Plant Leaf Classification Using Convolutional Neural Network

Plant classification systems, in general, could be a beneficial tool in the agricultural industry, especially when it comes to recognising plant types in a systematic and manageable manner. Previously, plant growers used to rely on observation and experienced personnel to distinguish between plant v...

Full description

Bibliographic Details
Published in:2022 8th International Conference on Control, Decision and Information Technologies, CoDIT 2022
Main Author: Othman N.A.; Damanhuri N.S.; Ali N.M.; Chiew Meng B.C.; Abd Samat A.A.
Format: Conference paper
Language:English
Published: Institute of Electrical and Electronics Engineers Inc. 2022
Online Access:https://www.scopus.com/inward/record.uri?eid=2-s2.0-85134298291&doi=10.1109%2fCoDIT55151.2022.9804121&partnerID=40&md5=f3c204389e85ee123204dab351c45e5e
Description
Summary:Plant classification systems, in general, could be a beneficial tool in the agricultural industry, especially when it comes to recognising plant types in a systematic and manageable manner. Previously, plant growers used to rely on observation and experienced personnel to distinguish between plant varieties. However, some plants, such as leaves and branches, have nearly identical traits, making identification difficult. Hence, there is a need for a system capable of resolving this issue. Thus, the focus of this research is on classifying plant leaves using a convolutional neural network (CNN) technique. Coriander and parsley were chosen as test subjects for this study because their leaves have comparable structures. The input image was subjected to a number of filter layers using CNN. A total of 100 coriander and parsley leaf photos are collected for this research. These photos were filtered using kernels. These kernels have a set size and extract features from the input photos to create a feature map. These extracted features will then be used to classify plant leaves according to its classes type. With the use of the Graphical User Interface (GUI), the end user will be able to determine the type of leaf. Results show that, using the ReLu activation layer with 15 layers of network design and a 70-30 training-testing proportion, this plant leaf classification system was able to attain a coriander and parsley classification accuracy of 90% with an error rate of 0.1. In addition, due to its great accuracy, this system can be extended for additional uses such as recognising plant diseases and species. © 2022 IEEE.
ISSN:
DOI:10.1109/CoDIT55151.2022.9804121