Classification and analysis the maturity and class variety of chili fruit using Convolutional Neural Network (CNN)

Mohd Hussin, Najihah (2022) Classification and analysis the maturity and class variety of chili fruit using Convolutional Neural Network (CNN). Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Classification and analysis the maturity and class variety of chili fruit using Convolutional Neural Network (CNN).pdf - Submitted Version Download (6MB)

Abstract

Autonomous robots have recently grown in popularity in the agriculture industry, although they are still unable to do tasks as well as humans. A thorough procedure produces erroneous results, which raises the cost of production. This study presents an alternative way for farmers who need to sort the fruit categories such as chilis according to tehir maturities. This kind of innovation could be tackled by adopting an Artificial Intelligence (AI) approach such as deep learning. Farmers can save labour expenses while increasing fruit harvests by implementing an innovative chili identification method. A total of 1200 256 X 256 pixel chilli images are used, with 840 being used for training and the remaining 360 being served for testing. In this experiment, a learning rate of 0.0001 is used, with a mini batch size of 64 and 400 iterations. Convolutional Neural Network (CNN) model is applied to learn and recognize the chili fruits into three categories; unmatured, moderately mature, and mature. Chili fruit maturity is determined by measuring the distance between the calyx and the apex. In addition,the variety class of chili fruits is also being recognized according to its colors into green, red, and overripe chili. Using ADAM and SGDM optimizers with multiple CNN architectures, this study is capable of recognising and classifying chilli fruits with an accuracy of above 85%. We intend to expand the experiment with 3D images in the future, considering the depth information of the images in developing an autonomous agriculture robot.

Actions (login required)

View Item