=Paper=
{{Paper
|id=Vol-2030/HAICTA_2017_paper11
|storemode=property
|title=OLEA Framework for non Refined Olive Oil Traceability and Quality Assurance
|pdfUrl=https://ceur-ws.org/Vol-2030/HAICTA_2017_paper11.pdf
|volume=Vol-2030
|authors=Stavros Valsamidis,Dimitra Papaefthimiou,Antonios Ventouris,Irene-Maria Tabakis,Ioannis Kazanidis,Sotirios Kontogiannis
|dblpUrl=https://dblp.org/rec/conf/haicta/ValsamidisPVTKK17
}}
==OLEA Framework for non Refined Olive Oil Traceability and Quality Assurance==
<pdf width="1500px">https://ceur-ws.org/Vol-2030/HAICTA_2017_paper11.pdf</pdf>
<pre>
 OLEA Framework for non refined olive oil traceability
             and quality assurance

     Dimitra Papaefthimiou3, Antonios Ventouris1, Irene-Maria Tabakis1, Stavros
             Valsamidis2, Ioannis Kazanidis2 and Sotirios Kontogiannis1
 1
  University of Ioannina, Dept. of Mathematics, Laboratory of Distributed Micro-Computers
                                       Ioannina, Greece,
                  {aventoyris, irma.tabakis}@gmail.com, skontog@cc.uoi.gr
   2
     TEI of East Macedonia and Thrace, Dept. of Accounting and Finance, Kavala, Greece,
                                {svalsam, kazanidis}@teiemt.gr
3
  University of Ioannina, Dept. of Biological Applications and Technology, Ioannina, Greece,
                              dimitra.papaefthimiou@gmail.com



       Abstract. This paper proposes a framework for the monitoring of olive oil
       production chain (OLEA management system). The deployment of the system
       initiates at the olive tree fields, where NFC technology is used as part of
       OLEA system capability for both pesticides and fertilizers control, fungicides
       use and olive oil traceability. Additionally, OLEA system uses sensors for the
       procurement of quantitative and qualitative olive oil characteristics at the
       extraction industrial process. Such characteristics are pertained to the depth of
       olive cluster geographical location as initial system’s kick off and up to the
       identifiable tree point when the OLEA system will fully deploy. The paper
       presents also OLEA technical characteristics as well as the structure of its
       integrated database and middle-ware communication protocol, which will
       positively affect Greek olive oil industry product marketing and exports.
       Furthermore, an ongoing case study of the OLEA system used for olive oil
       quality monitoring and traceability purposes is presented.

       Keywords: Olea europaea, olive oil management system, olive oil monitoring
       and traceability system, NFC technology, web applications, IoT applications
       and protocols, data mining algorithms.



1 Introduction

Olive oil is a natural fruit juice with excellent nutritional characteristics. It is a
typical source food of the Mediterranean diet which has been associated with a low
incidence of cardiovascular diseases, neurological disorders, breast and colon
cancers, as well as with antioxidant properties. Moreover, an increase of interest in
olive oil as a healthy food has been observed lately in areas other than the
Mediterranean countries mainly because of its fatty acid composition and content of
other functional food components, such as polyphenols (Vekiari et al., 2010).
   Greece is the world's third largest olive oil producer after Spain and Italy,
according to industry sources. More than 80 percent of the Greek annual production




                                               91
of olive oil is extra virgin oil, while 90% of the Greek olive oil is exported to the EU
(Bettini and Sloop, 2014; Mylonas, 2015). 80% of Greece’s olive oil production is
located at the areas of Crete, Peloponnese and West Greece. Most olive mills in
Greece are small sized and less technologically advanced than those in Spain thus
leading to higher milling costs. Similar milling conditions occur in Italy, however,
despite the small productivity potential, production is closely integrated with the
olive farming stage (ICT monitoring processes), as well as the distribution stage (ICT
marketing processes). Such controlled monitoring processes allowed Italy to
overcome problems of the disseminated structure of olive oil partnerships and to
brand exceptional olive oil products that meet exportation requirements.
    In Greece, both the fragmented and technologically outdated coalitions’
mechanisms regarding standardization and product quality control increases further
the risks from improperly to not at all disposability of premium quality olive oil
(Mylonas, 2015). According to the 2015 World's reports for the 10 best olive oil
brands (WBOOR, 2015), Greek Messenia’s branded “Lia” oil is placed at the 9th
position bellow two Italian, Spanish and US brands. While US produces only 2% of
olive oil worldwide (84% is Europe), managed to pertain the biggest number of
unique olive oil brands in the global market (after Spain), due to its production
monitoring and marketing infrastructure.
    Olive oil organic cultivation is the form of agriculture that relies on techniques
such as crop rotation, green manure, olive compost and biological pest control to
maintain soil productivity and control pests on an olive oil field. Organic cultivation
excludes or strictly limits the use of artificial fertilizers, pesticides (which include
herbicides, insecticides and fungicides), plant growth regulators such as hormones,
livestock antibiotics, food additives, and genetically modified organisms (Camarsa et
al., 2010; Ehaliotis et al., 2011). In the latest years there are only a small number of
farmers in Greece that are modifying their cultivation types to organic production
following LIFE EU directives, and the whole cultivation policy remains unchanged
due to the lack of both contemporary cultivation methods and educational-training as
well as the strict state frameworks and cultivation monitoring methodologies.
    In Greece, olive pressing companies, coalitions and partnerships are more than
400 in total. The pressing systems still preferably used are the traditional grindstone
removing processes. A few more contemporary olive mills located in the areas of
Peloponnese use the two phase decanter centrifugation method for the olive
extraction. Other methods such as the three phase decanter, sinolea method or cold
extraction methods are not at all used (Niaounakis and Halvadakis, 2006).
Contemporary decanter systems reduce oil leakage from the traditionally milling
process up to 15-20%. Since two phase mode virgin oils had high oxidative stability
and better organoleptic characteristics (no correlation between their stability and
phenolic concentration as appeared in the 3 phase process), ends up that the two
phase machines are the ideal extraction equipment for producing olive oil of more
quantity and comparative quality to that of the old fashioned grindstone milling
process (Niaounakis and Halvadakis, 2006).
    Despite the immersive problems of monitoring and marketing olive oil products in
Greece, the future of Greek olive oil economy seems promising due to the globally
increased demands for the product. These demands are expected to grow by a 40-
50% more than the 2015 Greek exports until 2020 (210,000 tons – 2015)




                                            92
(International Olive council World, 2012; European Commission Agricultural
Market Briefs, 2012). The increasing export activity of olive oil products offers great
economic profits to a large number of people who are involved in the respective
supply chain. Even if reports from Greek authorities are non optimistic in respect to
the increase of Greek oil production, due to gradual reduction of EU CAP subsidies
and scattered small scale production, the use of ICT production monitoring and
traceability technologies shall increase product exportation which is already of high
quality standards (European Commission Agricultural Market Briefs, 2012).
   Focusing on Greece, the main olive varieties cultivated for drupes and oil are the
following: (Niaounakis and Halvadakis, 2006; European Commission Agricultural
Market Briefs, 2012) (a) Koroneiki: It represents 60% of Greece’s production, its
seed weight is 0.3-1gr and height 12-15mm. (b) Athinolia: an olive variety that
produces low acidity oil, has 2.2-2.9gr seed weight and height of 7.5-25mm. (c)
Tsounati (Mastolia, Mouratolia) (15% of Greece’s production): variety giving high
oil quantity with seed weight of 1.2gr and height of 10-16mm. (d) Kalamon-
Chalkidikis (ChondroElia): (1-2% of Greece’s production) used mainly as table or
paste oils. (e) Manaki (Lianolia, Koutsourelia): (10% of Greece’s production), high
altitude variety of medium oil characteristics (acidity and oil quantity).
   This paper presents a new olive oil monitoring and management framework called
OLEA. The OLEA framework is consisted of the following hierarchical steps: 1.
olive trees traceability and cultivation monitoring, 2. olive oil extraction sensory
monitoring and 3 olive oil management.
   The proposed OLEA framework and framework test-bed system implementation
serves a two-fold monitoring purpose: a) It will maintain regional olive tree
cultivation process information. Such information can be proven valuable and be
used in several statistical analyses and cultivation findings for the years to come. b) It
shall preserve regional olive oil characteristics. Such information is useful for the
quality assurance of promoted products as well as branding of new products.
   The remainder of this paper organization is as follows: Section 2 provides related
work of existing olive oil quality monitoring systems. Section 3, outlines the
proposed OLEA methodology and OLEA system high level architecture, while
section 4 focuses on an OLEA algorithm case study. Finally, section 5 concludes this
paper.



2 Olive Oil Monitoring Systems

In recent years, olive oil monitoring has been performed with the use of offline
sample analysis. However, offline sampling needs costly instruments, such as
accurate electrical balances, microscopes, automatic particle counters, and human
processes of long measure time, which precludes early diagnosis of oil system
failures and prevention.
   Focusing on commodity work automation, Gao et al. (2004) proposed a multiple
sensor system. Multiple integrated sensors could characterize the oil situation better
than a single sensor. With multiple sensors systems, reliability reflects the capability
of the sensor that could identify the oil system accurately. Oil characteristics can be




                                             93
measured online by single-function sensors. The oil measurement system is made up
of three parts: Multiple sensors for temperature, pressure, moisture and viscosity;
Signal acquisition channels with signal amplification, a diverter switch for analog
signals and an A/D converter; and finally the analysis and processing unit made up of
computer (industrial computer).
   Towards that direction, Taouil et al. (2008) proposed a video supervision system
used for quality monitoring in olive oil conditioning line. Oil monitoring system can
be used for achieving high production rates that cannot be done by human workers
easily.
   Taking into account product quality control as a major factor of quality assurance,
Mailer and Beckingham propose a sensory (organoleptic) system for testing. Sensory
quality is the most important test to ensure the oil is acceptable for consumption
(Mailer and Beckingham, 2006). The organic parts of olive oil tested are: Aroma,
flavour, pungency and bitterness.
   According to the IOC standards, extra virgin olive oil, for example, should not
have any sensory defects and should have some fruitiness. That means that the
definition of the sensory quality control shall include far more attributes than the
ones proposed by Mailer and Beckingham. Moreover, in order to achieve significant
quality standards, it is both required to develop the capability to preserve sensory
quality consistently as well as to have a reasonably good knowledge of regional olive
oil varieties attributes. Primarily, this would be an augmented knowledge and ability
to detect product defects and faults. At a more advanced level it could provide the
ability to identify oil for specific markets and the capability to blend oils to meet
customer specifications.
   Kiama et al. (2004) proposed a low-cost RFID - based palm monitoring system.
Passive RFID tags are used in the plantation field to uniquely identify each palm oil
tree and their Fresh Fruit Bunches (FFB) production is collected and monitored by
scanning the passive RFID tags using high frequency RFID scanners. This
technology aims to convert the harvest data into digital information which can be
processed and analyzed by PMS system and presented as informative outputs such as
dynamic charts. This system has only one level of access, top management users
(GIS Managers and Administrative Managers), as it mainly deals with crucial and
sensitive information. The overall system is made up of three main components
namely RFID system, Central PMS and the GIS system.



3 Proposed Methodology and Architecture of the OLEA System

Authors propose an olive oil monitoring and quality assurance methodology called
OLEA framework, supported by a test-bed system called OLEA system. OLEA
system methodology includes the OLEA clustering algorithm and the steps that the
OLEA system implements, while OLEA system architecture is presented in Fig.1
Below are the OLEA system technical implementation details according to
framework steps.




                                           94
Fig. 1. OLEA High level system architecture

The authors of this article propose the utilization of the OLEA framework for the
process of olive oil quality evaluation, traceability and branding. The proposed
framework includes the following methodology steps:
   1. Traceability step: OLEA product traceability regions are divided into two
distinctive layers:
   Layer 1: The olive field spatial geographical location GIS system maintained by
either a local prefecture or the nationwide OPEKEPE (Greece EU CAP funding
organization)(OPEKEPE, 2017). Each field location is divided into smaller polygon
areas (see Fig. 1) that pertain similar microclimate characteristics (Theodosiou et al.,
2012).
   Layer 2: Olive tree traceability, presented in Fig. 1 as rounded areas of tree
clusters. Such traceability is achieved with the use of passive NFC tags placed on
every tree each farmer claims for OPEKEPE funding. Apart from ID traces these tags
are also capable of recording yearly fertilization, pesticide activities (date of
appliance, type of medicine used, agriculturist approval for biological or non
biological cultivation) following an NDEF message formulation similar to the one
presented by Kontogiannis et al for the purpose of NDEF NFC tagging for the sheep
industry (Kontogiannis et al, 2016). From 2015 and on, the use of pesticides requires
agriculturist prescription and farmer certification, thus the recording of olive trees
can be performed in a more organised way (Minister of Agriculture, 2015).
   2. Olive oil sensory monitoring step: The OLEA system includes sensory-based
equipment (shown at Fig. 1) installed at the oil presses, for the measurement process
of the olive oil quantity and the quality clustering based on metrics. Due to its special
purpose and function, this could be nationally patented as well.
   3. Olive oil Management step:
   Logging and reporting: The OLEA database shall record olive oil production
quantity and quality attributes connected to cultivation characteristics of tree cluster
areas. The database shall also provide information for tree clusters regarding
quantities produced, their quality, and market prices. Such information attained
nation-wide could be proven very helpful for the development of the oil industry and
therefore stimulation of the national economy.




                                              95
   Olive oil clustering algorithm: The proposed OLEA clustering algorithm of olive
fields is based on the following attributes:
    1. Olive oil field location characteristics based on the OLEA Traceability step.
        That is, location characteristics acquired by: Protected Designation of Origin
        (PDO), the Protected Geographical Indication (PGI) or Protected Origin
        Production names or Protected origin names of Highest Quality – OPAP.
    2. Olive oil micro-climate area characteristics, agriculturist yearly reports of
        cultivation type as recorded by the NFC tags.
    3. Olive oil quantity and quality characteristics of olive oil recorded extraction
        by the sensory system
   The OLEA clustering algorithm is outlined in Figure 1. The per-variety OLEA
clustering process offers the capability to brand new products based on quality
reports and cultivation conditions as well as pertain accurate quantity availability for
marketing purposes, no matter how scattered the production and cultivation areas are.

    The methodology is performed as follows:
    First, a set of metric qualitative measures of the extracted olive oil characteristics
is evaluated per tree field. This filtering and evaluation process can be performed
separately from the clustering process and the evaluated metrics data are stored back
at the OLEA database.
    The second step of the OLEA algorithm includes a per olive tree field sector
identification based on OPEKEPEs’ spatial data and registration validity of the sector
stored at the OPEKEPEs’ database of recorded olive tree fields.
    After successful validation, the microclimate separation process initiates where
each field area is bounded to a specific microclimate area polygon accordingly (Fig.
1, areas A, B) (Zinas et al., 2013). Upon first layer cultivation area separation, a field
clustering process is undertaken for each area. That is, clusters of trees of a tree area
with maximum space of 10000 sq. m. Such areas are uniquely identified with the use
of an NFC tag, that keeps track of data regarding cultivation process, pesticides
used, and adjustment dates. Olive tree field NFC recorded metrics are the following:
    C1] NFC maintained metrics with the use of OLEA mobile phone
application: Metrics associated with the tree farming process, cultivation and
harvesting such as:
    1. Fertilization indexes and metrics (Kgr of biological or chemical fertilization/Ha
- quantity (gr) of borium, nitrogen, phosphorus, potassium and iron per tree per year,
etc.), 2.Soil fertilization periods and soil minerals composition from the last recorded
chemical analysis (B and other minerals like Ν, P, K and iron), 3. Pesticide
periodicity and chemical composition metrics and indexes (chemical sprayed (%ww,
%wv)/Kg over 1000 sq. m multiplied by the number of cultivation years (1
cultivation year equals to 6 months) operations), 4. Olive trees age and average
distance, 5. Certified biological cultivation or not.
For each cultivation tree-cluster location and microclimate, measures are kept at the
OLEA database and set/updated by the farmer’s mobile application or OLEA web
GUI. More specifically the microclimate measurements and predictions are
automatically updated with the use of agents at the Wunderground service
(Wunderground API, 2014).




                                             96
Fig. 2. OLEA clustering and branding algorithm

C2] Location and microclimate measures maintained by the farmer and OLEA
application service agents:
   1. Field altitude, 2. Field distance in Km from the sea, 3. Field solar monthly-
average irradiance (W/sq. m), monthly sunlight average day recordings, flux-
luminosity measurements etc, 4. Field soil moisture measurements, 5. Field monthly
average rainfall measurements, 6. Yearly critical meteorological conditions recorded
for occurring over the field (criticality and occurrence) such as hail or drought or
frost. Daily microclimate area predictions.
For recording the extracted olive oil quantity and chemical consistency (quality
attributes) per tree cluster, the proposed OLEA framework includes a multi-sensory
equipment placed similarly to the ones proposed by Scarafia (Scarafia, 2011), in the
olive fruit decanter equipment (see Fig. 1). This multi sensory equipment includes
sensors such as: 1. Gravity liquid level sensors for monitoring produced oil
quantities, 2. Refractivity UV sensors for K232, K270 measurements of UV
absorbance, 3. Colour sensors for monitoring colour constancy, and 4. Peroxide,
acidity sensors (see Fig. 1). These measurements are then uploaded in real-time to
the OLEA application service. The capture attributes from this automatic sensory
process are the following:
   C3] Olive oil attributes identified by sensors measured during the milling
process and uploaded to the OLEA application service:
   1. Olive size (Number of olives per Kg – 230 up to 320 olives/Kg in Europe, this
   is set as large to brilliant size), 2.Peroxide value (meq O2/Kg), 3. %v/v Oleic acid
   and free fatty acid, 4.Omega-3/6 polysaturated acids-rancidity index, 5. UV
   absorbance (K232, K270, DeltaK), 6. Harvesting age (days/hours of harvesting
   passed), 7. Colour characteristics., 8. Organoleptic classification based on the
   median of any defect (Md) and median of fruity attributes (Mf) (tests based on
   human sensory perception).
   Oxide rancidity protection, additives and refinement process metrics are not taken
into account by authors, since refinement, if it occurs, is a process that follows the oil
chemical characterization. Oil refinement is done by using chemicals that are harmful




                                             97
to humans. This means that the oil is treated with acid, or purified with an alkali, or
bleached. It can also be neutralized, filtered or deodorized. This paper focuses only
on virgin oils characterization with no intermediate refinement or additive process.
   After NFC tree cluster tagging and metrics evaluation of each clustering area the
clustering of areas into groups is performed based on previously mentioned metrics
(see: C1, C2, C3). Metrics are interval continuous values and/or ordinal discrete
values that can be treated as continuous. The clustering process that follows is
performed on metrics that are preferably normalized over similar value scales.
   Such clustering process is achieved using the of K-means algorithm. The process
is initialized with n=3 clusters and is repeated by incrementing n, until metric
requirements are satisfied by at least one of the generated clusters. Then, that cluster
which satisfies metric requirements and has the highest total score, according to
requirements is selected. That is, for all clusters, a total score value is calculated in
order for clusters to get ranked according to market requirements. This total score
value is the sum of all normalized qualitative metric values multiplied by a
probability index called clustering or metric importance coefficient.
   The selection process that follows selects the cluster of highest rank based on total
score and continues top down selection, until quantity requirements are met or a
cluster is reached where at least one metric value is bellow the requested value for
that metric.
   For example, let’s assume an OLEA clustering process based on three qualitative
metrics (m1, m2, m3) and a market requirement for metric mean values of U1, U2 and
U3. Then, based on market or export requirements, each metric value U1, U2, U3 is
assigned with a statically set probability of importance pi, !!!! 𝑝𝑖 = 1 , where l=3 is
the number of metrics used (in our case three metrics). Then a market required score
value based on probability of market importance is calculated according to Equation
1:
                      TS! = !!!! p! U! ,       !
                                               !!! p! = 1          (1)
   where l=3 is the number of metrics used. Similarly, for each k-means cluster a
total score metric value is calculated and compared to TSM value as shown in
Equation 2, where k is the cluster number, which the total score value corresponds to.
                  TS! = !!!! p! ' M!                                   (2)
K-means algorithm is used for the creation of product clusters where on each cluster
the mean metric values of M1, M2, M3 and cluster total score are calculated, based on
Equations 3 and 4, and then compared to the market required total score value:
                                          ! !!
                                                !
                                            !!! !
                                         !                  !    !!
                  TSk = 𝑝! ′𝑀! = ( ! ! !       !!      )(        !!! µ! )   (3)
                                        !!!!! !!! !!
                                                            !!

   where nk is the total cluster elements from the K-Means process, i are the number
of tree fields in each cluster (i=1..nk), and µi the qualitative metric value for each tree
field id that belongs to a cluster. In depth, Mk is the mean metric value per cluster.
Probability 𝑝! ′     is expressed as the normalized Mk value and it expresses the
cluster’s metric value impact on the other clusters. If no probabilities are used for
metric values, the Total score for cluster k is calculated based on Equation 4.
                                 !
                                !! !!
                                        !     !!
                  TSk = 𝑝! ′𝑀! =              !!! 𝑚!                        (4)
                                        !!




                                                 98
   The K-means clustering process initially starts with k=n=3 (where n is the number
of metric set as requirements). This leads to the creation of three clusters with mean
cluster values M1, M2 and M3 accordingly per cluster (Equation 3) and a total score
value per cluster (Equation 2).
   If there is at least one cluster with TSk<TSM, then the cluster with the maximum
olive oil quantity value that fulfils metric requirements is selected as appropriate
(That is, the cluster with the max(TSk), where TSk<TSM. If there is no cluster that
fulfils the requirement, then the clustering K-means process is performed repeatedly
increasing each time the number of clusters by one, until the market requirements for
at least one cluster mean values is met (if metric vector values (m1,m2,m3) are less
than (U1,U2,U3) vector values).
   Based on total score, a clusters’ ranking process is maintained and from that
ranking process the high order clusters in rank, are denoted as of high exportation
and branding value clusters (selection process) and are selected from top to bottom
until quantity requirements are met.



4 OLEA clustering algorithm proof of concept

The case study experimentation in this work includes the Manaki Greek olive variety
cultivation in the area of West Greece (fields at Preveza and Lefkada). In this study,
a total area of 120 square meters of Manaki oil tree cultivation is divided into 1000
square meters production fields. Each field is identified by a unique NFC tag id that
includes field's geographical location, fertilization methods used, altitude, etc. An
average fields tree distance is about 6.5-7m away thus 25-30 olive trees are included
in each field. The average NFC tagged sector fields production is about 1000-1200
Kgr/0.1 Ha.
   All fields are situated at the two nearby microclimate areas of Preveza and
Lefkada and the altitude variation between all is from 50 up to 250 meters. Two
metrics were used in this study in order to verify the proposed methodology: The
%v/v of Oleic Acidity and the total net weight of olive production per NFC tagged
field block set to the size of 1000 sq. m. All such blocks were marked by NFC tags
and square edge-bounded by GPS coordinates.

Table 1. k-means clusters, average quantity per cluster field in it and %v/v average acidity of
per k-means cluster. Cluster 2 meets market acidity requirements and quantity requested.

    Cluster ID     Average cluster tree field   Average cluster oil     Total Cluster oil
    and symbol:    oil quantity (lt)            acidity %v/v            quantity (lt)
    C1(○)              2033                        0.56                    10166
    C2(∆)              1000                        0.49                    11000
    C3(+)              1216                        0.61                    24322
    C4(x)              769                         0.59                    12304
    C5(◊)              661                         0.55                    32389
    C6(∇)              861                         0.56                    16359




                                                99
Fig. 3. Clusters returned from the OLEA algorithm. Cluster C2 is marked with the Δ symbol
and is the OLEA algorithm selected cluster.


   Based on the market olive oil need requests to the farmers’ co-partnership, the
EVOO (Extra Virgin Olive Oil) maximum acidity regulations and market
requirements regarding olive oil acidity were set into maximum of 0.5% v/v and the
market requested olive oil quantity is 7000 lt. For that purpose using our proposed
OLEA methodology a k-means process was performed for the selection of the fields
that produced olive oil less that the acidity threshold set. Table 1 shows the clusters
where the requirement of acidity is met (cluster C2).



5 Conclusions

This paper presents a new framework called OLEA, for the process of olive oil
attributes acquisition, traceability and management. OLEA methodology uses an
NFC based olive tree clustering scheme that maintains cultivation and environmental
attributes information stored in its database. This information is reinforced with
qualitative olive product characteristics derived by the olive fruit milling process.
The attributes recorded by OLEA are landscape similarity attributes; meteorological
characteristics; agricultural field operations used and finally milled product quantity
and product quality evaluation metrics.
   Authors outline the system architecture based on the OLEA methodology. This
system includes NFC tags for tree clusters identification and sensors for the process




                                            100
of product metric values measurement. OLEA system also includes an integrated
application and database service where all measured NFC and sensor data are
collected and then application level queries towards the database can be performed
using web interface or mobile phone application interface.
   Authors propose a new clustering algorithm, used for the process of maintaining
uniform olive product characteristics as well as branding new products. This
algorithms uses as input a customer’s product requirements in terms of threshold
values of the OLEA olive oil recorded attributes and delivers back to them a
selection of the most appropriate clusters areas to harvest. To support their
methodology, a case study scenario is presented, where the proposed clustering and
selection algorithm is applied. From the case study results, it is clearly shown how
the OLEA clustering algorithm can achieve the standardization of olive oil products
quality as well as branding of new products based on market needs. Furthermore, the
utilization of the proposed OLEA framework shall provide adequate olive oil
attribute records with regional environmental and cultivation methodology as well as
product branding and traceability.



References

1. Bettini, O. and Sloop, C. (2014) Olive Oil Annual 2014 Greece – GAIN reports,
   United Stated Department of Agriculture, Foreign Agricultural service,
   http://gain.fas.usda.gov/Recent%20GAIN%20Publications/Olive%20Oil%20Ann
   ual%202014_Rome_Greece_5-14-2014.pdf, Apr. 2014.
2. Camarsa G., Gardner S., Jones W., Eldridge J., Hudson T., Thorpe E. and O'Hara
   E. (2010) “LIFE among the olives: Good practice in improving environmental
   performance in the olive oil sector", EU Commision Environment Directorate-
   General,          http://ec.europa.eu/environment/life/publications/lifepublications/
   lifefocus/documents/oliveoil_en.pdf.
3. Ehaliotis, C. Martinis C, A. and Minotou Ch. (2011) Towards a paradigm for the
   convergence of production, marketing and environmental protection in the field
   of the organic olive oil production chain-Conclusions from the Biolmed project,
   In Proc. of Int. Conference on organic agriculture and agroteco tourism in the
   Mediterranean.
4. European Commission Agricultural Market Briefs, (2012) Prospects for the Olive
   oil sector in Spain, Italy and Greece – 2012-2020.
5. Gao H., Zhou X., Cheng H., Zhao C. and Yan X. (2004) Information Fusion of
   Online Oil Monitoring System Using multiple Sensors, Tsinghua Science and
   Technology, 9(3), pp. 258 - 261, June 2004.
6. International Olive council World, (2012) Prospects For The Olive Oil Sector -
   EU Agricultural Market Briefs- International Olive council World Table Olive
   Figures-Greece country Profile-Policy, http://www.internationaloliveoil.org.




                                           101
7. Kiama J. W., Raman V., Patrick T. H. H., (2014) Low-cost RFID-based palm oil
    monitoring system (PMS): First prototype, In 8th International Symposium of the
    Digital Earth (ISDE8 2014), Sarawak, Malaysia, vol. 18.
8. Kontogiannis S., Tsipouras M., Kokkonis G. and Valsamidis S. (2016), Design
    and implementation of an identification productivity recording and breeding
    system for the sheep industry, International Journal of Sustainable Agricultural
    Management and Informatics, Vol. 2, No. 2-4, pp. 97-122, ISSN 2054-5819.
9. Mailer R. and Beckingham C. (2006) Testing olive oil quality: chemical and
    sensory methods, Profitable & Sustainable Primary Industries, PRIMEFACT vol
    231, NSW Dept. of Primary Industries, ISSN 1832-6688, pp. 1-5.
10. Minister of Agriculture, (2015) National Plan of Action for the Rational Use of
    Agricultural Medicines, Greece, http://www.minagric.gr/images/stories/docs/
    agrotis/Georgika_Farmaka/elenxoi/NAP-GR_consolidated.pdf,                   ΚΥΑ
    6669/79087/15-7-2015.
11. Mylonas, P., (2015) Olive oil: Establishing the Greek brand, National Bank of
    Greece       sectoral     report”,    https://www.nbg.gr/greek/.../press-office/e-
    spot/reports/Documents/Olive%20Oil_2015.pdf, May 2015.
12. Niaounakis, M. and Halvadakis, CP.            (2006) Olive Processing Waste
    Management, Elsevier Literature Review and Patent Survey 2nd Edition, ISBN
    978-0-08-044851-0.
13. OPEKEPE, (2017) Greek Organization for EU CAP funding control on line web
    applications, http://www.opekepe.gr/applications.asp.
14. Scarafia L. (2011) A Perspective on Extra Virgin Olive Oil and Premium Quality:
    California             olive           oils,             Agrolab             Inc.,
    http://www.agbiolab.com/files/agbiolab_California-EVOO-Premium-Quality.pdf.
15. Taouil K., Chtourou Z., Kamoun L. (2008) Machine Vision Based Quality
    Monitoring in Olive Oil Conditioning, In Proc. of Image Processing Theory,
    Tools and Applications (IPTA 2008), Sousse, Tunisia, pp. 1 - 4.
16. Theodosiou T., Valsamidis S., Hatziliadis G. and Nikolaidis M. (2012)
    Measuring, archetyping and mining Olea europaea production data, Journal of
    Systems and Information Technology, 4(4), pp 318-335, Emerald.
17. Vekiari, S.A., Oreopoulou, V., Kourkoutas, Y., Kamoun, N., Sallem, M.,
    Psimouli, V., Arapoglou (2010) Characterization and seasonal variation of the
    quality of virgin olive oil of the Throumbolia and Koroneiki varieties from
    Southern Greece, Grasas y Aceites, 61 (3), pp. 221-231, 2010.
18. WBOOR,           (2015),      World’s       best      olive      oil      reports,
    http://www.bestoliveoils.com/page/1/.
19. Zinas N., Kontogiannis S., Kokkonis G., Pikridas C., (2013) A novel
    microclimate forecasting system architecture integrating GPS measurements and
    meteorological-sensor data, in Proc of 6th Balkan Conference in Informatics
    (BCI), Thessaloniki, Greece, In ACM vol 1, pp 82-88, ISBN: 978-1-4503-1851-
    8.




                                          102
20. Wunderground API, A Weather API designed           for   developers,
    https://www.wunderground.com/weather/api/, 2014.




                                  103

</pre>