Page 1 - Faculty Researches
P. 1

INTERNATIONAL JOURNAL OF SCIENTIFIC & TECHNOLOGY RESEARCH VOLUME 9, ISSUE 04, APRIL  2020        ISSN 2277-8616


                Classification And Detection Of Nutritional

                Deficiencies In Coffee Plants Using Image


            Processing And Convolutional Neural Network

                                                       (CNN)


                                             Khenilyn P. Lewis, Juancho D. Espineli

        Abstract:  Coffee  farmers  or  growers  have  difficulty  in  classifying  nutritional  deficiencies  in  coffee  plants.  Proper  detection  of  these  nutritional
        deficiencies could help them in giving proper intervention to plants. The study was conducted to classify and detect the nutritional deficiencies in coffee
        plants  using  image  processing  and  Convolutional  Neural  Network  (CNN).  Once  the  nutritional  deficiency  is  identified,  the  prototype  will  display
        recommended fertilizer for the plant. One thousand images with eight nutritional deficiencies were used in the study namely, Boron (B), Calcium (Ca),
        Iron (Fe), Nitrogen (N), Phosphorus (P), Potassium (K), Magnesium (Mg) and Zinc (Z). The study covered the four varieties of coffee named Arabica,
        Robusta, Excelsa and Liberica existing in the Philippines. The capturing of images for testing and training the dataset were  done in coffee farms and
        nurseries in Cavite State University, National Coffee Research, Development and Extension Center (NCRDEC) and Amadeo, Cavite. Experimental and
        development research designs were  used.  Image processing  techniques was  utilized  in converting the  images  into grayscale and binary  values for
        threshold  and  segmentation.  Convolutional  Neural  Network  (CNN)  provides  the  predicted  nutritional  deficiencies  in  the  coffee  plants  through
        classification and detection. Results shows that CNN has a high accuracy in detecting and classifying the nutritional deficiencies in coffee plants. The
        prototype was evaluated, and results shows that it is an effective alternative for classifying and detecting the nutritional deficiencies in coffee plants.

        Index Terms: classification, coffee, convolutional neural network, detection, image processing, machine learning, nutritional deficiencies
                                            ——————————      ——————————

        1 INTRODUCTION                                         Adequate nutrients are essential to plants for growth. During
        Coffee is known as the most important crop commodity in the   the vegetation to pre-flowering of coffee plants, Nitrogen and
        world  [1].  People  around  the  world  drink  two  billion  cups  of   Potassium  promote  the  growth  of  tissues  in  new  plants.
        coffee every  day. A  total  of  25  million  of  families  depend  for   Calcium is for leaf growth that provides high yield for plants.
        coffee for living specifically for business or as growers. For the   Magnesium serve as fuel for developing tissues and Sulfur to
        last  15  years,  43  percent  of  coffee  consumption  has  arisen   maximize the growth of the plant through protein. During the
        worldwide [2]. The first coffee tree in Philippines was planted   post  flowering  to  berry  formation,  Nitrogen  and  Potassium
        in  Lipa,  Batangas  and  in  1860s  coffee  are  exported  to   serve  as  maintenance  for  plant  growth  and  support  berry
        America. Batangas reigned the coffee industry in the country   strength.  Calcium  is  for  strong  healthy  tissues.  Magnesium
        and  was  followed  by  Cavite  in  planting  coffee  in  1876.   and  micronutrients  are  for  growth  maintenance  and  berry
        However, due to insect manifestation coffee trees in Batangas   production. These nutrients are needed by the plants in their
        was  destroyed and  few  of  surviving  trees  was  transferred  to   berry expansion and berry maturity [9]. These nutrients are in
        Cavite.  As  of  2019,  Batangas  produces  13  percent  of  the   need  necessary  for  the  growth  of  coffee  plants.  A  healthy
        coffee  supply  in  CALABARZON  (Region  IV-A)  and  Cavite   coffee  plants  could  maximize  the  number  of  yields  that  may
        produces 67 percent [3]. The climate and soil condition in the   produce.  However,  identification  of  nutritional  deficiencies  is
        Philippines  are  suitable  for  planting  coffee.  Because  of  this,   done  manually  by  the  coffee  growers  or  experts.
        the country could produce the four varieties of coffee namely,   Characteristics and symptoms of plants in terms of nutritional
        Arabica  (Coffea  arabica),  Liberika  (Coffea  liberica),  Excelsa   deficiencies  are  usually  like  other  plants.  Coffee  growers  or
        and  Robusta  [5].  Arabica  is  the  most  expensive  variety  of   farmers should have enough knowledge to these symptoms so
        coffee and usually cultivated in high elevation areas. Liberika   that  they  could  perform  the  correct  interventions  [12].
        (Kapeng  Barako)  is  known  for  strong  flavor  and  aroma.   Numerous applications of digital image processing have been
        Excelsa  has bigger berries  compared  to Arabica.  Robusta  is   recorded  in  different  field.  Digital  image  processing  is  the
        being used for expresso and instant coffee mixes. Moreover,   manipulation  of  images  using  the  computer  [10].  It  converts
        Robusta  is  the  commonly  grown  variety  coffee  in  country   the physical images into corresponding images and extract the
        which has a total production of 69 percent in 2015. Followed   information  using  algorithms.  Digital  image  processing
        by  Arabica  which  contributed  24  percent  and  Excelsa  and   includes  image  collection,  image  processing  and  image
        Liberica [6].                                          analysis  [11].  This  study  was  conducted  to  utilize  an  image
                                                               processing  technique,  implement  Convolutional  Neural
                 ______________________________________        Network (CNN) and measure the effectiveness for classifying
                                                               and detecting nutritional deficiencies in coffee plants.
          •   Khenilyn  P.  Lewis  is  currently  pursuing  doctor  in  information
           technology in AMA University-Quezon City, Philippines. She is also   2 RELATED WORK
           working  as  a  faculty  member  in  Cavite  State  University,  Cavite,
           Philippines. E-mail: khenilyn@yahoo.com             This section discusses the related works in image processing
         •   Juancho D. Espineli is currently the dean of the school of graduate   and  algorithms  as  basis  for  classifying  and  detecting  the
           studies   in   AMA   University-   Quezon   City.   E-mail:   nutritional deficiencies in coffee plants. A neural network was
           jcespineli@gmail.com                                trained to detect the nutritional deficiencies in coffee plants. As
                                                               a  result,  a  Kappa  coefficient  of  0.96  for  Nitrogen  and
                                                                                                                 2076
                                                          IJSTR©2020
                                                          www.ijstr.org
   1   2   3   4   5   6