IEEE Geoscience and Remote Sensing Magazine - September 2022 - 43

images highly depends on the data acquisition method.
The mobile approach requires the shortest time for data
acquisition, e.g., 0.5 s/image [96]. On the other hand, the
stop-and-go mode with a tripod in dense forests takes
much longer, e.g., 14 s/image, on average [95]. The image
processing time to a dense point cloud is difficult to
compare. It mainly depends on computer configurations
together with the software that is and its version. Additionally,
different settings for image matching and dense point
cloud generation are used in postprocessing workflows,
which highly impacts postprocessing time. It is worth noting
that postprocessing time is not a period where the operator
actively works with the software. In most postprocessing,
the operator is required only to start the process
and evaluate the results.
The proportional time needed to capture one tree in the
field when the plot-based approach is used varies among
studies. To calculate the time, the number of trees that
were successfully detected on a plot was used. The time
needed varied from 0.1 to 2.7 min. For single-tree studies,
the average time required to capture a single tree was
2 min when a fisheye lens was used and the DBH RMSE
was no higher than 0.5 cm [82]. Concerning the software,
Agisoft Metashape (formerly Agisoft Photoscan) was mostly
used across the reviewed studies, whereas a few studies
employed Pix4D. This could be because of cost: Pix4D is
more expensive.
IMAGERY
Besides the recent popularity of point cloud data, various
terrestrial imagery-type sensors have a long history in forest
in situ observations. Earlier studies using film cameras to
measure tree diameters date to the 1950s, and many systems
were proposed in the following years [121]. For terrestrial
imaging systems, key technical issues include camera
calibration, feature extraction, scaling, and protocols. In
particular, determining the scale among objects in an image
and the field is challenging, and a wide range of solutions
exists.
CONSUMER-GRADE CAMERAS
Individual trees can be measured at close to a professional
level by using a digital camera with supporting equipment
and particular protocols. The authors of [122] tested a laser
camera under typical boreal forest conditions, which
integrated a Canon EOS 400D digital reflex camera with
a Mitsubishi ML101J27 laser line generator. The laser camera
measured the diameters of trees from the center of a
sample plot without visiting individual trees. The method
semiautomatically processed the images and achieved accuracies
of 6 and 2.5 mm in terms of the standard error
and bias of diameter measurements, respectively [123]. An
imaging system was developed to determine the stem curve
of individual Scots pines in one image [124]. It consisted
of a calibrated camera, laser distance measurement device,
and calibration stick. Stem curves were determined by stem
SEPTEMBER 2022 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
diameters from an image and a tapering model. Two [125]
and four [121] photos were taken in orthogonal directions
to derive the stem curve. Measurement settings (i.e., the inclination
angle of the camera, viewing angle, and distance
measurements) were also recorded in the field for data interpretation
in [125].
PANORAMA IMAGES
Panorama images have been used to determine stem locations
in forest plots [126]. The positions were determined by
field-measured DBHs together with angle and DBH measurements
of each tree in an image; 85% of the measured
trees were within 0.5 m of the field-measured tree locations.
Lately, smartphones have been used in in situ observations,
following ideas similar to the studies mentioned in the preceding.
More discussion is in the " Operations " section.
HEMISPHERICAL IMAGERY
Hemispherical, or fisheye, photography has been used in
forest research since the 1950s [127]. It is the most widely
used ground‐based method for describing canopy characteristics
and forest light regimes [128]. Since digital cameras
and computer processes became available, it has been
increasingly employed to indirectly obtain foliation and
canopy architecture. Hemispherical imagery is an indirect
method to study light transmittance. It does not inherently
vary with time and cloud cover, though the associated error
level can occasionally be substantial [129]. The light transmittance
of the canopy can be described by the percentage
of the incident solar radiation at a given site compared to
the total incident solar radiation in the open through the
same period [130], which requires gap positions and site
geographical locations so that the sun track can be superimposed
onto the hemisphere.
Hemispherical imageries are also used to estimate other
plant canopy properties, e.g., to invert the leaf area index
(LAI). Hemispherical photography is markedly cheaper
than the alternatives [131]. However, the derived effective
plant area index may be significantly affected by the exposure
of the photographs, and the determination of the
declination angle of plant elements may be substantially
different from that derived from a canopy analyzer, as demonstrated
in [132]. For canopy structures, original images
are typically converted to black and white, and a threshold
is used to determine, e.g., the canopy cover and openness.
One of the main problems in applying hemispherical photography
to determine the LAI is the selection of an optimal
bright threshold to distinguish leaves from the sky [131].
The image quality also affects the features that are estimated;
e.g., overexposure may lead to overestimating the
sky fraction. In addition, the determination of the clumping
index was also suggested to be part of plant area index
measurements derived using optical and radiation methods
[132] since clumping seems to be one of the main factors
causing errors in LAI estimation [131]. Similarly, hemispherical
photography from smartphones has been used as
43
IEEE Geoscience and Remote Sensing Magazine - September 2022

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2022
