565
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
Figure 4. Digital photographs of oat plot during the growing
season.
Besides, digital photographs and growth height from three
random places from each test plots were collected using tape
measure. Similarly, ground moisture measurements were
collected from five random places from each test plots using
moisture meter and Kotkaniemi meteorological station
observations were recorded for each scanning day.
Plots were threshed on the 16th August 2006 with a combined
harvester designed for harvesting trial plots. The whole plot was
threshed and total fresh weight of grains was weighed. A 1 kg
sample was taken for each plot to determine moisture content of
grains. Fresh grain weight was converted into grain yield value
(kg/ha) using grain moisture content and plot area. Measured
grain yield was reported at 15% moisture content.
4. RESULTS AND DISCUSSION
4.1 Initial works
Laser data was filtered using Faro Scene Software to remove
noise and points generated from both the ground and vegetation.
4.2 Growth height estimation of crops
Growth heights were determined from each test plot using laser
scanner data. A single test plot was divided into smaller grid
cells and growth heights were determined from each cell.
Growth height measures were compared to threshing results and
we found strong correlation between measured growth heights
and grain yield from each studied cultivars.
MEASURED VS SCANNED
Figure 5. Measured and scanned maximum growth heights of
Justina, Picolo and Belinda.
4.3 Ear detection
Ears of spring wheat cultivar Picolo were determined. An
algorithm was developed to automatically recognize ears and
estimate their size from laser scanner data. The algorithm was
based on the idea that point cloud was converted into voxel
model and voxels with enough laser points were marked.
Several marked voxels side by side in vertical direction were
associated as ear of wheat.
PICOLO
160N
120N
80N
40N
ON
Grain yield,
3556
2268
205
224
887
kg/ha
9
8
Estimated
2054
1133
123
115
709
ears
4
4
Table 1. Measured grain yield of Picolo at different N levels
and grain yields estimated using ear detection of scanner data.
Cultivar
Measured
height
Scanned
height
Estimated
ears
Picolo
0.93
0.93
0.97
Justina
0.90
0.95
0.96
Belinda
0.99
0.88
0.99
Mean
0.94
0.92
0.98
Table 2. Correlation coefficient between measured heights,
scanned heights, grain yields estimated using ear detection of
scanner data and precision harvesting. Correlations coefficients
are calculated using five different rates of fertilization.
Calculated ear size also correlates with the grain yield but the
problem was to find suitable parameters for the algorithm and
algorithm provides rather relative than absolute results of grain
yield. Detailed features were extracted from the voxels of ears
based on the idea that overlapping ear voxels contained
different amount of laser points and they provide more
information about the ear but this study was restricted by lack
of very detailed reference data.
5. CONCLUSIONS
In this study, terrestrial laser scanning was found to be a useful
tool for growth height and grain yield estimation. Growth
height of cereal plants was easy to estimate using laser scanner
data and estimated results correlates with tape measures.
Besides, ears of wheat were automatically recognized and their
size was estimated using laser scanner data.
Thus, our study shows that laser scanner could be used as
precision farming tool in agriculture. The scanner that we used
is not suitable for operational use but the similar methods can
be used for example to estimate data of laser scanner that is
mounted on a moving platform. Even if operational laser
scanning applications for agriculture seems not so relevant at
the time, it is worthwhile to study more this field because
development of instruments is fast and ongoing process.