Coriander Classification System Using Machine Learning Techniques

摘要

In agricultural science, the use of imagery is of great utility. By means of images, diseases can be detected and identified at an early stage. Another way to use images is for sorting and grading. The need of this work is to create a classification system able to classify the quality of coriander as close as possible to the human eye. The problem in classifying the quality of coriander is that it requires experience and good vision. With an ageing population, this task may become quite tedious. To counterpart this problem, machine learning technique is used to create a coriander classification system. Convolutional neural network (CNN) is used as the machine learning technique as it supports the use of images and is able to treat images of three channel depths. To train the model, the need of a database is of upmost importance. Therefore, a database containing images of fresh coriander and rotten coriander was created. A classification model was created and after evaluation, the optimal accuracy reached 98%. Considering the relatively small database used, the model is considered to be very efficient.