVs which are
currently present in the market, and even more due
to a high diversity of forest structure across the
planet, further research is required. Namely, in
the dense forests, it is very difficult to obtain
the regular and dense spatial distribution of
Ground Control Points (GCPs) for quality Indirect
Sensor Orientation (ISO). Another option for
sensor orientation in such scenarios is GNSS
supported sensor orientation. Quality of the
orientation in that case depends directly on the
GNSS observations quality (Dual-
frequency/singlefrequency etc.). Although various
Sensor Orientation (SO) methods exist, they always
depend on the UAV's technical characteristics as
well as on the characteristics of the surveyed
area. This study aims to evaluate the influence of
different orientation methods on the quality of
point cloud generation and tree height estimates
at the plot-level. The first method of the image
orientation was ISO. Second and third methods of
image orientation were GNSS supported SO, where
second and third differed in the type of GNSS
observations. Second method used absolute code
GNSS measuremets and the third used the Post
Processed Kinematics (PPK) of dual-frequency GNSS
observations. The study was conducted in the
lowland pedunculate oak (Quercus robur L.) forests
of Pokupsko Basin located 35 km southwest of
Zagreb, Croatia. Based on field measurements
conducted in 2017 on a sample of 54 circular
plots, Lorey's mean height (hL) were calculated
for each plot. The UAV-images with the ground
sampling distance of ≈8 cm were acquired using the
Trimble UX5 HP with Sony Alpha ILCE-R7 in May
2017. From UAV-images orientated by three
different methods, three sets of dense point
clouds were generated using images of the level 1
image pyramid. Each one of the point clouds was
normalized with the available LiDAR DTM, then
clipped to the spatial extent of each field plot,
and a full suite of statistical metrics commonly
used as independent variables in forest attributes
modelling was extracted using the FUSION LDV
software. For each point cloud, a regression model
was used to estimate plot-level hL based on field
measurements and various metrics extracted from
point clouds. The cross-validation results show
that all three point clouds, estimate plot-level
hL with reasonable accuracy. Although such results
may be unexpected, especially for the first method
which provides the less accurate orientation
results, they can be explained by the homogeneity
of forest structure around the sample plots.

POVEZANOST RADA