| Summary: | Classification and identification of agricultural products are usually done
manually. This method is prone to error and subjectivity of the user. The other
method, laboratory testing, can only be done by professional and takes time.
Classification and identification of cane sugar in Indonesia is also done with
similar process with no standardization. In the production of cane sugar, several
stages and condition produce different kinds of sugar, resulting in the need of
supervision. In automation and standardization of quality, quantized identification
process needs to be done. System was designed as Artificial Neural Network with
one hidden layer using Levenberg-Marquardt algorithm. Colour and textural
features were extracted from 120 images of cane sugar for Artificial Neural
Network inputs. After feature reduction, the designed system could identify 8 kinds
of cane sugar with success rate of 85%. Designed system was also tested for
different learning rate, activating function, ratio of training and testing set and
different testing conditions. Application with GUI was also made for user.
|
|---|