Content uploaded by Dušan Jovanović
Author content
All content in this area was uploaded by Dušan Jovanović on Jul 14, 2015
Content may be subject to copyright.
Research Journal of Agricultural Science, 46 (3), 2014
32
REMOTE SENSING AS A TREND IN AGRICULTURE
D. JOVANOVIĆ, M. GOVEDARICA, D. RAŠIĆ
University of Novi Sad, Faculty of Technical Sciences, Trg Dositeja Obradovića 6, Novi Sad, Serbia
Nikole Tesle 36, Bukovac 21209; email address: dusanbuk@uns.ac.rs
Abstract: Remote sensing are widely used in the field of agriculture, forests, urbanism,
transport and other fields for better performance and results. The aim of this paper is to show small part
possibility of using remote sensing in the area of agriculture. Main part of this research is to show the
process of classification and mapping of agriculture crops using satellite images from Rapideye satellite
platform. These satellite images were used for the classification of three basic crops: corn, beet and
soybean. Satellite images and basic concepts and techniques of classification are presented in order to
emphasize benefits of classification of satellite imagery for presented classification methods.
Key words: Remote sensing, GIS, Agriculture, Classification, Rapideye
INTRODUCTION
Agricultural goods belong to the most important renewable, natural resources. It is
very important to have accurate and timely information about agricultural resources. Due to the
increased needs for agricultural products (as a side effect of population growth) there is a need
for improvement in management of agricultural resources, i.e. increasing crop yields. Because
of that, it is necessary to get valid data about crop types their quality, quantity and location of
these resources. Remote sensing techniques were, and have an upward trend to continue to be,
an important factor in the improvement of existing systems for data acquisition and data
processing in agriculture
Remote sensing is defined as the acquisition of data using a remotely located sensing
device, and the extraction of information from the data. The increasing use of remote sensing
in agriculture is visible in growing researches in crops (production, crop types, harvest, crop
yield), lands (land condition, parcel area, location), forest types, quality of water bodies, types
of irrigation systems etc.
Progress in agricultural production mostly depends on the improvement of existing
and development of new and appropriate technical solutions. There is a lot of emphasis on
environmental protection, production of quality, healthy food, without neglecting the reduction
of production costs with respect to profit and yield increase and improvement of working
conditions.
MATERIAL AND METHODS
RapidEye AG, Germany, successfully launched its own satellite system, satellite
constellation which is comprised of five identical satellites, in August 2008. RapidEye
constellation is shown in Figure 1. RapidEye was first commercial operative satellite with the
capability of large-area coverage, frequent revisit intervals, high resolution and multispectral
imagery. With the ability to produce 5m spatial resolution imagery in five spectral channels,
Rapideye represents an excellent source of data for the monitoring in ecology, environmental
protection, agriculture and forestry.
Research Journal of Agricultural Science, 46 (3), 2014
33
Figure 1 RapidEye constellation of five satellites
Each of five satellites is equipped with one JSS-56 camera, which weights
approximately 150 kg. A seven-year operational lifetime is predicted for RapidEye satellite
constellation. The primary applications of RapidEye imagery are agricultural mapping, growth
monitoring, damage assessment, yield prediction and cartography.
Each of the multispectral sensors collects data in five different wavelength channels:
blue, green, red, red edge and near infrared, as illustrated in Figure 2. RapidEye is the first
commercial sensor which the ability to collect data in the red-edge band, which is useful for
detecting changes in chlorophyll content. It could allow better estimation of the ground cover
and chlorophyll content of the vegetation.
The significance of red-edge band is that it provides measurements of changes and
differences in vegetation, providing the distinguishing and monitoring of vegetation spieces.
Figure 2 RapidEye wavelength bands
Rapideye wavelengths are given in the Table 1.below. Table 1.
RapidEye basic characteristics
Band No
Band
Wawelength
Spatial resolution
1
Blue
(440 – 510µm)
5m
2
Green
(520 – 590µm)
5m
3
Red
(630 – 685µm))
5m
Research Journal of Agricultural Science, 46 (3), 2014
34
4
Red-edge
(690 – 730µm)
5m
5
NIR
(760 – 850µm)
5m
Agricultural enterprise “Sava Kovačević” A.D. was founded in 1946 (Figure 3).
Today, the company possess 3.880ha of soil of its own property and 750ha of state soil, which
is taken as a permanent rent. Out of total arable land (4.600ha), 40% is under the irrigation
system, with a tendency of continuing growth. Industrial crops are presented in 40% whereas
grains and forage crops are presented in 55% of the area.
The classification was performed in satellite images from August in 2009. The most
abundant crops in that time period, in the agricultural area “Sava Kovačević”, were: corn, sugar
beet and soya bean, which was the reason why those crops were chosen for the classification.
These crops are presented in most agricultural areas in the Republic of Serbia, especially in this
time period (August), which confirms selected crops for the process of classification.
Furthermore, soil areas in which there were not any crops sowed in August, or harvest had
already been done, were mapped. Such soils were introduced as a background training sample,
as practice has shown that this provides better results of classification.
Figure 3 Agricultural enterprise “Sava Kovačević”
Supervised classification (classification with supervision) requires a priori (already
known) information such as what type of classes are needed to be recognized and extracted
(soil types, vegetation, or something else). Using this type of classification, in order to select
the appropriate samples for classification, it is necessary to have the information about pixels
which fall into some of the classes. In the supervised approach, the system must learn from a
training sample (TS): a collection of examples previously analysed and identified by a human
expert. Those training samples represent a set of pixels which represent what is identified,
chosen as a representative of the class. It is important to choose representative TS of the class
that is needed to be identified and isolate in a certain area. These TS was create by information
from the employees from Agricultural enterprise “Sava Kovačević” (Figure 3).
Research Journal of Agricultural Science, 46 (3), 2014
35
Supervised classification is quite accurate for mapping classes, but depends heavily on
the cognition and skills of the image specialist. The strategy is simple: the specialist must
recognize conventional or meaningful classes in a scene from prior knowledge, such as
personal experience with what is present in the scene, or more generally, the region it is located
in, by experience with thematic maps, or by on-site visits. As a rule, the classifying person
locates specific training sites on the image - either a print or a monitor display - to identify the
classes.
RESULTS AND DISCUSSIONS
Classification, as a way of automatic mapping of agricultural crops, represents one
way to get the information about exact situation in the field, without on-site (on-the-spot) visits
or at least to a lesser extent. It implies the analysis of multispectral satellite imagery and usage
of statistically based conclusion rules for determination of the land cover identity for every
pixel of the scene.
The area of interest for the agricultural enterprise “Sava Kovačević” was drawn over
the satellite image, with the difference of 52.46ha (2.87%), compared to the real situation in the
field. Only those parcels which were sown with the crops of interest (crops which are used in
classification) in time of the imagery acquisition were included into the area of interest.
For every chosen area of interest, first step in the classification was mapping and
classification of single crops using Maximum likelihood algorithm in Erdas Imagine software
tool. The choice of pixels of interest, as the representatives of the classes, was done randomly.
During this process, different parameters were set for Euclidean distance and the number of
observed pixels in that distance. The results of classification of four crops simultaneously
(sugar beet, corn, soya, land background), by using maximum 10% of sampled area of every
class, are shown in Table 2.
Table 2.
The results of classification of four crops simultaneously
Area
Sugar beet
Corn
Soya
Background
class (land)
Total
Classified area
[ha]
Truth area [ha]
203,49
339,89
272,59
1006,95
1822,92
1875,38
Classified area [ha]
196,68
322,75
271,74
1057,58
1848,75
Classification error
[%]
3.35
5.04
0.31
-5.03
-1.4
The results of Rapideye image classification are illustrated in the Figure 4 The
results of Rapideye image classificationshown below. There are visually visible the
results of individual classification of agricultural crops: Corn is maroon, sugar beet is blue,
soya is yellow and land are represented light brown, respectively. There is also visible the
simultaneous classification of all of these crops in the first part of the picture.
Expanding the training set and the classification of four crop types simultaneously,
instead of the classification of individual crops by four runs, increased the accuracy of the
results. An Red-edge Band (705 – 745nm) significantly helps in the analyses of vegetation
Research Journal of Agricultural Science, 46 (3), 2014
36
conditions; Red-edge band is sensitive to chlorophyll and carry important information about
the content of chlorophyll in vegetation. It helps in mapping, identification and detection of
vegetation age, health and vegetation types.
Red-edge band is the transition between red and NIR band of electromagnetic
spectrum. In healthy plants with a high content of chlorophyll, REP position (REP Red Edge
Position) moves to the NIR and when plants are suffering from any illness, the REP is moved
towards the green band of the electromagnetic spectrum. Red edge band in RapidEye satellite
was created in order to capture the dynamic of REP. It is important to understand the
phenomenon of plants in order to get more precise information by analysing the satellite
images.
Figure 4 The results of Rapideye image classification
CONCLUSIONS
The accuracy of over 90% which was achieved by processing of Rapideye image,
basically is because of radiometric resolution of 16-bit and red-edge band with all benefits that
come with red-edge band. Red-edge band has proved to be responsive to the amount of
vegetation biomass presented in a scene. It is obviously very useful for crop identification, for
distinguishing between crops and soil, and for seeing the boundaries of agricultural parcels.
Since there have been major new developments in satellite technology, RapidEye imagery (in
terms of information content) and all other satellite images with high spatial and spectral
resolution seems to be particularly suitable for feature detection and land cover mapping of
agriculture landscapes.
BIBLIOGRAPHY
1. KAVZOGLU, T., (2009): Increasing the accuracy of neural network classification using refined
training data. Environmental Modelling & Software 24 (7), 850– 858.
2. FOODY, G.M., (2001) : Thematic mapping from remotely sensed data with neural networks: MLP,
RBF and PNN based approaches. Journal of Geographical Systems, 3, 217–232.
3. FOODY, G.M., (2002): Status of land cover classification accuracy assessment. Remote Sensing of
Environment 80, 185–201.
Research Journal of Agricultural Science, 46 (3), 2014
37
4. JENSEN, J.R., (2005): Introductory Digital Image Processing: A Remote Sensing Perspective.
Pearson, Prentice Hall, USA.
5. MCCLOY K. R., (2006): Resource Management Information Systems: Remote Sensing, GIS and
Modelling. 2nd ed. Taylor & Francis Group, US, 575 pages.
6. LU, D., WENG, Q., (2007). A survey of image classification methods and techniques for improving
classification performance. International Journal of Remote Sensing 26 (5), 823–870.
7. http://www.rapideye.de/
8. http://www.savakovacevic.rs/sr/html/onama.html